Erasca, Inc. (NASDAQ: ERAS) — A Borrowed Molecule Priced Like an Owned Franchise

Report date: 2026-07-25 · Price: $19.11 (2026-07-24 close) · Market cap: ~$6.54B (pro forma) · Enterprise value: ~$5.7B Sector: Health Care · Biotechnology (Precision Oncology / RAS-MAPK) Coverage: Initiation of coverage

Sections 1–15 of this article carry no recommendation and no price target. The sole exception is the clearly-labeled Claude's Take block immediately below, which is the author’s own subjective opinion. This article is general information, not investment advice.

⚡ Claude’s Take

Claude’s own independent, subjective opinion. This is general information, not investment advice. The analytical body of this article (Sections 1–15) carries no position and contains no price target.

Verdict: AVOID here. Not a short. Revisit only on confirmed-response and duration data, or on a resolved IP position — not on price.

Tag: “A borrowed molecule priced like an owned franchise.”

Erasca has one thing that matters — ERAS-0015, a pan-RAS molecular glue with a genuinely striking Phase 1 signal (62% unconfirmed ORR at 8 weeks in 2L+ KRAS-G12X NSCLC; 14/14 patients clearing ≥75% of ctDNA variant allele frequency). That is real pharmacology, not noise, and I do not dismiss it. But the market is paying ~$5.7B of enterprise value for it, and three facts sit under that number that I think are underweighted. First, Erasca does not own the asset: ERAS-0015 is in-licensed from Guangzhou Joyo Pharmatech, and Erasca owns no issued composition-of-matter patent on it anywhere in the world — the single issued US patent in the family is Joyo’s, and Erasca is contractually barred from challenging it. Second, the category leader is attacking that licensed IP on two theories at once — patent infringement under the doctrine of equivalents, and trade-secret misappropriation reaching Erasca “as a licensee” — a defect which, if real, sits inside the licence and is not curable by Erasca’s own conduct. Third, the evidence is thinner than the price implies: every efficacy figure is an unconfirmed response rate at 8–14 weeks, pooled across two differently-designed trials in two countries, benchmarked entirely cross-trial against a competitor that now has randomized Phase 3 overall-survival data (mOS 13.2 vs. 6.7 months, HR 0.40) and an FDA-accepted NDA. There is no disclosed PFS, no OS, and no duration of response for ERAS-0015. Not one.

The framing is not “momentum” and not “falling knife” — the factor model is unusually clear that neither label is earned: Momentum, Value, Quality, LowVol and Growth all load at zero across every nested model, and ~86% of the variance is idiosyncratic. This is a pure binary that has already re-rated twice, and the tape has already agreed with the bull case: the stock is up ~18x from its April 2025 low, sits 9% above the July placement price, and trades at 14.8x book — the 97th percentile of its own history. I am not being paid to take the binary here. My discomfort is compounded by capital allocation: management paid $150M for a territory option that was priced at $50M before its own licensor tripped the trigger — a $100M unforced cost that walked guided runway back from 1H2029 to 2H2028 — and wrote off naporafenib, an asset taken all the way into Phase 3, for nothing. Against that, the CEO bought 2.3M shares of his own stock at the lows with real money, and killing five failing programs was the right call each time. This is a competent, honest team; it is not a business with a moat. If I wanted this exposure I would want it materially below the July placement price — call it roughly $10–14/share, an EV of about $2.5–4.0B — which is where the body’s own scenario work lands the base case and the probability-weighted value, and where the option is priced as an option rather than as a franchise.

Conviction: medium. Flips bullish on confirmed ORR with a durable median duration of response (approaching RMC-6236’s 15.1 months) in expansion cohorts plus a RevMed settlement or licence that removes the derivation overhang — or a big-pharma partnership that validates the IP through a third party’s diligence. Flips bearish on either a second treatment-related pneumonitis death at a go-forward dose, or RevMed converting the demand letter into a filed complaint seeking a preliminary injunction.

📈 Stock Price Action — Five-Year Event Map

Erasca has completed a full round-trip and then some in five years. It went public at $16.00 in July 2021, closed as high as $24.34 that September, collapsed 96% to an all-time closing low of $1.06 on 8 April 2025, then re-rated roughly 21x to a closing high of $22.47 on 23 April 2026 — before breaking 60% in four sessions. The stock closed 24 July 2026 at $19.11, inside a 52-week closing range of $1.39–$22.47, 15% below its April 2026 closing high and 21% below the September 2021 all-time high. The five-year compounded return is approximately zero (+1.3% annualized) against a 95.6% maximum drawdown. The telling detail: the stock trades 21% below its 2021 peak price at roughly 2.3x the market capitalization of that peak — the share count has nearly tripled.

1. The IPO priced at $16.00 and closed its first session at $17.43; the $24.34 closing peak came six weeks later — the entire premium was awarded before a single efficacy dataset existed. 2. The unwind was roughly two-thirds sector and one-third company. The 9 December 2022 session is instructive: Erasca in-licensed naporafenib from Novartis for $20M cash plus 12.3M shares at $6.50, while simultaneously selling 15.4M shares at that same $6.50 — an 11% discount to the prior close — for a −22.3% day. 3. The November 2023 prioritization, which bought runway by killing programs, cost a quarter of the equity in three sessions. 4. May 2024 was the inflection that actually mattered: the Joyo licence brought in what became ERAS-0015, and the market tolerated a 99.5M-share raise at $1.85 to fund it — but the stock still bled to $1.06 by April 2025. 5. January 2026 began as a sector event, not a company one: takeout speculation around Revolution Medicines re-rated every listed pan-RAS asset, and Erasca — the most levered pure-play — gained 42.4% on ~27x normal volume with no filing of its own. Its own data and J.P. Morgan week then carried the move, and management monetized it immediately with an upsized $258.8M raise at $10.00. 6. The run to $22.47 was pure readout anticipation. 7. The 27–28 April break is the defining event in the file. The efficacy in that release was strong on its face; the damage came from the safety table and the forward-looking-statements block — a death at a recommended dose, and the market leader demanding Erasca cease all US activity on its lead asset. That two-day repricing is now the measure of damages in a securities class action covering purchasers from 14 January 2025 to 26 April 2026. 8. The recovery has retraced roughly three-quarters of the break on the strength of the July dataset and a registrational path — and management again sold equity into it, the third raise in twenty-six months and the second in six.

Three material single-day moves (19 Mar 2026 +13.0%, 29 Jun 2026 +14.2%, and 29 May 2025 +31.8%) have no identified company disclosure and are deliberately left unattributed rather than assigned a speculative cause.

1. Executive Summary

Erasca is a clinical-stage, pre-revenue precision oncology company with 103 employees, singularly focused on the RAS/MAPK pathway. It has never earned a dollar of product revenue, has an accumulated deficit of $1.1B, and has raised roughly $1.8B since inception. At $19.11 it carries a market capitalization of about $6.54B and, against roughly $875M of pro-forma cash, an enterprise value of about $5.7B. That entire enterprise value rests on one asset in Phase 1 dose escalation.

The asset is real and the data are genuinely interesting. ERAS-0015, a pan-RAS molecular glue, produced a 62% unconfirmed objective response rate at 8 weeks in 2L+ KRAS-G12X non-small-cell lung cancer (N=37), 40% at 14 weeks in second-line pancreatic cancer (N=20), and — the most credible datapoint — ≥75% reduction in KRAS-G12X circulating-tumor-DNA variant allele frequency in 14 of 14 evaluable patients. It appears active at roughly one-eighth to one-tenth the dose of the class leader. No dose-limiting toxicities were observed and no patient discontinued for a treatment-related adverse event.

Three things sit under the valuation that we think are underweighted. First, Erasca does not own its lead asset. ERAS-0015 is in-licensed from Guangzhou Joyo Pharmatech; Erasca owns no issued composition-of-matter patent on it in any jurisdiction. The only issued US patent in the family belongs to Joyo, and the licence contractually forbids Erasca from challenging it. Second, the category leader is attacking that licensed IP on two theories simultaneously. On 24 April 2026 Revolution Medicines demanded Erasca cease all US making, using and selling of ERAS-0015, alleging infringement of US Patent 12,409,225 under the doctrine of equivalents and — more seriously — that a third party misappropriated RevMed trade secrets in connection with the ERAS-0015 patent, with Erasca liable as a licensee. If that theory has substance, the defect lies inside the licence and is not curable by Erasca’s own diligence. Third, the evidentiary standard is well below what the price implies. Every efficacy number is an unconfirmed response rate at 8–14 weeks, in cohorts of 2 to 37, pooled across two separately-designed trials in two countries, and benchmarked entirely by cross-trial comparison. There is no disclosed PFS, no OS, and no duration of response for ERAS-0015.

The competitive position is that of a fast follower, roughly four years behind. Revolution Medicines’ daraxonrasib — same cyclophilin-A tri-complex mechanism — reported a randomized Phase 3 in previously-treated metastatic pancreatic cancer with median overall survival of 13.2 months versus 6.7 for chemotherapy (HR 0.40), was published in the New England Journal of Medicine, and had its NDA accepted by the FDA on 22 July 2026 under the Commissioner’s National Priority Voucher pilot. RevMed has 883 employees and ~$1.9B of cash. Erasca has not begun a pivotal trial; its first Phase 3 is guided to start in 2027, meaning first pivotal data realistically in 2029 or later — in a world where a RAS inhibitor is already standard of care. Erasca’s own guidance concedes the point: it skipped second-line pancreatic entirely and went straight to first-line.

Capital allocation is the weakest link, with one genuinely indefensible episode. The Joyo licence priced a territory-expansion option at $50M if exercised before the first Phase 2 patient was dosed by either party, and $150M after. Erasca exercised in March 2026 — after its own licensor had tripped the trigger in China — and paid $150M for what had been available at $50M. That $100M walked guided runway back from “into 1H2029” to “into 2H2028” within six weeks of the earlier guidance. Separately, naporafenib — in-licensed from Novartis for $20M cash plus $80M of stock, advanced through Fast Track designation and into a pivotal Phase 3 — was abandoned and the licence terminated effective 3 June 2026, with nothing recovered. Every clinical program Erasca held from IPO through 2024 has now been terminated: five for five.

Against that, the honest counter-case. Killing failing programs quickly is a virtue, not a vice, and the naporafenib decision was probably correct even though it was expensive. CEO Jonathan Lim bought 2.3M shares on the open market for ~$4.7M of his own money at prices between $1.69 and $4.99, repeatedly and into weakness — genuine conviction, not optics. The management pedigree is real. The January and July 2026 raises, whatever else one thinks of the pattern, were opportunistic and well-timed after a decade of dilutive ones. And there is no going-concern issue: with ~$875M the company is funded into 2028.

The verdict. This is not a business with a competitive advantage; it is a single-asset call option on a molecule the company licensed rather than discovered, whose exclusivity it does not own and cannot defend by challenge, in a class that has already demonstrated — through the KRAS G12C precedent, where two triumphant first-in-class approvals produced a combined $572M of 2025 sales against a “millions of patients” narrative — that incidence-based market sizing over-predicts realized revenue by roughly an order of magnitude. The science may well work. The question this memo puts to the committee is whether ~$5.7B is the right price for a Phase 1 probability, and what the market must believe for it to be.

2. Business Overview

2.1 What the company is

Erasca, Inc. was incorporated in Delaware on 2 July 2018 and is headquartered at 3115 Merryfield Row, San Diego. It describes itself as a clinical-stage precision oncology company “singularly focused on discovering, developing, and commercializing therapies for patients with RAS/MAPK pathway-driven cancers.” As of 28 February 2026 it had 103 full-time employees — 70 in R&D and 33 in G&A, of whom 34 hold doctorates. It operates as a single reportable segment, and its chief operating decision maker is the CEO. [FACT — FY2025 10-K]

The company’s name is its thesis: the RAS/MAPK pathway is the most commonly dysregulated signaling cascade in human cancer, and Erasca’s stated approach is to shut it down via three strategies — targeting upstream and downstream signaling nodes, targeting RAS directly, and targeting the escape routes that emerge under treatment. It calls itself “modality-agnostic,” meaning it will use small molecules, antibodies or other formats depending on the target. [FACT — FY2025 10-K]

2.2 How it makes money — it does not

This is the single most important structural fact about Erasca and it should not be softened. The company has never generated any product revenue since inception in 2018. It states it “does not expect to generate any revenues from product sales unless and until [it] successfully completes development and obtains regulatory approval for any of its product candidates, which will not be for at least the next several years, if ever.” [FACT — Q1 2026 10-Q]

Conventional revenue segmentation is therefore not meaningful, and we will not manufacture it. The correct analog for a pre-revenue biotech is program segmentation by phase and indication, which we give in §2.3. Value can accrue to shareholders through exactly three channels:

1. Equity issuance — the only source of funding to date, and the mechanism by which the company converts scientific news into balance-sheet capacity. Erasca has raised roughly $1.8B since inception.
2. Partnering, out-licensing or milestone income — Erasca has signed none for its RAS franchise. It states it “own[s] or control[s] worldwide development and commercialization rights to [its] entire pipeline.” This is presented as a strength; it is equally a statement that no large pharmaceutical company has yet put capital behind the asset after diligence.
3. Eventual product launch — Erasca intends to self-commercialize in the US, and concedes it “currently ha[s] no sales, marketing, or commercial product distribution capabilities.” [FACT — FY2025 10-K]

The company also owns no manufacturing. It states it does “not own or operate, and currently ha[s] no plans to establish, any manufacturing facilities,” relying wholly on contract manufacturers for preclinical, clinical and any eventual commercial supply, and on third parties to package, label, store and distribute. [FACT — FY2025 10-K]

2.3 The pipeline as it actually stands

Two facts jump out of that table and both belong in the thesis.

First, the pipeline is two clinical assets and one discovery program. That is a narrow base for a $6.5B market capitalization, and ERAS-4001 has produced no human data at all.

Second — and this is the business-quality headline — every clinical program Erasca held from IPO through 2024 has been terminated. Five for five. The FY2021 and FY2022 10-Ks are dominated by ERAS-007, ERAS-601, ERAS-801 and ERAS-3490; the FY2025 10-K is dominated by ERAS-0015 and ERAS-4001. There is no overlap. The company has been through four separate “strategic prioritizations” in 39 months — February 2023, November 2023, May 2024 and May 2025 — each of which killed assets that had previously been described as differentiated. [FACT — 10-K series; 8-Ks of 2023-02-08, 2023-11-28, 2024-05-14, 2025-05-13]

2.4 The pipeline is entirely borrowed

Nothing Erasca discovered internally is in the clinic today. The company is explicit about this and frames it as a virtue: “we have in-licensed or acquired novel therapies from multiple geographic regions, including ERAS-0015, which we in-licensed from Joyo, ERAS-4001, which we in-licensed from Medshine Discovery Inc., and our ERAS-12 program, which we acquired from Emerge Life Sciences.” Management’s own words — “We pursue the best science in the world, regardless of its origin” — are an explicit statement that the differentiator is deal-sourcing, not proprietary chemistry. [FACT — FY2025 10-K]

That is a legitimate business model; Royalty Pharma and, in its way, Ignyta made it work. But it has direct consequences for the competitive assessment that follows: an in-licensing vehicle’s competitive advantage cannot rest on a research platform it does not have, and its exclusivity is only as good as the counterparty’s patent estate.

Licence economics:

The Joyo terms themselves are not onerous — a low-to-mid-single-digit royalty and roughly $182.5M of remaining milestones is light against a $6.5B market capitalization. The problem is not the price of the licence. It is the provenance and control of the underlying intellectual property, which we take up in §4.4.

2.5 Scale and capital position

At 31 March 2026 Erasca held $408.5M in cash, cash equivalents and marketable securities against an accumulated deficit of $1.1B. On 13 July 2026 it priced 31,428,572 shares at $17.50 for approximately $516.0M net (up to ~$593.5M if the 4,714,285-share underwriters’ option is exercised in full; that 30-day window remains open as of this report date and we do not assume exercise). The offering closed 15 July 2026, taking shares outstanding to 342,235,460 (346,949,745 if the option is exercised). [FACT — 10-Q; 424B5 of 2026-07-14]

At the 24 July 2026 close of $19.11 that is a market capitalization of approximately $6.54B, pro-forma cash of roughly $875M, and an enterprise value of approximately $5.7B. For context, enterprise value was approximately $109M at 30 June 2025 — a ~52x increase in thirteen months with no approved product, no revenue, and no pivotal trial started.

Verdict on the business overview: Erasca is a two-asset, 103-person in-licensing vehicle with no revenue, no manufacturing, no commercial infrastructure, no partner, and a demonstrated pattern of terminating everything it has ever put into the clinic. What it has is one molecule with a striking early signal, a very large cash balance, and a capable team. Whether that constitutes a business or an option is the subject of §4.

3. Industry Dynamics

3.1 The base rate is the most important number in this memo

Any assessment of Erasca must start from what the industry’s own statistics say about a Phase 1 solid-tumor asset. The authoritative dataset is BIO/Informa/QLS, Clinical Development Success Rates 2011–2020, covering 12,728 phase transitions across 9,704 programs:

A Phase 1 solid-tumor program has a 4.6% likelihood of reaching approval. Biomarker preselection — which ERAS-0015 has, since patients are selected by KRAS mutation — roughly doubles that, to about 16% for a generic biomarker-selected asset. Blending the two and allowing for the unusual strength of the early signal, a defensible range for ERAS-0015 is roughly 9–12%. Note also that likelihood of approval from Phase 3 in solid tumors is only 39.8% — even reaching a pivotal trial leaves most of the risk unresolved. Mean time from Phase 1 to approval is 10.5 years. Cost per approved asset is roughly $2.3B (Deloitte 2026) on a capitalized, failure-inclusive basis.

Two long-standing structural findings from the pharmaceutical-economics literature sharpen this further. First, on the economics conditional on approval: the average after-tax present value of US sales by decile of FDA-approved drugs runs $1,880M for the top decile, $701M for the second, $434M for the third and essentially nothing by the ninth — against an average after-tax R&D cost of roughly $800M including failures. The conclusion is stark: fewer than one in ten approved drugs recoups its cost of development. Approval is not the same as value. Second, on portfolio construction: a pipeline built entirely on novel mechanisms carries correlated mechanism risk — if the underlying biology disappoints, everything fails together. Erasca’s “singular focus” on one pathway is precisely this shape. (After-tax present value of US sales by decile of approved drugs: Vernon, Golec & DiMasi, Health Economics Letters, 2009; DiMasi & Grabowski, Managerial and Decision Economics 28, 2007.)

3.2 Sizing the RAS opportunity honestly

The RAS pitch is genuinely large: KRAS is mutated in roughly 25% of human cancers, ~90–98% of pancreatic adenocarcinoma, ~31% of lung adenocarcinoma and 40–50% of colorectal cancer. Erasca states an addressable population of ~2.7M RAS-mutant diagnoses globally per year.

That figure is not a market. The relevant population is metastatic, treatable, biomarker-positive patients, and in the US it is roughly:

At roughly $180,000 net annual price, six-to-ten-month treatment duration and 50% penetration, that supports approximately $3.9B of US and $7–8B of global revenue for a class-leading pan-RAS franchise, with a plausible global peak of $8–12B. [ASSUMPTION — our derivation from SEER 2026 incidence, published mutation frequencies, and analog pricing]

That is a real prize. But the sizing must be disciplined by what actually happened the last time this exact story was told.

3.3 The KRAS G12C precedent — the single most important disconfirming evidence

Two first-in-class KRAS G12C inhibitors were approved on the same “millions of patients” narrative. Their realized commercial performance:

A scientifically triumphant, genuinely first-in-class precision-oncology category produced $568M of global revenue in its fifth year. Lumakras is 1.0% of Amgen’s revenue — less than twenty-year-old Vectibix. Its FY2025 growth was +4% with volume +13% and net price −9%, with only two competitors in the market. NICE is removing it from the UK Cancer Drugs Fund in 2026. Bristol Myers Squibb paid $4.8B ($5.8B with the CVR) for Mirati; the CVR has not paid.

The causes are instructive and transfer directly to pan-RAS: the addressable G12C NSCLC pool is only ~7,000 US patients; the confirmatory CodeBreaK 200 trial delivered just +1.1 months of PFS and no overall-survival benefit; resistance emerges at around six months; and price eroded despite minimal competition.

And the category is now being obsoleted. On 1–2 July 2026 Roche’s divarasib beat both sotorasib and adagrasib head-to-head in a randomized Phase 3 (KRASCENDO-1, N=338) on both PFS and OS — the only head-to-head trial ever run in the class. First-mover advantage in a druggable oncology target proved to be about five years and under $1.5B of cumulative revenue.

Incidence-based TAM arithmetic has over-predicted realized revenue in this exact target class by roughly an order of magnitude. Any pan-RAS forecast that does not confront this is not a forecast.

3.4 Competitive intensity: crowded, and accelerating

There are approximately 106 SHP2/SOS1/KRAS-directed agents in development. The reference competitor, Revolution Medicines, carries a market capitalization of roughly $40.2B (212,596,462 shares per its 10-Q cover at 1 May 2026, at $189.23 on 24 July 2026), holds over $3.9B of cash after an April raise, guides 2026 operating expense of $1.7–1.8B, and grew headcount from 534 to 883. Its daraxonrasib NDA is filed and accepted with an FDA National Priority Voucher; zoldonrasib is in three Phase 3s. Behind it sit Amgen, Bristol Myers Squibb, Roche (currently winning the G12C fight), Eli Lilly, Boehringer Ingelheim, Novartis, BridgeBio Oncology, plus at least four approved G12C agents in China and entrants including Jacobio and GenFleet. RAS-directed business development in 2026 alone includes AstraZeneca/Jacobio (~$2.0B), AbbVie/Kestrel ($1.45B) and Johnson & Johnson/Firefly (~$1B).

This crowding pattern is not unique to RAS; it is the industry’s standard response to a validated target. Hepatitis C direct-acting antivirals went from about 2 candidates to about 30 between 2005 and 2010; JAK inhibitors went from 1 to 9 over the same period. The durable lesson, well established in the pharmaceutical strategy literature and vindicated by G12C: “being first in class is often short lived; once a company is seen as having potential in an untapped disease, several other biotech companies usually follow its lead… First-in-class is often more of a short-term win, as share is usually lost quickly when a better drug comes along.” The same source offers a formula worth holding onto — success requires meaningful disease impact, which is mandatory, plus best-in-class or first-in-class status; first-in-class alone is insufficient.

Being second or third in this field is worth very little, and that proposition has already been tested inside this very target class.

3.5 Regulation

Project Optimus (final FDA guidance, August 2024) now requires randomized multi-dose comparison before registration, replacing the old maximum-tolerated-dose paradigm. This bites hardest where a class has on-target wild-type toxicity — daraxonrasib carries 81–91% rash, 8% grade 3+, and 30% dose reductions. Erasca’s dual 24/32 mg recommended-dose-for-expansion structure reads as Optimus compliance rather than a tolerability ceiling. We note explicitly that ERAS-0015 has reported no dose-limiting toxicities, 100% median relative dose intensity, and zero treatment-related discontinuations; the Grade 5 pneumonitis was a serious adverse event, not a formal DLT, and the two should not be conflated.

The FDA in 2026 is faster but less predictable. A one-trial approval default has been articulated by agency leadership; the Commissioner’s National Priority Voucher pilot has produced oncology approvals in 44 and 55 days. These mechanisms are discretionary rather than statutory, and discretion tends to favor the incumbent who already holds the voucher — which, in this class, is Revolution Medicines.

The IRA “pill penalty” is a structural negative specific to this modality. Small molecules become eligible for Medicare price negotiation nine years after approval versus thirteen for biologics. The EPIC Act, which would equalize the two, has been reintroduced but not enacted. This is not theoretical for oncology: Imbruvica entered negotiation in cycle 1, Ibrance in cycle 2, and Verzenio and Kisqali were announced for cycle 3 in January 2026, effective 1 January 2028. Median age at diagnosis across pancreatic, lung and colorectal cancer is roughly 66–71, making these populations Medicare-heavy. Any valuation that assumes a thirteen-year commercial runway for ERAS-0015 overstates terminal value by roughly four years of peak cash flow.

3.6 The capital cycle — textbook late boom

Applying Marathon’s supply-side lens: biotech went through a genuine bust from 2021 to 2025. The XBI drew down 63.9%; the count of public US and European biotechs fell from 977 to 758; 39% of the sector had under a year of cash in 2024; biotech venture funding fell from $30.8B to $11.7B.

The 2026 turn has been violent. XBI is up roughly 70% over twelve months. There has been $106B of M&A across 201 deals year-to-date. Fifteen IPOs priced in the first half of 2026 versus seven in the first half of 2025, at a median raise of $287.5M — the highest since 2021. Over $5.5B of RAS-specific capital has been raised in 2026 alone. Erasca itself raised at $10.00 in January and $17.50 in July.

The demand driver is real: a $236B patent cliff between 2025 and 2030 (Keytruda alone is $29.5B, 56% of Merck’s revenue, losing exclusivity in 2028) against roughly $180B of pharmaceutical dry powder. Take-out premiums run 30–60% for late-stage assets and 50–150% for exceptional Phase 2 data.

But the capital-cycle reading is unambiguous: this is the boom phase, and the capital arriving now will compete for the same 70,000–80,000 US patients in 2030–2033. The G12C sub-cycle ran the complete boom → entry → obsolescence loop in seven years. The asset-growth anomaly — heavy equity issuance predicting low subsequent returns — applies directly to a company that has roughly doubled its capital base in two quarters.

One important qualification on the M&A thesis, drawn from the academic literature: in the empirical data, being acquired is correlated with weakness, not strength. Among small pharmaceutical firms, low Tobin’s q, few marketed products and low cash predict being acquired; each marketed drug reduces the probability of being a target by about 5 percentage points, and financially strong firms are more likely to do nothing at all. The unconditional base rate of being acquired is roughly 2% per firm-year. “Takeout optionality” is a real but low-probability, weakness-correlated event, and should not be underwritten as a floor. (Danzon, Epstein & Nicholson, Mergers and Acquisitions in the Pharmaceutical and Biotech Industries, NBER Working Paper 10536, 2004.)

3.7 Barriers to entry at the industry level

Applying Greenwald’s tests to the industry rather than the company:

• Market-share stability fails catastrophically. Lumakras went from 100% of class revenue to roughly 64% in four years and is now being displaced by the third entrant. Greenwald’s threshold for “no barriers to entry” is share movement above five points over five to eight years; this is roughly 36 points in four.
• Returns on capital fail. Top-20 pharmaceutical late-stage R&D internal rate of return was 7.0% in 2025 — at a cyclical peak — and 38% of projected inflows are GLP-1s, i.e. the headline number is flattered by one unrelated category. Against a 15–25% hurdle, that is value destruction in aggregate.
• Source of advantage. Supply/cost advantage reduces to a single molecule’s patent — the weakest and most transient barrier in Greenwald’s hierarchy. Demand captivity is zero: oncologists re-optimize to the newest NCCN-listed randomized data at no switching cost. Scale is absent in discovery — a 500-person biotech beat both Amgen and Roche to a validated target — and present only in commercial capture, which a clinical-stage company does not own.

3.8 Verdict

Structurally bad — one of the worst industries by every test we apply. A 4.6% Phase 1 likelihood of approval in solid tumors; approximately $2.3B of near-certain cost over 10.5 years against contingent revenue now statutorily truncated at nine years by the IRA; no durable barriers to entry; and a realized commercial outcome for a successful precision-oncology class of $568M globally in year five against an incidence-based TAM of millions.

The disconfirming evidence, weighed fairly. The prize is genuinely large — $8–12B of global peak revenue for a class leader — and RASolute 302’s hazard ratio of 0.40 validates the underlying biology in a way that was not established two years ago. Biomarker selection really does roughly double the odds. And critically, the exit mechanism partially rescues shareholder economics: with $236B coming off patent and $180B of dry powder, a clinical-stage company needs one convincing randomized dataset and a buyer — not a franchise, not a sales force, not scale.

That is the honest conclusion, and it is a narrow one. This is an industry in which the profit pool accrues to the large-pharma acquirer, who has commercial scale and diversification across many draws, and to the seller who happens to hold one of the ~10% of programs that work. The other ~90% is capital destroyed. A clinical-stage biotech is therefore a tradeable option on a binary event, not a good operating business — and it should be valued, and sized, as such.

4. Competitive Position

4.1 The data behind the “best-in-class” claim

Erasca’s entire competitive argument rests on one dataset. It deserves to be stated in full, fairly, before it is stress-tested.

From the 27 April 2026 release (data cut-off 4 April 2026 US / 27 February 2026 China, pooling AURORAS-1 and Joyo’s China trial JYP0015M101), at pharmacologically active doses of 16–32 mg:

From the 13 July 2026 update (cut-off 25 May 2026): 57% uORR at 8 weeks (N=7) in 2L+ KRAS-G12X PDAC at the 32 mg dose; safety unchanged; median relative dose intensity 100% at both 24 and 32 mg; rash 72% all-grade (N=72), diarrhea 32%; no dose-limiting toxicities, zero discontinuations for treatment-related adverse events. The panitumumab combination cleared its 16 mg cohort with no DLTs.

The signal is genuinely strong for dose escalation, and the ctDNA data are the most credible element — 14 of 14 patients clearing at least 75% of variant allele frequency is pharmacology, not statistical noise. Activity at 8 mg, roughly one-eighth to one-tenth of daraxonrasib’s dose, is a real observation.

But the evidentiary quality is well below what a $5.7B enterprise value implies, and this must be said plainly:

• Every response figure is an unconfirmed ORR at 8 or 14 weeks. Unconfirmed responses regularly fail to confirm.
• Cohort sizes are 2 to 37. The 57% headline is N=7 — four responders. A single reclassification moves it 14 points.
• Results are pooled across two separately designed trials in two countries, one of them run by a licensor whose operations Erasca concedes in its own risk factors it “did not control.”
• All benchmarking is cross-trial, against a competitor that has since produced randomized Phase 3 overall-survival data.
• There is no disclosed progression-free survival, no overall survival, and no duration of response for ERAS-0015. None. Duration is the variable that determines commercial value in this class, and it is entirely unmeasured.

4.2 Where Erasca actually ranks

The benchmark figures Erasca itself cites for daraxonrasib: 2L/3L RAS-G12X NSCLC ORR 38% (N=40), median duration of response 15.1 months, mPFS 9.8 months, mOS 17.7 months; 2L PDAC ORR ~29% rising to ~35%. Erasca’s claimed edges — 62% vs. 38%, 40% vs. 29% — are unconfirmed-response, short-follow-up, cross-trial comparisons against a molecule with randomized Phase 3 survival data.

The scale gap is not close. Revolution Medicines: 883 employees, over $3.9B of cash, 2026 operating expense guidance of $1.7–1.8B, ~$40.2B market capitalization. Erasca: 103 employees, ~$896M of cash, roughly $100M of annual burn stepping up toward $150–180M, ~$6.5B market capitalization.

One datapoint deserves particular weight because it is a revealed preference rather than a claim. Amgen — which owns an approved KRAS inhibitor of its own — disclosed in February 2026 that it is combining its PRMT5 inhibitor AMG 193 with Revolution Medicines’ RAS inhibitor in pancreatic cancer. The incumbent with its own franchise chose the competitor’s molecule as the backbone. In the same commentary Amgen conceded a class-level immuno-oncology combinability wall (its first-line NSCLC trial is restricted to PD-L1-negative patients) and the need for vertical combination with its own anti-EGFR agent — both of which are class limitations that apply to ERAS-0015 as much as to sotorasib. Notably, Amgen volunteered specific revenue figures for other oncology products across two investor events while giving none for Lumakras. (Amgen investor events, February 2026.)

Two calibration points from the wider RAS and precision-oncology literature are worth carrying over. First, KRAS G12C — the entire commercial track record of this target class — represents only about 13% of KRAS mutations, so the $568M G12C outcome was achieved on a comparatively small slice of the opportunity; that cuts in Erasca’s favour on market size and against it on the reliability of incidence-based sizing. Second, and less favourably: a reasonable probability of success for daraxonrasib after its positive Phase 3 sits in the 65–75% range. Any base case for ERAS-0015, working from an N=7 Phase 1 readout, must be calibrated dramatically below that.

Verdict on rank: a fast follower, second at best in pan-RAS and mid-pack in pan-KRAS. ERAS-4001 has produced no human data at all and meets BridgeBio and Boehringer on the same 2H2026 timeline. The only dimension on which Erasca leads is a dose-potency claim and an unconfirmed cross-trial response rate.

We also flag a contradiction between management narrative and evidence. The FY2025 10-K frames the competitive landscape as “Coopetition” — “we view other companies in this sector more as potential allies and collaborators than as competitors.” That framing is directly contradicted by the market leader demanding Erasca cease all US activity on its lead asset. We resolve in favor of the evidence.

4.3 Is there a moat? Applying Greenwald

Market-share stability: not meaningful (no revenue). The correct pre-revenue analogs all fail or are unproven: share of development lead time (roughly four years behind); share of the finite biomarker-selected trial-eligible patient pool and of KOL and site capacity (held by RevMed); and program survivorship — Erasca’s realized portfolio stability is 0%: five clinical programs in, five terminated.

ROIC: not meaningful (no revenue; TTM operating loss $293M; negative book capital). The correct analog is realized return on in-licensing and R&D capital, and it is unambiguously negative: roughly $1.1B of accumulated deficit has produced zero surviving assets from the 2018–2024 vintage. The historical return on Erasca’s deployed capital is approximately −100%, and the entire enterprise value rests on molecules in-licensed twenty-six months ago.

Scarce talent is real but is not a moat — Greenwald is explicit that star talent is owned by the talent, not the firm. The pedigree is genuine: CEO Jonathan Lim built Ignyta and sold it to Roche; Michael Varney ran Genentech’s research organization. But note an awkward fact: Dr. Kevan Shokat co-founded both Erasca and Revolution Medicines. Shared founder science is definitionally not a proprietary advantage, and it is uncomfortable context for a trade-secret derivation allegation.

Network effects: none. The nearest analog — first-mover lock-up of key opinion leaders, trial sites and the scarce biomarker-selected patient pool — exists, and belongs to the competitor.

4.4 Intellectual property: the load-bearing weakness

This is where the thesis concentrates, and it is documented rather than inferred.

Erasca’s entire patent estate at 31 December 2025, across all programs combined, is one issued US patent and one issued foreign patent, plus 4 US provisionals, 6 US non-provisionals, 11 PCTs and 39 foreign applications. That is the whole estate behind a $6.5B market capitalization.

For ERAS-0015 specifically, the composition-of-matter family is in-licensed from Joyo and contains that sole issued US patent, expiring 2043 absent extensions. Erasca’s own and co-owned ERAS-0015 families are all still applications. Erasca owns no issued composition-of-matter patent on its lead asset anywhere in the world. For ERAS-4001, there is no issued patent anywhere. In the standard hierarchy of pharmaceutical patent robustness — composition of matter above mechanism of action above method of use above formulation — Erasca does not own the top tier on either clinical asset.

The attack. On 24 April 2026 Erasca received a letter from counsel for Revolution Medicines alleging that (1) ERAS-0015 is “substantially equivalent” to compositions claimed in US Patent No. 12,409,225 and infringes under the doctrine of equivalents; (2) a third party misappropriated RevMed trade secrets in connection with a patent relating to ERAS-0015, and Erasca is liable “as a licensee”; and (3) Erasca’s public preclinical comparisons against RMC-6236 are deceptive. RevMed demanded Erasca cease all US making, using, offering, selling and importing of ERAS-0015 outside the Hatch-Waxman safe harbor. Erasca calls the claims “without merit” and intends to contest them vigorously, while conceding in its Q1 risk factors that remedies “could include injunctive relief or monetary damages” and that it “may be required to obtain licenses, modify our products or processes, or otherwise adjust our development or commercialization plans.”

We read the licence itself, and it materially worsens the picture. The Joyo License Agreement of 15 May 2024, filed as Exhibit 10.1 to the Q2 2024 10-Q, provides at Section 11.7 that Joyo “EXPRESSLY DISCLAIMS ANY … WARRANTY AS TO THE VALIDITY OF ANY PATENTS OR THE NON-INFRINGEMENT OF ANY INTELLECTUAL PROPERTY RIGHTS OF THIRD PARTIES”; that the licensed IP is provided “‘AS IS’, ‘WITH ALL FAULTS’, AND ‘WITH ALL DEFECTS’”; that Joyo “WILL HAVE NO LIABILITY TO Erasca FOR ANY ACT OR OMISSION IN THE … PROSECUTION, MAINTENANCE, ENFORCEMENT, DEFENSE OR OTHER HANDLING OF THE LICENSED PATENTs”; and that Erasca “ASSUMES ALL RISK AND LIABILITY” for determining whether the licensed patents are usable.

Joyo’s indemnity at Section 12.2 runs only to its own activities in its own territory and to breach of its own representations. There is no general third-party-IP-infringement indemnity running from Joyo to Erasca. Erasca’s only recourse is breach of Joyo’s representations — of which the relevant one, that no facts existed that would form a reasonable basis for an infringement or misappropriation claim, is knowledge-qualified, subject to an exclusion of consequential damages, enforceable only in binding arbitration, and collectible only against a private Chinese counterparty.

Two further structural constraints compound this. First, the licence permits termination if Erasca or any affiliate or sublicensee challenges the licensed patents — Erasca is contractually barred from attacking the validity of the very patents on which its asset depends. Second, the agreement contemplates that some licensed patents may themselves be in-licensed by Joyo from a further upstream third party, whose identity is redacted from the public filing. RevMed’s theory names an unnamed “third party”; Erasca may sit two or more steps from the origin of the IP it must defend.

The patent chronology explains the shape of RevMed’s claim. RevMed’s US 12,409,225 (“Ras inhibitors,” inventors Koltun, Cregg, Liu and Gill) carries a priority date of 15 September 2020 and was granted 9 September 2025; it claims macrocyclic RAS(ON) inhibitors that form a tri-complex with cyclophilin A — precisely ERAS-0015’s mechanistic class. Joyo’s own issued US patent, 12,458,647 (“Macrocyclic derivative and use thereof”), carries a priority date of 29 September 2022 and was granted 4 November 2025 — roughly two years after RevMed’s priority. That ordering is why RevMed pleads infringement under the doctrine of equivalents together with a derivation and trade-secret theory, rather than contesting priority directly. It is the weaker legal posture on the patent count and the more dangerous one on the trade-secret count.

One further concentration risk surfaced in the patent record. Shuhui Chen is a named inventor on both Joyo’s '647 patent and numerous Medshine patents — meaning the same chemist stands behind both of Erasca’s clinical assets. Whatever one concludes about the RevMed allegations, this means ERAS-0015 and ERAS-4001 are less independent than a two-asset pipeline implies: they share not only a pathway but a chemical origin.

We are careful about what is not established. No public document names the “third party” RevMed alleges misappropriated its trade secrets, and there is no public evidence linking any Revolution Medicines personnel to Joyo; the inventor sets do not overlap. The Joyo licence itself contemplates upstream in-licence agreements whose counterparties are redacted, so the third party need not be Joyo at all. This remains an allegation, and we treat it as one. Guangzhou Joyo Pharmatech (广州嘉越医药科技有限公司) is documented as founded in 2017, having raised roughly $92M across about twelve investors including HongShan, and sponsors NCT06895031 — a seamless Phase 1/2 study begun 31 March 2025. That trial design is itself notable: it is the mechanism that placed the $50M-to-$150M option step entirely within the licensor’s control.

Status, stated precisely: as of this report date this remains a pre-suit demand letter, not a filed complaint. Erasca stated in its Q1 10-Q that it “is not currently a party to any material proceedings,” and press coverage through late July indicates no suit has been filed. That distinction matters and we do not overstate it. Erasca’s Q1 10-Q states that “RevMed may seek to initiate litigation against us,” and Revolution Medicines’ own 10-Q of 6 May 2026 refers to the dispute as one that may arise “whether or not it results in litigation” — both parties treat suit as contingent. A federal-docket search returns no case with Revolution Medicines as a party against Erasca. (One trade headline reads “Revolution Medicines Sues Erasca”; its own body text states neither action has been initiated. We disregard it.)

A securities class action has been filed: Ching Cheng v. Erasca, Inc., Jonathan Lim and David Chacko, No. 3:26-cv-03481-AJB-JLB (S.D. Cal.), filed 10 June 2026 by Block & Leviton LLP, class period 14 January 2025 to 26 April 2026, lead-plaintiff deadline 10 August 2026 — alleging misleading statements about competitive advantage, safety and the “IP moat.” There is one complaint; multiple plaintiffs’ firms are marketing the same case. No Erasca filing has yet disclosed it, as the Q1 10-Q predates it.

We cannot quantify this risk, and we say so. Neither ERAS-0015’s chemical structure nor the claim scope of the '225 patent is public. No outsider can assess doctrine-of-equivalents exposure from public sources. A binary, unmodellable legal risk sits underneath the large majority of the enterprise value.

4.5 Verdict

No durable competitive advantage. Erasca fails all three Greenwald tests. The only barrier that matters in pharmaceuticals — government-granted exclusivity — it does not yet possess, does not itself own, cannot defend by challenge, and is currently being attacked by the market leader on two independent theories.

The disconfirming evidence, weighed honestly. The Phase 1 efficacy signal is unusually strong for dose escalation. The ctDNA clearance data are credible pharmacology. The licensed family does contain an issued US composition patent running to 2043. Management killed five failing programs rather than funding them, and the naporafenib termination was the right call at real cost. If the data survive expansion and randomization, a best-in-class pan-RAS agent would be a genuinely valuable asset.

But value is not the same as moat. What Erasca would own in the good scenario is a licensed compound with time-limited exclusivity in a class that has already demonstrated — through KRAS G12C — rapid multi-entrant fragmentation of the profit pool. The financial metric that would deteriorate if the “moat” disappeared is not margin, retention or pricing. It is the equity value itself, because there is nothing else there. That is the definition of an option, not a competitive advantage.

5. Growth History and Forward Opportunities

5.1 There is no revenue growth to analyze

Erasca has produced zero revenue in every year of its existence. Conventional growth analysis — revenue by segment, organic versus acquired, cohort behavior — has no object. The honest analogs are pipeline progression, spend allocation, and the shape of the addressable opportunity.

5.2 What the spending record shows

R&D expense by program is the most informative table in the filings, because it shows a company that has completely re-based itself:

The two in-licensed RAS assets went from 14% of R&D in FY2024 to 49% in FY2025 to 75% in Q1 2026, while the naporafenib bucket collapsed from $58.3M to $4.7M and discovery collapsed from $40.7M to $2.1M annualizing near $8M. Annualizing Q1 2026, Erasca runs roughly $109M of R&D, of which about $55M goes to ERAS-0015 and $27M to ERAS-4001.

Two readings follow. Favorably, this is cheap clinical development — roughly $47M of annual R&D per clinical asset is efficient by large-biotech standards, and total R&D has been range-bound at $93–115M for four years while the entire pipeline was replaced. Unfavorably, the discovery engine has been switched off. Preclinical and discovery spend has fallen roughly 95% from FY2024 to the Q1 2026 run rate. Erasca is no longer attempting to originate assets; it is a development and in-licensing vehicle running two borrowed molecules and an $8M-a-year research stub. Whatever one thinks of that as a strategy, it means there is no internal source of the next asset if these two fail.

5.3 Forward opportunity

The stated development plan, from the 13 July 2026 release, is genuinely ambitious:

Two features of that schedule matter more than its ambition.

First, these are initiation dates, not readouts. No pivotal trial has begun. With Phase 3 starts in 2027–2028 and typical enrollment and follow-up in these indications, first pivotal data is realistically 2029 or later, and approval 2030–2031. The market is currently paying $5.7B for an outcome three to five years and several binary events away.

Second, the choice of indications is a concession. Erasca has skipped 2L pancreatic entirely — the exact population where its 40% and 57% response rates were generated — and gone straight to first line. The reason is that if daraxonrasib is approved on its accepted NDA, 2L PDAC will no longer be a chemotherapy-comparator setting. The trial Erasca would have to run in 2L is a randomized comparison against an approved pan-RAS inhibitor with a 13.2-month median overall survival, which is slower, larger, costlier and far harder than the single-arm accelerated-approval path that the Phase 1 data would have supported eighteen months ago. The commercial white space is closing while Erasca is still in dose escalation.

There are genuine additional shots: the panitumumab combination in colorectal cancer cleared its first dose-escalation cohort with no DLTs; a Merck clinical-supply collaboration with KEYTRUDA was signed in May 2026; a Tango Therapeutics combination began in March 2026. None of these involve a partner paying Erasca money — they are clinical-supply arrangements, not validating economic deals.

5.4 Verdict

Growth quality cannot be assessed on revenue, and on the metrics that do exist the record is poor but the forward setup is real. The historical record is five clinical programs started and five terminated — a 0% survivorship rate on roughly $450–500M of attributable program spend. The forward opportunity genuinely exists: a $8–12B global peak for a class-leading pan-RAS franchise, in indications with severe unmet need. But it is entirely prospective, entirely dependent on two molecules in-licensed twenty-six months ago, timed to arrive after the competitor, and requiring capital beyond what is on the balance sheet. This is option value with a long fuse, not growth.

6. Financial Quality

6.1 Framing: this is a burn-and-runway story, not an earnings story

Every conventional profitability metric here is undefined or actively misleading. There is no revenue concept in Erasca’s XBRL taxonomy at all. Gross margin, operating margin, ROIC, ROE and free-cash-flow conversion do not exist. Reported ROE of −33% for FY2025 says nothing about business quality and everything about the rate at which raised capital is converted into clinical data.

The correct sector analogs, used throughout: quarterly cash burn, runway in quarters, cost per clinical program, annualized dilution, and cumulative capital consumed per surviving asset.

6.2 The multi-year P&L

The headline loss is nearly useless as a run-rate, and this is the single most important quality-of-earnings point in the file. Reported net loss swings between $123M and $243M with no change in the underlying business, because Erasca expenses all in-licensing consideration immediately as in-process R&D and capitalizes nothing. Q1 2026’s $183.4M loss is 82% one-time — strip the $150.0M Joyo territory payment and underlying operating expense was $37.9M and the underlying net loss $33.4M. Anyone quoting the $183.4M quarterly loss or the $277.0M TTM loss as a run rate is wrong by roughly 4.5x.

Normalized, the business is remarkably stable and modest: underlying operating loss of roughly $117M a year and underlying cash burn of roughly $97M a year. R&D has been range-bound at $93–115M for four consecutive years while the entire pipeline was replaced. G&A has been sticky at $33–42M despite an 18% headcount cut, and now runs $38.6M on 103 employees — roughly $375,000 per employee, which is high.

We flag a management-versus-evidence contradiction: the 10-K states G&A “will increase substantially for the foreseeable future,” while G&A has in fact been flat-to-down for three years.

6.3 Cash burn and runway

Underlying quarterly operating cash burn ran $31.6M / $20.5M / $21.7M / $21.7M through 2025 and $27.4M in Q1 2026. The clean all-in annual burn — change in cash plus marketable securities excluding financing — is approximately $97–99M, or $24.5M a quarter.

One accounting subtlety materially matters: Erasca classifies cash paid for in-licensed IPR&D in investing activities, not operating. Reported operating cash burn therefore understates true cash consumption in any year with a licence payment. True Q1 2026 cash consumption was $27.4M operating plus $150.0M of IPR&D plus $0.6M capex — approximately $178M in a single quarter.

The July 2026 financing. On 13 July Erasca filed an automatically-effective shelf as a well-known seasoned issuer, launched a $500M offering, and upsized and priced it the same day: 31,428,572 shares at $17.50, $550.0M gross, underwriters taking at $16.45 (a 6.00% gross spread worth $33.0M), net proceeds approximately $516.0M, closed 15 July. A 4,714,285-share greenshoe runs to roughly 12 August and had not been exercised as of this report date. There was no concurrent private placement, no warrants, no pre-funded warrants and no convertible structure — a clean common-stock deal. Pricing was a 5.6% discount to the prior close of $18.53, and the deal was launched the same morning as the updated AURORAS-1 data: a textbook data-then-raise sequence.

The prospectus’s own dilution table is worth quoting because it is the honest arithmetic of the asset class: net tangible book value was $1.27 per share before the deal and $2.66 after; new investors took immediate dilution of $14.84 per share, paying 6.6x as-adjusted tangible book. Roughly 85% of the purchase price is option value on data.

Pro-forma cash is approximately $896M ($974M with the greenshoe), assuming a ~$28M Q2 burn.

The runway gap runs toward more burn, not less — and this is important. Management guides that $408.5M at 31 March 2026 is sufficient “into the second half of 2028.” That implies $149–182M of annual burn against FY2025 actual of $98.7M. Management is telling you, through its own runway arithmetic, that burn is about to roughly double — which is exactly consistent with committing to a registration-enabling NSCLC trial in 1H2027 and two Phase 3s in 2027–2028. Three pivotal-scale trials cannot be run on $25M a quarter.

Our independent runway on ~$896M pro forma:

Our honest read: pro-forma cash funds the company into roughly 2030 on a realistic escalating profile. That is genuinely long runway for a clinical-stage biotech — about four years — and it is enough to carry ERAS-0015 through at least the first pivotal readouts without a forced raise. It is not enough to fund three Phase 3 trials to registration plus a commercial build. A further raise or a partnership before approval is close to certain, but on this evidence it would be a raise of choice on data, not a raise of necessity on fumes.

Note that no updated runway guidance appears in the July prospectus. The only live guidance predates the raise. Expect an update at the Q2 print, expected around 11 August 2026.

6.4 Balance sheet

Erasca has zero debt of any kind — no term loan, no revolver, no convertible notes, no royalty monetization. Full-text searches of the FY2025 10-K return nothing for “credit facility,” “term loan,” “revolving” or “convertible note.” Total liabilities of $67.7M are overwhelmingly the $40.7M San Diego operating-lease liability and $26.7M of payables and accrued clinical costs. There is no creditor ahead of equity, no covenant and no maturity wall. Pro-forma cash is therefore also pro-forma net cash.

Important correction to third-party data: ROIC.ai reports roughly $46.0M of “short and long term debt” for Erasca. It is not borrowing. ROIC’s own detail shows null for both short-term debt and long-term borrowings, while short- and long-term capital-lease obligations of $5.251M and $40.702M sum to exactly $45.953M. The balance has declined monotonically over six quarters — the signature of amortizing rent. ROIC’s enterprise value for Erasca is also unusable because it nets only $47.3M of cash and ignores $361.3M of marketable securities. The filing wins; we use it.

There is no going-concern language in any filing, ever. Searches of the FY2025 10-K and Q1 2026 10-Q for “substantial doubt” and “going concern” return zero hits in the audit opinion, notes and risk factors. That is a genuine distinction in a sector where a large share of the clinical-stage universe carries one.

Tax attributes at 31 December 2025: federal NOLs of $413.8M, California NOLs of $576.7M, federal R&D credits of $19.4M — all fully valuation-allowanced, and near-certain to be §382-limited by the 2024 and 2026 offerings. They should not be capitalized in any valuation.

6.5 Dilution — the real cost

The share count is up 2.87x since 2021 — roughly 26% a year compounded. The issuance ladder tells the story: $16.00 (IPO) → $6.50 (Dec 2022) → $2.06 (Apr 2024) → $1.85 (May 2024) → $2.37 (ATM) → $10.00 (Jan 2026) → $17.50 (Jul 2026).

The defining number: in 2024 Erasca issued 130.5M shares — 38% of today’s count — at a blended $1.92, raising $250.9M. Those shares are worth roughly $2.49B at $19.11. That is a value transfer of approximately $2.2B away from pre-2024 holders, and it happened weeks after the company in-licensed the two assets that subsequently drove the re-rating.

Overhang. 58,411,166 options outstanding at 31 March 2026 at a $4.66 weighted-average exercise price — up 10.9M in a single quarter — plus 19,822,874 shares reserved. Fully diluted, that is 400.6M to 420.5M shares, a 17.1% to 22.9% overhang, on top of evergreen provisions adding roughly 6% a year (5% under the equity plan, 1% under the ESPP) through 2031. Option intrinsic value at $19.11 is roughly $844M for a 103-person company — GAAP stock-based compensation of $24.9M a year badly understates the economic transfer, because grant-date fair values were $1.24–1.31. The $200M ATM remains fully undrawn.

6.6 R&D productivity — the pre-revenue return metric

Cumulative R&D of roughly $565M plus cumulative in-process R&D of roughly $366M equals approximately $931M consumed since inception, producing zero surviving internally-discovered clinical assets. Both current clinical programs were in-licensed from Chinese biotechs in a single month — May 2024 — for $22.5M of upfront consideration. Cost per clinical program now runs about $41M a year; G&A runs about $21M per clinical asset and has not fallen as the pipeline consolidated.

6.7 Verdict

Do economics improve with scale? The question does not apply, and that is itself the finding — there is no revenue, no operating leverage, an immovable ~$40M G&A base, and costs about to rise 50–85% by management’s own implied guidance.

On financial condition, however, the verdict is genuinely favorable, and we say so plainly. With ~$896M of pro-forma cash, zero debt, no covenants, no going-concern qualification in company history, a $200M undrawn ATM, an unlimited well-known-seasoned-issuer shelf, and roughly $35M a year of interest income offsetting a third of the burn, Erasca is funded into approximately 2030. Survival is not the question here. That distinguishes it from most of the clinical-stage universe and is the strongest single element of the bull case.

The cost of that survival has been severe and is not finished. A 2.87x share count since 2021, a ~$2.2B value transfer to buyers of the 2024 raises, a 17–23% forward overhang and roughly 6% a year of evergreen issuance. The real financial questions are whether $41M a year per program buys enough shots on goal before 2030 — and whether the next raise happens at $30 or at $3.

7. Capital Allocation

For a pre-revenue biotech, capital allocation is the business: management raises equity and spends it on shots on goal. The verdict turns on research productivity per dollar raised and the price at which dilution was taken.

7.1 The financing record, with price context

Cumulative raised: roughly $1.85B. Against that, current pro-forma cash is $896M and cumulative accumulated deficit is $1.08B.

7.2 Where the money went — the kill list

Every program Erasca owned before May 2024 is dead or handed back. No exceptions.

Attributable killed-program spend is roughly $450–500M — about 40–45% of the accumulated deficit — for zero surviving assets.

Against that, the entire current enterprise value rests on roughly $180M of in-licensing: Joyo at $12.5M upfront plus the $150.0M territory payment, and Medshine at $10.0M upfront.

7.3 The $100M unforced error

This is the sharpest capital-allocation fact in the file and it deserves to be stated precisely.

The Joyo licence priced the option to add mainland China, Hong Kong and Macau at $50.0M if exercised on or prior to the first dosing of the first patient in a Phase 2 trial by either Erasca or Joyo, and $150.0M after that first dosing. Erasca held the option for twenty-two months. In March 2026 it exercised — and, “based upon feedback from Joyo that Joyo had dosed the first patient in a Phase 2 clinical trial,” paid the $150.0M tier.

Erasca paid $100M more than the contract’s cheap tier because its own licensor tripped the trigger first. The company held $341.8M at 31 December 2025 and had just taken in $242.7M of fresh January proceeds — it was not capital-constrained. The consequence was immediate and disclosed: guided runway was walked back from “into the first half of 2029” (8-K, 29 January 2026) to “into the second half of 2028” (10-K, 12 March 2026) within six weeks. The $100M step-up cost roughly two to three quarters of runway for Greater China rights to an asset that will not be commercialized anywhere for years.

We can construct a defense — the option had to be exercised eventually, and China rights have real long-run value. We do not find it adequate. The trigger was foreseeable, monitorable and cheap to pre-empt.

7.4 Restructurings and planning discipline

There has been one formal reduction in force — May 2024, roughly 18% of staff, a $2.2M charge, headcount from 126 to 103. But there have been four separate pipeline prioritizations in 39 months (February 2023, November 2023, May 2024, May 2025), each killing assets previously described as differentiated. The 78,000-square-foot San Diego headquarters has been carved into subleases with $4.7M of impairments taken within 30 months, against $60.4M of undiscounted lease commitments running to 2032.

Interpretation: this is fast correction of repeated over-commitment, not disciplined planning. Killing failing programs quickly is genuinely a virtue in this industry; needing to do it four times in three years is evidence about the quality of the original commitments.

7.5 Incentives — quoting the actual metrics

The proxy’s disclosure of its own bonus determination, quoted verbatim:

“For 2025, annual bonuses were based on corporate performance relative to key corporate objectives tied to clinical and regulatory priorities as well as operational and financial objectives. Following a comprehensive review… our compensation committee approved the payment of annual bonuses to our NEOs at 120% of target levels.”

That sentence is reproduced essentially verbatim in all five proxies. Only the percentage changes: 125% (FY2021), 90% (FY2022), 95% (FY2023), 115% (FY2024), 120% (FY2025). In five years as a public company, not a single goal, weighting or threshold has ever been disclosed, and the payout has never fallen below 90% of target. It paid 115% in FY2024 — the year of the 18% reduction in force, three program deprioritizations, and raises at $2.06 and $1.85. It paid 120% in FY2025 — the year the company killed its most advanced clinical asset.

On equity, also verbatim: “our equity awards vest over four years, subject to the employee’s continued employment.” One hundred percent of named-executive equity is time-vesting stock options. There have been zero performance share units and zero performance conditions, ever.

FY2025 compensation: Jonathan Lim $4,298,838 total (salary $670,400, options $3,135,250, incentive $482,688); David Chacko $1,778,496; Shannon Morris $1,750,898. Aggregate $7.83M, or 5.6% of operating expense; cash compensation alone is 1.9%. Base salaries were raised for 2026 to $700,000 / $540,000 / $536,400, with January 2026 option grants of 2,000,000 / 640,000 / 625,000 shares.

Two further governance items. In May 2024, 7,478,918 options were repriced to $2.35 — properly excluding Section 16 officers and the board, which is a real mitigant. But the named executives then received fresh grants dated 20 June 2024 at $2.03, a lower strike, thirty days later. And FY2025 grants of 15,078,786 options were struck at $1.75 — a 5.3% burn rate one quarter before a 19x move.

There has never been a say-on-pay vote. As an emerging growth company Erasca files with no compensation discussion and analysis, no pay-versus-performance table, no pay-ratio disclosure and no advisory vote. The 2026 annual meeting put only director elections and auditor ratification to shareholders. EGC status lapses on 31 December 2026, making the 2027 proxy a genuine event.

Governance is otherwise mixed: six of eight directors independent (Michael Varney is a paid employee), a combined chair and CEO, a classified board with two-thirds removal — but a single share class, prohibitions on both pledging and hedging, a Dodd-Frank clawback since 2023, and no evidence of option timing around material information. The board voluntarily cut its own annual grants by 75% in April 2026 by moving to fixed-value awards.

7.6 Insider behavior — and a signal that is not one

The Form 144 cadence is a false positive and should be dismissed. Thirty-eight of forty-one Form 144 notices are the Erasca Foundation selling exactly 8,333 shares per month through Merrill Lynch since July 2023, liquidating the 1,093,557-share charitable donation made at the IPO. That is roughly 0.03% of shares outstanding per year and carries zero information. Any screen reading “forty Form 144s in three years” as heavy insider selling is misreading a charity’s standing order.

Actual insider selling totals 220,000 shares — 0.06% of the company — in three transactions, all option-exercise-and-sell under a Rule 10b5-1 plan adopted 30 June 2024, when the stock traded near $1.90: General Counsel Ebun Garner 120,000 at $5.59 (7 January 2026) and 80,000 at $16.40 (1 April 2026), and CMO Shannon Morris 20,000 at $15.04 (4 March 2026). Because the plan predates the ERAS-0015 clinic by roughly ten months, these carry no inference about the April data. Two observations survive anyway: all three sales were executed in the four months before the −48% break, and the Chief Medical Officer of a clinical-stage oncology company now owns zero shares outright while the General Counsel owns 25,076 — against 625,000 and 560,000 fresh options granted in January 2026. Executive economic exposure is essentially entirely option-shaped, and therefore convex to volatility rather than symmetric with shareholders.

On the other side of the ledger, and it matters: CEO Jonathan Lim bought 2,300,000 shares on the open market for roughly $4.71M of his own money, in six separate transactions, all into weakness: 40,000 at $4.99, 60,000 at $3.86, 100,000 at $2.84, 100,000 at $2.75, 1,000,000 at $2.03, and 1,000,000 at $1.69–1.72 between December 2022 and December 2023. He has never sold a share, and holds 36,574,772 shares — 11.6% of the company, roughly $700M, unhedged and unpledgeable. Director Alexander Casdin bought 150,000 shares for ~$384K on the same pattern. This is genuine, repeated, sizeable conviction buying at the lows and it is the strongest evidence in the file that management believed its own thesis.

Two qualifications: there have been zero open-market purchases by anyone since 21 May 2024 — including on the −48% break — and the six Form 4s filed on 29 June 2026 are routine annual director option grants, not purchases.

7.7 Verdict

Mixed, improving from a poor base — but the improvement is a purchased asset, not a demonstrated competence.

Management is a competent, decisive capital recycler and a poor capital originator. It kills failing programs quickly, it bought well when it bought, and it has built nothing itself.

The defense has real force. Attrition is the oncology base rate. Killing five programs rather than funding them was correct each time, and the naporafenib termination in particular was the right decision taken at real cost. The company issued nothing in 2025, burning its balance sheet down to $341.8M rather than printing stock at $1.06 — that is genuine discipline and deserves credit. Both 2026 raises were opportunistic, at-market, into strength, with a tier-one syndicate. Cash compensation is modest. The founder holds 11.6% unhedged and has never sold a share. The board cut its own pay.

The indictment is equally real. Four prioritizations in 39 months. A headquarters sublet in four pieces. $100M of unforced cost on the Joyo option. Sixty percent of every share ever issued went out in 2024 at a blended $1.92, transferring roughly $2.2B of value. And a compensation system that has never disclosed a metric in five years, has never paid below 90% of target, paid 115% in the worst year of the company’s life, gates 100% of executive equity on nothing but the passage of time, struck a 5.3% option burn at $1.75 one quarter before a 19x move, and sits inside a 17–23% overhang with a 6%-a-year evergreen that no shareholder has ever voted on.

Net: one very good purchase and decisive triage, funded by heavy dilution at the lows, rewarded by a compensation system that would have paid substantially the same if the purchase had failed. The correct reading is that 2024–2026 represents one decision that worked, not a repeatable process. The 2027 proxy — the first after EGC status lapses, with say-on-pay and pay-versus-performance disclosure required — is the test worth watching.

8. Changes and Headwinds — Last Two Years

The last twenty-six months have transformed Erasca more completely than most companies change in a decade. The company that exists today shares a name, a CEO, a pathway thesis and a building with the company of mid-2024 — and essentially nothing else.

8.1 The event timeline

8.2 The four headwinds that matter

First, the competitive clock has moved decisively against Erasca. In April 2026 daraxonrasib reported median overall survival of 13.2 versus 6.7 months (HR 0.40) in previously treated metastatic pancreatic cancer, was published in the New England Journal of Medicine, and had its NDA accepted on 22 July under a priority-voucher pathway that has produced approvals in 44 and 55 days. Eighteen months ago Erasca’s Phase 1 response rates pointed toward a single-arm accelerated approval against a chemotherapy backdrop. That door is plausibly closing. If the FDA requires randomization against an available effective therapy, Erasca’s registration path becomes slower, larger, costlier and much harder to win — and Erasca has already conceded the point by skipping second-line pancreatic entirely in its Phase 3 planning.

Second, the IP position is under direct attack and the contractual protection is thin. Covered in the intellectual-property discussion above; the essential point is that the claim reaches an asset Erasca does not own, through a licence that disclaims any non-infringement warranty and assigns all risk to Erasca.

Third, safety now carries a real asterisk. A treatment-related death from Grade 3 pneumonitis progressing to Grade 5, at 24 mg — a recommended dose for expansion — in a first-in-class mechanism. We are careful here: there were no dose-limiting toxicities, zero discontinuations for treatment-related adverse events, and 100% median relative dose intensity, and one event in a heavily pretreated patient with pulmonary metastases and prior lung cryoablation may well be idiosyncratic. But a 72% all-grade rash rate indicates substantial on-target pathway toxicity, and Erasca’s entire differentiation claim is a wider therapeutic window at lower dose. That is precisely the claim a fatal pneumonitis undercuts.

Fourth, the KRAS G12C commercial precedent has deteriorated further. Roche’s divarasib beat both approved G12C inhibitors head-to-head on PFS and OS in July 2026 — demonstrating inside this very target class that first-mover advantage lasts about five years and under $1.5B of cumulative revenue before displacement.

8.3 What genuinely improved

Balance: three things got materially better. Capital position — from $341.8M at the end of 2025 to roughly $896M pro forma, funded into approximately 2030 with zero debt. Asset quality — the two in-licensed RAS programs are, on the available evidence, better assets than anything Erasca previously owned; the pivot worked. Focus — from five clinical programs across four mechanisms to two programs on one pathway, with R&D spend concentrated 75% on the two lead assets.

8.4 Verdict

On balance these developments weaken the thesis, and the weakening is structural rather than sentiment-driven.

The strengthening is real but largely financial and already reflected: the company is better capitalized and better focused than at any point in its history, and it owns a more interesting molecule than it has ever owned. The weakening is competitive, legal and regulatory — the three areas where a clinical-stage company has the least control. The single most consequential change of the period is not anything Erasca did; it is that Revolution Medicines won its Phase 3 and filed. That one event simultaneously validated the mechanism Erasca is pursuing (helpful) and closed the fast, cheap regulatory path Erasca was pointed toward while raising the efficacy bar its own Phase 3s must clear (much less helpful). Target validation and commercial white space are, in this instance, substitutes rather than complements.

9. Risk Analysis

9.1 Risk matrix

9.2 What is not a material risk here

Two risks that dominate most clinical-stage biotech write-ups do not apply and should not be double-counted. There is no financing-distress risk in the near term — zero debt, no covenants, no maturity wall, no going-concern qualification in company history, and roughly four years of runway. And there is no accounting-aggressiveness risk of consequence — Erasca capitalizes nothing, expensing all in-licensing consideration immediately as in-process R&D. The income statement is lumpy and near-useless as a run-rate, but it is not flattering; if anything it overstates losses relative to underlying cash burn.

9.3 The risk that dominates

Risks 1, 2 and 3 are the file. Risk 1 (efficacy regression) and Risk 2 (IP) are genuinely binary and are not diversifiable within this security — there is no second asset of consequence to absorb a failure, because ERAS-4001 has no human data and targets the same pathway. Risk 3 (regulatory path) is the most under-appreciated, because it does not announce itself with a press release: it changes the shape of the investment quietly, by adding years and hundreds of millions of dollars to the path, and it is largely outside Erasca’s control.

10. Valuation Discussion — Embedded Expectations

No price target and no recommendation appears in this section or anywhere else in the analytical body of this article. The purpose here is to establish what the current price requires the world to look like.

10.1 The enterprise-value bridge

Sensitivities. Treating the $46.0M of lease liabilities as debt-like gives $5,706M — immaterial. The greenshoe is effectively EV-neutral (cash and shares rise together), so its unconfirmed status does not affect the analysis. On a treasury-stock basis, however, the 58.4M options add roughly 46M net new shares, giving 388.6M fully diluted shares and a fully diluted EV of about $6,310M. The honest range is $5.66B to $6.31B; the analysis below uses the lower figure, which is the friendliest to the bull case.

10.2 Why the standard multiples are useless, and what replaces them

EV/Revenue, EV/EBITDA and P/E are not merely unhelpful here — they are undefined. ROIC returns null for all three. The AZI valuation index returns null P/E and null P/S with n_components = 1; the only definable metric is P/B at 14.78, the 97.5th percentile of Erasca’s own multi-year history. That tells you the market is paying more per dollar of balance-sheet cash than at any point in this stock’s life. It is own-history context only — never cross-sectional — and it is not a valuation.

Four frameworks replace them: (1) EV as the market’s implied risk-adjusted NPV of the pipeline — since there is no debt, EV is the price of ERAS-0015 plus two backups; (2) rNPV, the workhorse below; (3) the cash-plus-backups residual, which brackets the downside; and (4) precedent take-out values, the reality check.

10.3 Building the rNPV

Step 1 — peak unadjusted global sales. [ASSUMPTION set, sourced inputs]

Derivation: 67,530 US pancreatic cases (ACS 2026) × 90% PDAC × 55% metastatic × 85% KRAS-G12X = 28,413 US, ×1.7 for EU5+Japan. 229,410 US lung cases × 87% NSCLC × 62% advanced × 28% RAS-mutant = 34,648 US, ×2.2. Net price of $150K/yr is anchored on Krazati’s ~$237K and Lumakras’s ~$223K annual gross WAC, less 30–35% gross-to-net and ex-US discounting. China is excluded entirely, which is conservative given the territory expansion. We model $3.5B to favor the bull. For reference, sell-side consensus for daraxonrasib — which has already won its Phase 3 — is $5–7.6B.

Step 2 — the success-case NPV. Launch 2030; uptake ramp 5/20/40/60/80/95/100%; plateau to 2040; erosion 2041–44. Operating margin 45% at scale, less the 5% Joyo royalty, at a 15% blended cash tax rate reflecting the NOL shield — a 34% after-tax margin on net sales. Discount rate 13% (midpoint of the 12–15% clinical-stage band), mid-year convention.

PV of commercial cash flows = $2,866M, less the PV of remaining Joyo milestones ($57.5M regulatory, $125.0M commercial) = conditional NPV of $2,782M. That is 0.795× peak sales — a reusable constant: every $1B of peak is worth about $795M of present value if approved.

Step 3 — the development cost required to find out. Three Phase 3-scale programs plus ongoing Phase 1/2, G&A and a commercial build: $80M (2H26), $250M (2027), $350M (2028), $400M (2029), $400M (2030), $300M (2031) = $1,780M nominal, $1,184M present value, $1,006M probability-weighted.

This is the quiet fact that shapes everything: pro-forma cash of ~$882M funds roughly two of the five-and-a-half years required. Further equity issuance is embedded in every success case, not only the bear case.

Step 4 — the rNPV at a defensible probability. Equity value = PoS × $2,782M − $1,006M + $882M:

At a $3.5B peak, no probability of success — not even 50%, which no pre-pivotal asset deserves — produces an equity value above one-fifth of today’s market capitalization.

10.4 Embedded expectations — the back-solve

Invert the model and solve for what the price requires:

EV = PoS × k × Peak − PV(development spend)
5,660 = PoS × 0.795 × Peak − 1,006
PoS × Peak = $8,385M

The market is underwriting a product of probability and peak sales of roughly $8.4 billion. Decomposed:

Read the last row carefully. Even assuming ERAS-0015 were already as de-risked as a drug with a positive Phase 3 and a filed NDA — which it demonstrably is not — the current enterprise value still requires a $9.9B peak. That is roughly Keytruda scale, and above the consensus range for daraxonrasib, the pan-RAS asset that has already won.

On a disciplined rNPV there is no combination of defensible probability and defensible peak sales that reaches the current enterprise value.

10.5 The calibration check — is the model too harsh, or is the sector rich?

Intellectual honesty requires the counter-test, and it matters. Run the same 13% discount rate and 34% margin on Revolution Medicines and you also get a value far below its own ~$38B enterprise value. The framework is not singling Erasca out; the RAS franchise as a whole is priced above a 13%-discount intrinsic value. The fair test is therefore relative — calibrate to the multiple the market actually pays the leader:

RVMD EV / risk-adjusted peak ≈ $38.3B / (0.85 × $10B franchise peak) ≈ 4.5×
ERAS required risk-adjusted peak ≈ $5,660M / 4.5 ≈ $1,258M

Even on the sector’s own generous yardstick, Erasca requires a 20–25% probability of success for a molecule that has not started a pivotal trial — against a published benchmark of 5.3–15.9% — and a peak at or above the leader’s consensus. That is the most charitable statement the evidence supports, and it remains aggressive.

10.6 Scenarios

Implied values use the market-calibrated multiple. On the disciplined rNPV the same three scenarios give roughly −$0.1B / $0.35B / $2.1B — even the bull case lands at about 0.3× today’s market capitalization. The gap between the two sets of numbers is not a modeling error; it is the size of the sector’s optimism premium, and understanding that gap is the single most important thing about this valuation.

Probability-weighted (30/45/25, market-calibrated): approximately $4.97B against a ~$6.54B market capitalization — roughly 24% below today’s price, using the sector’s own generous multiple.

The downside is knowable. If ERAS-0015 fails or the IP claim succeeds, the residual is cash less wind-down plus the backups. The pipeline is 100% in-licensed and thin: naporafenib is terminated, ERAS-4001 has no human data, ERAS-12 is preclinical, and — as discussed above — both clinical assets share a chemical origin. Modeled residual: roughly $800M of cash at the point of failure × 0.85 (busted biotechs historically transact at 0.7–1.0× net cash) plus ~$150M for the backups = ~$830M of equity, about 13% of today’s market capitalization. This is not hypothetical: the 2026 low was $3.49 and the 2025 low $1.06. Note the timing asymmetry — if failure arrives late, at a 2030 pivotal readout, cumulative spend will be $1.1–1.3B and at least one more large raise will have occurred, so the residual per unit of ownership is materially worse. Late failure is the expensive kind.

10.7 Comparables and precedent transactions

Erasca trades at the fourth-highest enterprise value in this set while being the least clinically advanced. It carries roughly 2.0× the EV of Relay — the closest stage-matched comparable, a Phase 1/2 precision-oncology name with an uncontested IP position — and about 15% of Revolution Medicines’, a company that has already won the Phase 3 Erasca has not started. Janux’s negative enterprise value is the sharpest datapoint in the table: this same market, in this same week, assigns less than zero to a Phase 1 oncology pipeline. What it is paying Erasca is not a sector-wide phenomenon; it is specific to this story.

Every recent Phase 1/2 oncology take-out sits below Erasca’s current enterprise value. The only larger transaction is BMS/Mirati — and Krazati was already FDA-approved and generating revenue. The current EV therefore already prices in a strategic outcome better than any comparable Phase 1 precedent of the last five years. At this level, M&A is not a safety net; it is a stretch. That conclusion is reinforced by the academic evidence cited in the industry section above: among small pharmaceutical firms, being acquired correlates with weakness rather than strength, and the unconditional base rate is roughly 2% per firm-year.

10.8 What the market is underwriting correctly

1. The pan-RAS mechanism works. RASolute 302 is genuine, hard-won target validation and legitimately raises the prior on any well-designed pan-RAS molecule.
2. 2L+ pancreatic cancer is a real unmet need. A 6.7-month median overall survival on chemotherapy is an open field.
3. Erasca is well capitalized and unlevered — ~$882M, zero debt, funded to roughly 2030.
4. Biomarker selection genuinely improves the odds — 15.9% versus 7.6% is a large, real effect that applies here.
5. Management has sold an oncology company before. Jonathan Lim built Ignyta and sold it to Roche. That matters when the plausible exit is a trade sale.

10.9 What the market may be underwriting incorrectly

1. It is pricing target validation as though it were asset validation. Daraxonrasib’s Phase 3 proves the mechanism works. It says nothing about whether ERAS-0015 works. The market appears to have collapsed two distinct propositions into one.
2. It is treating N=7 as a de-risking event. A 57% unconfirmed 8-week response rate in seven patients carries a confidence interval spanning roughly 18% to 90%. That is a reason to run a larger trial, not a valuation input.
3. It is not discounting the regulatory consequence of RevMed’s win — the most under-appreciated item here. Daraxonrasib reaching market first with a 13.2-month overall-survival benchmark plausibly closes the single-arm accelerated-approval door for a second pan-RAS agent in 2L+ pancreatic cancer. If randomization against an active comparator is required, the path becomes slower, larger, costlier and much harder to win. That single regulatory question moves both the probability of success and the time to revenue simultaneously, and is worth more to this valuation than any other assumption.
4. It is treating the RevMed claim as noise. The asset is entirely in-licensed, the patent estate is prosecuted and controlled by a counterparty Erasca does not control, the licence disclaims non-infringement warranty and provides no third-party IP indemnity, and Erasca is contractually barred from challenging the patents.
5. It appears not to price the full claim on future value — roughly 46M net new shares from in-the-money options, a 6%-a-year evergreen, a $200M untapped ATM, and at least two further equity raises embedded in every success case.
6. The therapeutic index may be narrower than the headline suggests. A 72% all-grade rash rate and a Grade 3 pneumonitis that progressed to Grade 5 are the early shape of a tolerability ceiling in a drug whose commercial case depends on chronic dosing in combination.

Section verdict. The enterprise value of roughly $5.66B — up to $6.31B fully diluted — requires a product of probability and peak sales near $8.4B. Against a published oncology Phase 1 likelihood of approval of 5.3%, or 15.9% with biomarker selection, that implies a peak of $53–158B, a number that does not exist in oncology. Calibrating instead to how this market actually prices the leader, the requirement softens to a 20–25% probability of success on a $5.0–6.3B peak. That is the most charitable reading the evidence permits, and it still asks the market to grant a pre-pivotal, in-licensed, IP-contested Phase 1 asset a probability of success roughly one-and-a-half to five times the published benchmark, and a peak at or above the consensus for the drug that already won. The current price embeds assumptions that the available clinical, regulatory, competitive and legal evidence does not currently support.

11. Variant Perception

11.1 What consensus believes

Consensus, as expressed by the price and by post-data sell-side actions, holds that ERAS-0015 is a best-in-class pan-RAS molecular glue whose potency advantage — activity at roughly one-eighth to one-tenth of daraxonrasib’s dose — will translate into a superior therapeutic window, allowing deeper pathway inhibition, better combinability, and ultimately a share of a validated multi-billion-dollar market. On this view the April crash was an over-reaction to an idiosyncratic death in a heavily pretreated patient and to a legal letter with no complaint behind it; the July data confirmed the thesis; and a well-capitalized company with a proven dealmaker at the helm is now a credible acquisition target in the middle of a $236B patent cliff.

11.2 The strongest bull case

Stated at its best, because it is not a weak case:

The mechanism is proven — RASolute 302 removed the central scientific risk that hung over this entire class, and it did so after Erasca had already in-licensed its asset for $12.5M. ERAS-0015’s early data are, on their face, better than daraxonrasib’s at the same stage: 62% versus 38% in 2L+ NSCLC, 40% versus 29% in 2L pancreatic. The ctDNA data are the tell — 14 of 14 patients clearing at least 75% of variant allele frequency is deep, consistent target engagement, and it is the kind of pharmacology that usually precedes real clinical benefit. Achieving it at 8–32 mg rather than 300 mg suggests a genuinely different molecule, not a copy. The combination programs are progressing: panitumumab cleared its first cohort with no dose-limiting toxicities, and Merck and Tango have both put their drugs alongside it. The company has ~$882M, no debt, and four years of runway — it will not be forced to sell or dilute at a bad price. Insider alignment is real: the CEO owns 11.6% and bought 2.3M shares with his own money at the lows. And in a sector where large pharma must replace $236B of revenue with $180B of dry powder, a differentiated pan-RAS asset is exactly what gets bought — at a premium.

11.3 The strongest bear case

Also stated at its best:

You are paying $5.66B for a molecule the company does not own, on evidence that would not survive a confirmatory scan. Every efficacy figure is an unconfirmed response rate at 8–14 weeks, in cohorts as small as seven, pooled across two trials in two countries, one run by a licensor whose operations Erasca concedes it did not control. There is no PFS, no OS, and no duration of response — and duration is what determines commercial value in this class, where the competitor’s 15.1-month median is the bar. The one hard safety datapoint is a treatment-related death at a go-forward dose in a drug whose entire differentiation claim is a wider therapeutic window.

Meanwhile the commercial opportunity is being closed from the front. Daraxonrasib will reach the market roughly four years earlier with a 13.2-month overall-survival benchmark, which likely converts Erasca’s registration path from a single-arm accelerated approval into randomized trials against an active comparator — slower, larger, costlier, harder. Erasca has already conceded this by abandoning 2L pancreatic for its Phase 3. And the class’s own history is damning: two triumphant first-in-class KRAS G12C approvals produced $568M of combined revenue in year five and are now being displaced head-to-head by a third entrant.

Underneath all of it sits an unquantifiable legal binary: the market leader alleges that Erasca’s asset infringes its patent and derives from its trade secrets, and Erasca’s licence provides no third-party IP indemnity, no warranty of validity or non-infringement, “as is, with all faults and with all defects,” with all risk expressly assumed by Erasca — and bars Erasca from challenging the patents it depends on. The historical return on this management team’s ~$931M of deployed capital is approximately −100%, and the entire enterprise value rests on a $12.5M in-licence made twenty-six months ago, into which they subsequently overpaid by $100M.

11.4 The three to five assumptions that actually matter

1. Does the unconfirmed response rate confirm, and does it last? Everything downstream depends on this and it is currently unmeasured.
2. Is ERAS-0015 differentiated, or merely active? Being a second working pan-RAS is worth a fraction of the current price; being a demonstrably better one is worth all of it. This is the single load-bearing word in the bull case.
3. Does the FDA still permit a single-arm accelerated approval for a second pan-RAS agent once daraxonrasib is approved? This one regulatory judgment moves probability of success and time to revenue simultaneously.
4. Does the RevMed claim have substance? Unknowable from public sources — the compound structure and the claim scope are both undisclosed — and the licence gives Erasca almost no protection if it does.
5. Does the pan-RAS profit pool resemble the $8–12B promise or the $568M G12C outcome? The class has already run this experiment once.

11.5 What the tape and the factor model say

The positioning read is unusually clean and it disciplines the framing. The factor model zeroes Momentum, Value, Quality, LowVolatility and Growth across all four nested models, despite a twelve-month return above +1,000%. What survives is Market (+1.18 to +1.25), SmallSize (+1.13 to +1.50, biased upward by serial issuance and to be discounted), a Biotech industry loading (+0.81 to +1.03), and a materially negative liquidity loading. R² is 14–20% and idiosyncratic volatility is 97.4% annualized against 105–110% total. Roughly 86% of the variance is company-specific clinical and legal event risk.

Nothing in factor space explains this stock. Calling it “momentum” would be asserted, not evidenced; calling it a “falling knife” would be equally wrong. It is a pure binary that has already re-rated twice — and the tape has already agreed with the bull case. The price sits above a fully positive moving-average stack, 9% above the July placement price and only 15% below the April closing high, at a $6.5B market capitalization on a Phase 1 asset. The fundamental investor is not early here.

Two further disciplines from the same data. First, the sector factor is crowded and extended: the biotech industry factor is +25.2% over 252 days at a z-score of +2.5, which is an extreme reading — and an extended sector factor raises the cost of being wrong on a binary, because a sector de-rate and a company-specific disappointment can arrive together. Second, the long-horizon record: five-year annualized return +1.3%, five-year Sharpe −0.009, five-year maximum drawdown −95.6%. The +9.67 one-year Sharpe is real arithmetic but is a single-event artifact of the same series. Quoting one without the other would be misleading.

Where consensus may be offsides: it has priced a Phase 1 asset as though the binary had already resolved, in a sector factor at +2.5 standard deviations, on evidence that contains no duration data and sits under an unresolved legal claim against IP the company does not own.

11.6 Verdict

The variant perception is not that the science is bad — we think the molecule may well work, and the ctDNA data are genuinely persuasive pharmacology. The variant perception is that “works” and “worth $5.66B” are separated by four distinct hurdles that the market appears to be treating as one: confirmation and durability, differentiation versus an approved competitor, a regulatory path that has quietly become harder, and ownership of the intellectual property. Consensus is pricing the first hurdle as if it cleared all four.

12. Fact vs. Interpretation

13. Open Questions

1. What is the confirmed response rate, and what is the duration of response? No PFS, OS or DoR has ever been disclosed for ERAS-0015. This is the single largest evidentiary gap in the file. Resolvable: expansion data guided 1H2027.
2. Will RevMed file suit, and on which theory? As of this report date it is a demand letter. A filed complaint seeking preliminary injunctive relief would be a materially different fact. Watch: any 8-K; the Q2 10-Q, due ~11 August.
3. Does the '225 patent read on ERAS-0015? Unknowable — neither the compound structure nor the claim scope is public. This is an unquantifiable binary sitting under the majority of enterprise value, and no amount of further public research resolves it.
4. Who is the “third party” alleged to have misappropriated trade secrets? Unnamed by either party. The Joyo licence contemplates upstream in-licence agreements whose counterparties are redacted (Exhibit 1.77, not filed). No public evidence links RevMed personnel to Joyo.
5. Will the FDA accept a single-arm accelerated-approval path for a second pan-RAS agent? The most consequential unresolved question for valuation, and one Erasca has partly pre-answered by skipping 2L pancreatic in its Phase 3 plan.
6. What is the actual 2027–2029 development cost profile? Management has committed to three pivotal-scale programs without disclosing their cost. Our $1.78B estimate is an assumption.
7. What is updated runway guidance post-raise? The July prospectus contains none; the only live guidance predates the raise. Expect an update at the Q2 print, ~11 August 2026.
8. Was the 4,714,285-share greenshoe exercised? Window closes ~12 August 2026. Worth up to $77.5M of additional net proceeds.
9. What is ERAS-4001’s Phase 1 monotherapy data? Guided 2H2026 — the nearest binary, and the only diversification in the pipeline. Note, per the intellectual-property discussion above, that it shares a named inventor with ERAS-0015, so the assets are less independent than they appear.
10. Does Erasca have, or intend to assert, a claim against Joyo under §12.2(b)? Undisclosed. Recourse is knowledge-qualified, arbitration-only and PRC-collectible.
11. What was the Joyo China trial’s data quality? Erasca pooled it into headline efficacy while its own risk factors concede it “did not control Joyo’s operations or compliance systems.”
12. Will the 2027 proxy — the first after EGC status lapses, with say-on-pay and pay-versus-performance required — show a change in compensation design?
13. Three unexplained price moves (+31.8% on 2025-05-29, +42.4% on 2026-01-07, +14.2% on 2026-06-29) have no identified company disclosure. We leave them unattributed rather than assign a speculative cause.

14. What Must Be True

14.1 For the bull case

14.2 For the bear case

14.3 The single most important test

If we could observe only one thing, it would be the confirmed objective response rate with mature duration of response in an expansion cohort of at least fifty patients. It resolves assumption 1 on both sides simultaneously, it is scheduled (1H2027), and it is the input to which every valuation scenario in §10 is most sensitive. Everything else in this file — the IP claim, the regulatory path, the competitive position — modulates the value of a drug that works. That test determines whether there is a drug at all.

15. Source Appendix

See Appendix B — Source Appendix below, which lists every primary and secondary source relied upon — SEC filings, the Joyo License Agreement exhibit, clinical and scientific literature, regulatory sources, industry data, financial data providers, internal research library materials, and news and trade press — with publisher, date, URL and the claim each supports, and marks primary versus secondary.

Sections 1–15 contain no investment recommendation and no price target. The Claude's Take block at the head of this article is Claude’s own independent, subjective opinion and is clearly labeled as such. Nothing in this article is investment advice. The author holds no position in the securities discussed. Readers should do their own research and consult a qualified adviser before making any investment decision.

APPENDIX A — Standard Diligence Questionnaire

Report date: 2026-07-25 · Price: $19.11 (2026-07-24 close) · Market cap: ~$6.54B pro forma · EV: ~$5.66B

Where a question does not map to a pre-revenue, clinical-stage biotechnology company, we say so explicitly and give the correct sector analog rather than forcing an answer.

General

What thoughtful questions have other investors asked about this company?

The serious questions cluster in five places, and they are the right ones.

1. “Is the 62%/57% response rate real, or is it small-N selection?” This is the central question and it is currently unanswerable. Every figure is an unconfirmed ORR at 8 or 14 weeks in cohorts of 2 to 37, pooled across two separately designed trials in two countries. The 57% headline is four responders out of seven patients. FACT.

2. “Does the potency advantage translate into a therapeutic window advantage?” Erasca’s differentiation claim is that ERAS-0015 is active at roughly one-eighth to one-tenth of daraxonrasib’s dose. The bull reading is a wider window; the bear reading is that a 72% all-grade rash rate and one fatal pneumonitis at a recommended dose suggest the window is narrower than the dose ratio implies. OPEN QUESTION.

3. “How much does the RevMed letter actually matter?” The sophisticated version of this question focuses not on the patent count but on the derivation theory and on the licence architecture — because if the defect is inside the licensed IP, Erasca cannot cure it. INTERPRETATION: this is the question that most separates informed from uninformed holders.

4. “What does RVMD’s approval do to Erasca’s regulatory path?” Under-discussed relative to its importance. If the FDA will no longer grant single-arm accelerated approval for a second pan-RAS in a setting where an effective therapy exists, the path lengthens by years and hundreds of millions of dollars.

5. “Is this a takeout?” Frequently asked, and the evidence is unfavorable: every recent Phase 1/2 oncology take-out has been below Erasca’s current enterprise value, and the academic literature finds that among small pharmaceutical firms, being acquired correlates with weakness rather than strength, at a base rate near 2% per firm-year.

A sixth question is asked less often than it should be: “What did this team actually build?” The answer is that roughly $931M of cumulative R&D and in-process R&D has produced zero surviving internally discovered clinical assets, and the entire enterprise value rests on a $12.5M in-licence signed in May 2024.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low?

N/A — there are no earnings and never have been. The correct analog is where the company sits in the biotech capital cycle and in its own newsflow cycle, and on both readings it is near a high, not a low. Enterprise value rose from roughly $109M at 30 June 2025 to roughly $5.66B today — a ~52x increase in thirteen months with no approved product, no revenue and no pivotal trial started. Sector-wide, the XBI is up roughly 70% over twelve months following a 63.9% drawdown, biotech IPO issuance is at its highest since 2021, and the biotech industry factor sits at a +2.5 standard-deviation reading. INTERPRETATION: this is a cyclical high in the price of optionality, not a cyclical high in earnings.

Driven by the external environment or internal actions?

Both, and the split is instructive. The January 2026 move began as an external event — takeout speculation around Revolution Medicines re-rated every listed pan-RAS asset, and Erasca gained 42.4% on roughly 27x normal volume with no filing of its own. The subsequent moves were internal: the January data release, the April data-and-disaster, the July update. FACT/INTERPRETATION.

How stable are revenues?

N/A — zero revenue. The relevant stability question is cash burn, and there the answer is genuinely favorable: underlying operating burn has been remarkably steady and modestly declining at roughly $97–99M a year, or $24.5M a quarter, for three years. That stability is about to end — management’s own runway guidance implies burn stepping up to $149–182M a year as three pivotal programs start.

Outlook for products/services?

There are no products. Two clinical assets: ERAS-0015 in Phase 1 dose escalation with expansion data guided for 1H2027 and pivotal starts in 2027–2028, and ERAS-4001 with first human data guided for 2H2026. First pivotal readout is realistically 2029 or later; approval 2030–2031 at the earliest.

How big will this market be — growing, shrinking, domestic or international?

The addressable population is genuinely large and stable-to-growing with an ageing population: roughly 70,000–80,000 US metastatic KRAS-mutant treatable patients a year across pancreatic (~25,000–28,000), NSCLC (~21,000–23,000) and colorectal (~22,000–25,000). Global, with Erasca now holding worldwide rights including Greater China after the $150M option exercise.

The critical caveat, and it is the most important disconfirming datapoint in the file: incidence does not equal revenue. Two approved, scientifically triumphant KRAS G12C inhibitors produced $568M of combined global revenue in 2025, their fifth year, against an identical “millions of patients” narrative — because single-agent response rates run 30–40%, median PFS is 5–7 months, resistance is rapid, and multiple entrants fragment the pool. Incidence-based sizing has over-predicted realized revenue in this exact target class by roughly an order of magnitude. A class-leading pan-RAS franchise plausibly supports $8–12B of global peak sales; a third entrant supports a small fraction of that.

Business Quality & Competitive Moat

Is the industry getting more or less competitive?

Decisively more. There are approximately 106 SHP2/SOS1/KRAS-directed agents in development. In pan-RAS specifically, Revolution Medicines is roughly four years ahead with an accepted NDA, 883 employees and over $3.9B of cash; BridgeBio Oncology and Boehringer Ingelheim compete directly with ERAS-4001 on the same 2H2026 readout timeline. RAS-directed business development in 2026 alone includes AstraZeneca/Jacobio (~$2.0B), AbbVie/Kestrel ($1.45B) and J&J/Firefly (~$1B). FACT.

The historical pattern is unambiguous: hepatitis C direct-acting antivirals went from about 2 candidates to about 30 in five years; JAK inhibitors from 1 to 9. First-in-class lead times in hot mechanism classes have collapsed from years to months.

How profitable is the business (ROIC, ROE)?

Not meaningful, and saying “N/M” alone would be evasive. There is no revenue; ROIC is undefined; reported ROE of −33% for FY2025 measures the pace of capital conversion, not business quality. The correct analog is realized return on in-licensing and R&D capital, and it is approximately −100%: roughly $931M consumed since inception has produced zero surviving assets from the 2018–2024 vintage. Five clinical programs entered; five were terminated.

How profitable is the industry — how many competitors, what barriers to entry?

Structurally poor, and this is one of the harsher industry verdicts we render. Top-20 pharmaceutical late-stage R&D internal rate of return was 7.0% in 2025 at a cyclical peak — and 38% of projected inflows are GLP-1s, an unrelated category flattering the number. Against a 15–25% hurdle that is aggregate value destruction. Cost per approved asset is roughly $2.3B over 10.5 years. Fewer than one in ten approved drugs recoups its cost of development, because approved-drug economics are power-law distributed.

Barriers to entry at the industry level are weak: Lumakras went from 100% of class revenue to roughly 64% in four years and is now being displaced by the third entrant — roughly 36 points of share movement in four years against Greenwald’s five-point-in-five-to-eight-years threshold for “no barriers.”

Can the business be easily understood?

Yes at the business-model level — it in-licenses molecules and runs trials — and no at the level that determines value. Whether ERAS-0015 infringes US 12,409,225 cannot be assessed by anyone outside the litigation, because neither the compound structure nor the claim scope is public. A binary, unmodellable risk sits under the majority of enterprise value. That is a genuine limit on diligence, not a failure of effort.

Can it be undermined by foreign low-cost labor?

Not in the conventional sense, but the question has an unusual and pointed application here. Both of Erasca’s clinical assets were in-licensed from Chinese biotechs — Joyo and Medshine — in a single month, for $22.5M of upfront consideration combined. The competitive threat from lower-cost drug discovery is not hypothetical for this company; it is Erasca’s own business model. The corollary is that the same channel is open to every competitor, and that Erasca’s sourcing advantage is not defensible.

Do brands matter?

Not in oncology. Prescribing is gated by NCCN guideline inclusion and by molecular testing, and oncologists re-optimize to the newest randomized data at essentially zero switching cost. The nearest analog to brand — a therapeutic-area franchise reputation that earns confidence in a company’s next molecule — is real in pharmaceuticals, but Erasca does not have one and Revolution Medicines is building one.

What is the nature of competition?

Competition on clinical evidence, speed to registration, and eventually price. The decisive variable is order of arrival with randomized data. Erasca’s own 10-K frames the landscape as “Coopetition” — “we view other companies in this sector more as potential allies and collaborators than as competitors.” That framing is directly contradicted by the market leader demanding Erasca cease all US activity on its lead asset. Flagged as a management-versus-evidence contradiction; we resolve in favor of the evidence.

Customers’ switching costs?

Zero — and the switching cost runs in the wrong direction. Once daraxonrasib enters guidelines and formularies it becomes both the standard of care and the comparator arm Erasca’s trials must beat. Its FDA expanded-access protocol in pretreated metastatic pancreatic cancer is already draining the trial-eligible pool Erasca needs to enroll from.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet?

Yes, and they are the whole company. Erasca capitalizes nothing — all in-licensing consideration is expensed immediately as in-process R&D. The Joyo licence, on which the large majority of the ~$5.66B enterprise value rests, was expensed as incurred ($12.5M in FY2024, $150.0M in Q1 2026) and carries zero balance-sheet value. Book equity of $393.5M at 31 March 2026 is essentially the marketable-securities portfolio. This is conservative accounting, and it means book value has no informational content about the asset.

Also unrecognized: federal NOLs of $413.8M and California NOLs of $576.7M, fully valuation-allowanced. They should not be capitalized — §382 ownership-change limitations are near-certain given the 2024 and 2026 offerings, and they only matter in a world where Erasca reaches profitability.

Off-balance-sheet liabilities?

Modest and well disclosed. Contingent milestone obligations: up to $57.5M development/regulatory plus $125.0M commercial to Joyo, and $30.0M plus $130.0M to Medshine, plus tiered royalties (low-to-mid single digits to Joyo, low single digits to Medshine). Operating-lease commitments of $60.4M undiscounted through 2032, carried at $40.7M.

The genuinely unquantified off-balance-sheet exposure is legal: the RevMed demand letter carries no accrual and no reserve, correctly under the accounting standard, but the potential remedies Erasca itself discloses include “injunctive relief or monetary damages.” The securities class action is likewise unreserved.

How conservative is the accounting?

Very. No capitalization of licensing or development costs, no goodwill, no intangibles to impair, no revenue recognition judgments to make, no going-concern qualification in company history. Auditor is KPMG. The one presentation issue worth flagging is not a violation but does obscure: the FY2024 10-K broke out naporafenib as a named R&D line ($51.2M FY2024, $31.6M FY2023); the FY2025 10-K folded it into an unnamed “Other clinical programs” bucket after the decision to kill it. Reconstructed, naporafenib absorbed $110M+ of direct R&D on top of ~$100M of licence consideration.

A second presentation point materially affects burn analysis: Erasca classifies cash paid for in-licensed IPR&D in investing activities, not operating. Reported operating cash burn of $27.4M in Q1 2026 excludes the $150.0M Joyo payment; true cash consumption was approximately $178M that quarter.

How CapEx-hungry is the business?

Barely at all — under $1M a quarter. Erasca owns and operates no manufacturing facilities and has “no plans to establish” any, relying entirely on contract manufacturers. Its only meaningful fixed commitment is the San Diego lease, which it has subleased in four pieces with $4.7M of impairments taken. The capital intensity of this business is entirely clinical-trial spend, not physical assets — which is why $1.78B of modeled development cost, not capex, is the binding constraint.

Capital Allocation & Management

How much FCF does the business generate, how does management use it, what is the philosophy?

Free cash flow is deeply negative and always has been — roughly −$97M a year on a clean all-in basis. The correct question for a pre-revenue biotech is inverted: how does management raise and deploy capital?

The philosophy, from the record: raise early, raise on data, never negotiate from weakness. Management has never let cash fall below ~$298M and has consistently raised with six or more quarters of runway remaining. Most tellingly, it issued nothing in 2025, burning the balance sheet down to $341.8M rather than printing stock at $1.06. That is genuine discipline and deserves credit.

Deployment philosophy is explicit and management-stated: “We pursue the best science in the world, regardless of its origin.” Erasca is an in-licensing vehicle, not a discovery engine. Discovery spend has been cut roughly 95% from $40.7M in FY2024 to a run-rate near $8M.

Significant acquisitions recently?

Not acquisitions of companies, but the licensing record is the substance:

The $150.0M Joyo territory payment is the single sharpest capital-allocation fact in the file. The contract priced that option at $50.0M if exercised before first Phase 2 dosing by either party, and $150.0M after. Erasca held it twenty-two months, its licensor dosed first, and Erasca paid the higher tier — $100M of unforced cost, when it held $341.8M plus $242.7M of fresh proceeds and was not capital-constrained. Guided runway was walked back from “into 1H2029” to “into 2H2028” within six weeks.

Buying back shares?

No, and it would be inappropriate. A pre-revenue company burning $100M a year should not repurchase stock. Correctly, Erasca has never done so and has no authorization.

Issuing large amounts of new shares to insiders?

Yes, and this is a real concern. 58,411,166 options outstanding at 31 March 2026 at a $4.66 weighted-average exercise price — 18.8% of shares outstanding, up 10.9M in a single quarter — plus 19,822,874 reserved. Total overhang 17–23% fully diluted, with evergreen provisions adding roughly 6% a year (5% equity plan, 1% ESPP) through 2031, board-discretionary and never shareholder-approved. Option intrinsic value at $19.11 is roughly $844M for a 103-person company; GAAP stock-based compensation of $24.9M a year badly understates the economic transfer because grant-date fair values were $1.24–1.31.

FY2025 grants of 15,078,786 options were struck at $1.75 — a 5.3% burn rate one quarter before a 19x move. In May 2024, 7,478,918 options were repriced to $2.35, properly excluding Section 16 officers and directors (a real mitigant) — but the named executives received fresh grants at $2.03, a lower strike, thirty days later.

Compensation policy of directors/management?

The weakest element of the governance file. Quoted verbatim from the 2026 proxy:

“For 2025, annual bonuses were based on corporate performance relative to key corporate objectives tied to clinical and regulatory priorities as well as operational and financial objectives. Following a comprehensive review… our compensation committee approved the payment of annual bonuses to our NEOs at 120% of target levels.”

That sentence appears essentially verbatim in all five proxies; only the percentage changes: 125% (FY2021), 90% (FY2022), 95% (FY2023), 115% (FY2024), 120% (FY2025). In five years as a public company, not one goal, weighting or threshold has ever been disclosed, and the payout has never fallen below 90% of target. It paid 115% in FY2024 — the year of an 18% reduction in force, three deprioritizations and raises at $2.06 and $1.85 — and 120% in FY2025, the year the lead Phase 3 asset was killed.

On equity, also verbatim: awards “vest over four years, subject to the employee’s continued employment.” One hundred percent of named-executive equity is time-vesting options. There have been zero performance share units and zero performance conditions, ever.

FY2025 totals: Lim $4,298,838; Chacko $1,778,496; Morris $1,750,898 — $7.83M aggregate, 5.6% of operating expense, with cash compensation only 1.9%. Modest in absolute terms; poorly structured.

There has never been a say-on-pay vote. As an emerging growth company Erasca provides no compensation discussion and analysis, no pay-versus-performance table, no pay-ratio disclosure and no advisory vote. EGC status lapses 31 December 2026, making the 2027 proxy a genuine event. Other governance: combined chair and CEO, classified board with two-thirds removal, six of eight directors independent — against a single share class, prohibitions on pledging and hedging, a Dodd-Frank clawback since 2023, no evidence of option timing around material information, and a board that voluntarily cut its own annual grants 75% in April 2026.

Motivations of management?

Mixed, and the evidence cuts both ways honestly.

Strongly aligned: CEO Jonathan Lim owns 36,574,772 shares — 11.6% of the company, roughly $700M — unhedged and unpledgeable, and has never sold a single share. He bought 2,300,000 shares on the open market for ~$4.71M of his own money in six transactions between December 2022 and December 2023, all into weakness, at prices from $1.69 to $4.99. Director Alexander Casdin bought 150,000 shares on the same pattern. That is genuine conviction, not optics.

Weakly aligned: the compensation structure above. And note that the Chief Medical Officer now owns zero shares outright and the General Counsel owns 25,076, against 625,000 and 560,000 fresh options granted in January 2026 — executive economic exposure is essentially entirely option-shaped, and therefore convex to volatility rather than symmetric with shareholders.

A signal that is not one, flagged to prevent a false conclusion: thirty-eight of forty-one Form 144 notices are the Erasca Foundation selling exactly 8,333 shares monthly since July 2023, liquidating the charitable donation made at the IPO — roughly 0.03% of shares outstanding a year, with zero information content. Actual insider selling totals 220,000 shares (0.06% of the company) across three transactions, all option-exercise-and-sell under a Rule 10b5-1 plan adopted 30 June 2024, when the stock traded near $1.90.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer?

No. Erasca is a Delaware C-corporation listed on the Nasdaq Global Select Market, single share class, common stock, par $0.0001. Ordinary Form 1099 treatment; no K-1, no UBTI, no ADR fees, no withholding complexity.

Dividend policy?

None, and none should be expected. Erasca has never declared a dividend and will not while it burns ~$100M a year. Any capital return would be a misallocation.

How profitable is the business?

Not profitable; see above. FY2025 net loss $124.5M; underlying operating loss normalized ~$117M a year. The company will not be profitable before 2031 at the earliest under any scenario in this file.

Is net income diverging from cash from operations?

Yes, dramatically, and understanding why matters. Q1 2026 net loss was $183.4M against operating cash use of only $27.4M — a 6.7x divergence. The reconciling item is the $150.0M Joyo territory payment, which is expensed on the income statement but classified in investing activities on the cash-flow statement.

The direction of the divergence is unusual and worth noting: for most companies a large net-loss-versus-cash gap signals aggressive accruals; here it signals the opposite — conservative immediate expensing of an asset purchase. Reported net loss overstates the operating burn. The honest measures are underlying operating loss (~$117M a year) and all-in cash consumption including investing-classified IPR&D. Anyone using the $183.4M quarterly loss or the $277.0M TTM loss as a run rate is wrong by roughly 4.5x.

Valuation context (no price target is offered): market capitalization ~$6.54B pro forma; enterprise value ~$5.66B, or ~$6.31B fully diluted for the option overhang. Conventional multiples are undefined — no revenue, no earnings, no EBITDA. The only definable own-history metric is price-to-book of 14.78, at the 97.5th percentile of Erasca’s own multi-year range — its richest ever on the one metric that can be computed, reflecting that the market is paying more per dollar of balance-sheet cash than at any point in the stock’s life. Own-history context only; never cross-sectional.

Risks & Downside

What factors would cause the stock to decline?

In descending order of expected impact:

1. Expansion-cohort data showing confirmed ORR below ~30% or median duration of response under ~8 months (guided 1H2027). The cleanest and most likely falsifier.
2. RevMed converting the demand letter into a filed complaint, particularly one seeking preliminary injunctive relief.
3. FDA feedback requiring randomization against an approved pan-RAS in 2L+ pancreatic cancer — adding years and hundreds of millions to the path.
4. A second treatment-related pneumonitis death at a go-forward dose, converting an idiosyncratic event into a class or dose signal.
5. ERAS-4001 disappointing in 2H2026, removing the only pipeline diversification.
6. Daraxonrasib launching to a Lumakras-like trajectory, which would re-base the whole category’s peak-sales assumption downward.
7. An equity raise below the July $17.50 placement price.
8. A sector de-rating — the biotech industry factor sits at a +2.5 standard-deviation reading, and a crowded factor plus a company-specific disappointment can arrive together.

Risk of a catastrophic loss?

Yes, and it is the base case in the bear scenario rather than a tail. Our modeled residual on failure is approximately $830M of equity — about 13% of today’s market capitalization, comprising roughly $800M of cash at the point of failure marked at 0.85x (busted biotechs historically transact at 0.7–1.0x net cash) plus ~$150M for the backups. That is a −87% outcome, and it is not hypothetical: the stock’s 2026 low was $3.49 and its 2025 low $1.06.

The timing asymmetry matters and is under-appreciated. If failure arrives late — at a 2030 pivotal readout — cumulative spend will be $1.1–1.3B and at least one more large raise will have occurred, so the residual per unit of ownership is materially worse than the figure above. Late failure is the expensive kind.

Chance of a total loss?

Low in the near term; non-trivial on a five-year view. A total loss requires ERAS-0015 to fail and ERAS-4001 to fail and cash to be exhausted. Against that: ~$896M of pro-forma cash, zero debt of any kind, no covenants, no maturity wall, no going-concern qualification in company history, and roughly four years of runway. There is no creditor ahead of equity and no mechanism for a forced wipeout. Even in failure, cash-shell value provides a floor well above zero.

The concentration qualifier, however, is real: the two clinical assets target the same pathway and — per the patent record — share a named inventor, so their failure modes are correlated rather than diversified. A pathway-level disappointment would impair both simultaneously. This is the extreme case of the correlated-mechanism risk that the standard pipeline-construction framework warns against.

Recent News & Events

Has the business environment changed recently?

Profoundly, in four ways, three of them adverse.

1. Target validation arrived — from a competitor. Revolution Medicines’ RASolute 302 reported median overall survival of 13.2 versus 6.7 months (HR 0.40, p<0.0001) in previously treated metastatic pancreatic cancer, was published in the New England Journal of Medicine, and had its NDA accepted 22 July 2026 under the Commissioner’s National Priority Voucher. This proves the mechanism works — and simultaneously closes the commercial white space and raises the bar Erasca’s own trials must clear.
2. The IP position came under attack — the 24 April 2026 RevMed demand letter (see the section on that subject above).
3. Safety acquired an asterisk — a treatment-related Grade 3 pneumonitis progressing to Grade 5 at a 24 mg recommended dose, disclosed 27 April 2026.
4. The balance sheet was transformed — from $341.8M at end-2025 to roughly $896M pro forma, funded to approximately 2030. The one unambiguously favorable change.

Significant acquisitions?

No corporate acquisitions. The material transaction was the $150.0M Joyo territory-option exercise in March 2026, taking worldwide rights including mainland China, Hong Kong and Macau. Two clinical-supply collaborations were signed with no economics to Erasca: Tango Therapeutics (March 2026) and Merck, pairing ERAS-0015 with KEYTRUDA (May 2026).

Change in accounting policies?

None. No changes to critical accounting policies were disclosed in the FY2025 10-K or the Q1 2026 10-Q. Erasca remains an emerging growth company availing itself of the JOBS Act extended transition for new accounting standards and is not subject to Section 404(b) auditor attestation of internal controls — both exemptions lapse at the end of FY2026. Management concluded disclosure controls were effective at 31 March 2026, with no changes in internal control over financial reporting.

The one presentation change worth flagging, discussed above, is the reclassification of naporafenib R&D from a named line into an unnamed “Other clinical programs” bucket between the FY2024 and FY2025 10-Ks.

Recent changes — new markets, facilities, management?

Markets: worldwide rights to ERAS-0015 including Greater China, acquired March 2026 for $150.0M.

Facilities: none added. The 78,000-square-foot San Diego headquarters has been carved into four subleases with $4.7M of impairments taken within thirty months, against $60.4M of undiscounted lease commitments to 2032.

Management: no changes to the three named executive officers — Jonathan Lim (Chairman and CEO), David Chacko (CFO and Chief Business Officer), Shannon Morris (Chief Medical Officer). Ebun Garner serves as Chief Legal Officer. Three Class II directors — Alexander Casdin, Julie Hambleton and Michael Varney — were re-elected on 26 June 2026, with withhold votes of roughly 12% of votes cast on Casdin. Headcount has been stable at 103 for two years following the May 2024 reduction in force.

Pipeline: the most consequential change of all — naporafenib development stopped and the Novartis licence terminated effective 3 June 2026, completing the turnover of every clinical program the company held before May 2024.

Supplemental appendix. Contains no investment recommendation and no price target. Nothing herein is investment advice.

APPENDIX B — Source Appendix

Report date: 2026-07-25 · All URLs accessed: 2026-07-25 unless noted

Conventions. P = primary source (filing, court document, patent grant, trial registry, issuer press release). S = secondary source (news, trade press, aggregator, law-firm marketing). Where a P and an S source cover the same fact, the memo cites the P source. Every non-obvious fact in the memo should map to a row below.

A. SEC Filings — Erasca, Inc. (CIK 0001761918)

All mirrored locally in output/ERAS/sources/ (see MANIFEST.csv); read in place.

B. SEC Filings — Third Parties

C. Court Records — Litigation

D. Patents & Intellectual Property

E. Clinical Trial Registry & Scientific Literature

Date-integrity flag (E3–E5). The comparator datasets are 2024–2025 data cuts benchmarked against Erasca’s 2026 cuts. Erasca’s own filings concede these are “cross-study comparisons… not based on any head-to-head clinical trials.” The comparison is the specific conduct RevMed calls “deceptive” and the securities complaint calls misleading. Treat every “exceeded comparator by X ppt” figure as management’s cross-study framing, not evidence.

F. Company Press Releases & Licensor / Counsel Disclosures

Reconciliation note (F1 vs. F4). Erasca’s filings and PR say $12.5M upfront + up to $176.5M milestones (= $51.5M dev/reg + $125.0M commercial). Joyo’s counsel describes up to $20M upfront/near-term + up to $345M milestones. The gap is presentation, not substance: the licensor aggregates the $150M territory-option payment and near-term milestones into its headline. The 10-K (A1/A2) governs.

G. Licensor Background — Guangzhou Joyo Pharmatech Co., Ltd.

H. Financial & Market Data Providers

I. News & Trade Press

J. Plaintiff-Firm Press Releases (marketing — use with care)

Near-identical releases from at least six firms describe the same single complaint (C1/C2). They are solicitation material, not filings, and several recycle April facts under July datelines. Cited only to establish the lead-plaintiff deadline, which is corroborated across all of them.

K. Internal / Google Drive

L. Verification Results — High-Stakes Claims

Each claim below was checked against the primary document named. Verified = the filing’s own words match the log’s assertion.

M. Flagged Items — Unsupported, Imprecise, or Stale Claims

Ⓕ1 — Date error risk on the fatal event (material). Some secondary coverage attributes the Grade 5 pneumonitis to the 2026-07-13 8-K (A5). The event was first disclosed on 2026-04-27 in the separate Item 8.01 8-K (A4, filed ~4:17 p.m. ET per the complaint, C2), and that disclosure — not the July one — drove the largest price event in the company’s history (H1: close $19.15 on 2026-04-27 → $9.90 on 2026-04-28, −48.3%; $21.49 → $9.90 over two sessions, −53.9%). The July 8-K repeats the identical footnote. The memo must date the disclosure to 2026-04-27; the securities class period ends 2026-04-26 precisely because of it.

Ⓕ2 — Secondary headline contradicts its own body. The AllSci page (I3) is titled “Revolution Medicines Sues Erasca Over RAS Patent” while its body states litigation has “not been initiated.” Do not cite the headline. Endpoints (I1) has it right: “threatens to sue.”

Ⓕ3 — Two paywalled/unverified news citations in the log. The Bloomberg headline (I6) and the Fierce Biotech “42% stock drop” piece (I5) could not be independently opened or date-verified. Both describe April 2026 events. Replace with A4 + H1 (the 8-K and the actual price series), which prove the point directly; or verify the dates before citing. A “42% drop” does not match the actual −48.3% single-day or −53.9% two-day move — do not put that number in the memo.

Ⓕ4 — Cross-study efficacy comparisons are management framing, not evidence. Every “exceeded comparator by N percentage points” claim rests on comparing Erasca’s 2026 data cuts to RevMed’s 2024–2025 published cuts (E3–E5). Erasca’s own 8-K carries a full-paragraph “Cross-Study Comparisons” disclaimer, and this exact practice is (a) allegation #3 in RevMed’s letter and (b) a pled theory in the securities complaint. Any memo section using these deltas must label them INTERPRETATION (management’s cross-study comparison), never FACT.

Ⓕ5 — Pooled US + China efficacy. The April headline ORRs (A4) pool AURORAS-1 with Joyo’s JYP0015M101. Erasca’s own risk language states the China data “are presented as received and have not been independently verified by the Company.” Any pooled figure carries a licensor-data-integrity caveat.

Ⓕ6 — “Runway into 2H 2028” is stale. Guidance from A1/A2 predates the ~$516M July raise (A6) and post-dates the $150M Joyo payment. Do not present it as the current runway; derive runway from pro-forma cash and the ex-IPR&D burn.

Ⓕ7 — Third-party quantitative data must be reconciled. The AZI valuation_index (H2), ROIC EV series (H3), and FactorsToday loadings (H5) are aggregated third-party estimates. The 44x EV move cited in the log is derived from H3, not from a filing; label it as such. ROIC’s “debt” in EV is unreconciled and is likely operating-lease liabilities — the log already flags this and it remains OPEN.

Ⓕ8 — Transcript coverage gap. ROIC’s most recent ERAS transcript is 2024-03-28 (H4) — ~28 months stale, predating the entire ERAS-0015 story. get_company_news returned empty. Management’s current framing (the April 27 and July 13 investor calls) is therefore absent from the evidence base. Any statement about “what management says now” must come from the 8-Ks/press releases or a transcript sourced from company IR or a public transcript service, or be flagged as an evidence gap.

Ⓕ9 — Licensor deal-value discrepancy. F4/F5 (Joyo’s counsel) quote “$20M upfront/near-term + up to $345M milestones” against Erasca’s “$12.5M + up to $176.5M.” Use the 10-K figures; note the difference only if discussing licensor-side framing.

Ⓕ10 — Law-firm press releases are not evidence of new events. J1–J6 carry July datelines but describe April facts, and several imply new developments where none occurred. Only C1/C2 establish the case. Multiple firms ≠ multiple lawsuits: one complaint is on file.

Ⓕ11 — “Erasca announced a US patent covering ERAS-4001” is not supported by the FY2025 10-K. The Valuation agent’s open question presumes an issued ERAS-4001 patent. Per A1/D7, the in-licensed Medshine family contained only two pending US non-provisional applications as of 2025-12-31 — no issued US patent. Erasca’s separately owned six RAS-inhibitor families are likewise all applications. Any claim of an issued ERAS-4001 patent needs a post-2025-12-31 primary source (USPTO grant or a 2026 8-K/press release) or must be cut. The underlying point still stands and is now documented: ERAS-4001’s IP estate is legally distinct from the contested Joyo estate.

Ⓕ12 — RVMD market-cap figures in the log are both wrong. The Industry agent flagged a conflict between $30.7B and ~$37.9B. Neither is right. RVMD’s own 10-Q cover states 212,596,462 shares outstanding as of 2026-05-01 (post-April raise) (B3); at the 2026-07-24 close of $189.23 that is a market capitalization of ~$40.2B. The $37.9B figure used the stale 31-Mar share count; $30.7B appears to be a stale ROIC snapshot. Use ~$40.2B, sourced to B1/B3 plus the 2026-07-24 close.

Ⓕ13 — “No DLTs” vs. a fatal TRAE is not a contradiction, and the memo must not blur them. The Industry agent’s open question is resolved by the primary filings: A4/A5 report zero dose-limiting toxicities (a protocol-defined, DLT-window construct) and one Grade 5 treatment-related pneumonitis, and zero TRAE-driven discontinuations. Both are true simultaneously because the fatal event fell outside the DLT-evaluation window and because “no discontinuations due to TRAEs” excludes a patient who died. The dual 24/32 mg RDE is consistent with FDA Project Optimus dose-optimization, not with a tolerability ceiling. Report all three facts together; reporting “no DLTs, 100% dose intensity, zero discontinuations” without the Grade 5 death would replicate exactly the disclosure the securities complaint attacks.

N. Open Questions the Record Does Not Answer

1. Provenance of the Joyo chemistry. RevMed’s letter alleges “a third party misappropriated RevMed’s alleged trade secrets in connection with a patent relating to ERAS-0015” (A3). Neither RevMed’s filings (B1), Erasca’s filings (A1–A5), the patent record (D1, D3), nor any public reporting names the third party or describes the alleged chain of transfer. No public evidence links any RevMed personnel to Joyo. This is an allegation, unadjudicated and unelaborated, and must be labeled as such throughout the memo. The only reportable facts: RevMed’s '225 priority date (2020-09-15) precedes Joyo’s '647 priority date (2022-09-29) by ~2 years (D1, D3).
2. Who prosecutes and controls the Joyo patent estate. Partly resolved. The License Agreement itself (A14) has now been read: Joyo disclaims any warranty of validity or non-infringement, licenses the IP “as is, with all faults,” accepts no liability for its handling of prosecution/enforcement/defense, and gives Erasca no general third-party-IP indemnity — only a knowledge-qualified representation (§11.2(o)) enforceable in arbitration against a private PRC counterparty. Still open: the identity of any upstream “Joyo In-License Agreement” counterparty (Exhibit 1.77, not publicly filed) — relevant because RevMed’s unnamed “third party” need not be Joyo itself; and whether Erasca has asserted a §12.2(b) claim against Joyo.
3. Greenshoe exercise. No 8-K or 424B5 as of 2026-07-25 confirms exercise of the 4,714,285-share option; the 30-day window runs to ~2026-08-12. Pro-forma share count should be stated as a range.
4. Cumulative naporafenib spend and any termination consideration. A1 discloses the termination and that no IPR&D was recorded in FY2024/FY2025 for Novartis, but not the cumulative program cost or any payment received. Not separately disclosed.
5. RevMed’s next step. As of 2026-07-25 there is no docket, no 8-K, and no public statement beyond B1’s risk-factor sentence. Whether RevMed files, and where, is unknowable from the record.