Fastly, Inc. (NASDAQ: FSLY) — A Commodity CDN’s Real Turnaround, Re-Rated Before It’s Proven — With Free Cash Flow Paid in Stock

Independent equity research note Report date: 2026-07-04 · Sector: Information Technology · Internet Software & Services (Edge Cloud / CDN) Primary sources: SEC filings (FY2025 10-K filed 2026-02-25; Q1-2026 10-Q; DEF 14A 2026-04-22), the Q1-2026 earnings call (2026-05-06), and public market data. All figures reconciled to filings unless flagged.

⚡ Author’s Take

This block is the author’s own subjective opinion and general information only. It is not investment advice. The analytical body that follows takes no position and carries no price target — this block is the single exception.

Verdict: HOLD / not-a-short / accumulate-on-real-weakness sub-$12. Conviction: medium. Tag: “The CDN that finally learned to cross-sell — repriced before it proved the model, with its free cash flow written in stock.”

Fastly at ~$18 is a genuinely better company than it was two years ago, priced as though the turnaround is already de-risked. The good news is real and not manufactured: revenue re-accelerated from a +7% trough (FY24) to +15% (FY25) and +20% in Q1-26; security is compounding ~47% and is now a fifth of revenue; net retention has climbed from 100% to 113%; RPO is up 63%; and the company is, for the first time, self-funding. That is a legitimate inflection and it deserved a re-rating off the $5 capitulation low. But three things stop me from paying up here. First, ~77% of revenue is still commodity delivery growing ~11% into structural mid-single-digit price deflation — the melting core has not gone away, it has merely been out-run for a few quarters. Second, the “profitability” and the “free cash flow” are both substantially an accounting artifact of adding back $117M of stock-based comp (≈19% of revenue): net of SBC, the business burned ~$52M of real owner cash in FY25, and the share count is up ~54% since 2020. Third, and decisively for the price, at ~4.4x forward EV/sales and ~60x forward non-GAAP EPS the stock trades right on top of profitable Akamai’s multiple while being GAAP-lossmaking — and its own-history valuation percentile is now mid-range (≈42nd on P/S), not the cheapest-ever setup that made it interesting at $5–8.

So this is a momentum-driven turnaround that has already earned its first-leg re-rate and now needs to prove it — durable NRR above 110% through a delivery-deflation cycle, security scaling to a real second leg, and cash comp displacing SBC — before the next leg is underwritten by anything but narrative. I’d own it on weakness (sub-$12, roughly 3x forward EV/sales, where you’re paying for the mix-shift optionality rather than its success), trim into strength above the low-$20s, and I would not short it — the operating trend is genuinely up and the float is a high-beta squeeze magnet. Flips bullish if security + compute clears a $250M+ run-rate with NRR holding >112% and SBC falls toward 10% of revenue (real FCF turns positive). Flips bearish if network-services growth rolls back toward low-single-digits or a large customer (ByteDance ≈10% of revenue) optimizes away, exposing the core as still-melting under a full multiple.

📈 Stock Price Action — Five-Year Event Map

Factual price history — not a recommendation. Price moves are FACT (AZI daily CSV); attributed causes are INTERPRETATION.

The arc. Over the trailing ~60 months Fastly has round-tripped violently. It IPO’d at $16 (May 2019), became a COVID edge-cloud/CDN darling and closed at an all-time high of $131.45 (14-Oct-2020), then de-rated for three-plus years to a capitulation low of $4.92 (07-Apr-2025). A 2026 turnaround rip carried it to a 52-week high of $34.05 (08-Apr-2026), roughly half of which has since been given back. At $18.13 (02-Jul-2026) the stock is ~86% below its 2020 all-time high, ~47% below its April-2026 52-week high, and ~78% above its year-end-2025 close of $10.18. The 52-week range is $6.38 – $34.05. The widely-cited “doubling in 2026” is real but incomplete: the honest description is a triple off the bottom followed by a halving — a round trip, not a clean breakout.

Cycle narrative. (1) The IPO-to-ATH melt-up was the purest expression of the 2020 “everything that benefits from internet traffic” trade. (2) The unwind began the moment Fastly disclosed that a single large customer (TikTok/ByteDance) had reduced usage and cut guidance — the first public tell that this was a whale-concentrated consumption business, not a durable subscription compounder. (3) The June-2021 outage briefly advertised Fastly’s centrality to the internet, then reminded enterprises it was a single point of failure — accelerating multi-CDN adoption that would later cap pricing power. (4) The 2021–22 collapse was macro (rates) plus fundamental (decelerating growth, no profitability, revolving-door management). (5–6) The 2024 guide cut and 2025 capitulation marked maximum pessimism — the stock traded below $5 at ~1.5x sales. (7) The 2026 rip is the FCF-inflection/edge-AI narrative repricing a left-for-dead name; the Q4-25 print (record revenue, positive FCF) was the catalyst. (8) Q1-26 was strong but “priced in” — the stock sold off ~28% on the print and has drifted with the broader high-beta unwind. Earnings-date moves cross-referenced to 8-Ks: Q3-25 (2025-11-05), Q4-25 (2026-02-11), Q1-26 (2026-05-06).

1. Executive Summary

Fastly is a subscale edge-cloud / content-delivery-network (CDN) operator that spent 2021–2024 as one of the most conspicuous casualties of the post-COVID growth unwind, and has spent the last four quarters staging a credible operational recovery. The investment question is no longer “is this company falling apart?” — it plainly is not — but “is a genuine but early mix-shift, funded by stock and run from behind against a far larger rival, worth a multiple that already resembles a profitable incumbent’s?”

What is working (FACT). Revenue re-accelerated from a +7.4% trough in FY24 to +14.8% in FY25 ($624.0M) and +20% in Q1-26 ($173M) — the first acceleration in five years. Security revenue grew +21% in FY25 and +47% in Q1-26 to 22% of the total; compute (“other”) grew +64%/+67%. Trailing-twelve-month net retention recovered from 102.3% (FY24) to 110.1% (FY25) to 113% (Q1-26). RPO rose +45% in FY25 and +63% year-over-year in Q1-26. The company generated positive reported free cash flow (+$65.8M FY25) and positive non-GAAP operating profit, and guided FY26 to $710–725M revenue (+15%), non-GAAP operating profit of $58–68M, and FCF of $40–50M.

What is not (FACT / INTERPRETATION). The company remains GAAP-unprofitable: a −$121.7M net loss in FY25 (−$0.83 EPS), a −$40.7M GAAP EBITDA, and a −19% GAAP operating margin. The positive “adjusted EBITDA” (+$77.4M) and positive “free cash flow” (+$65.8M) are both produced almost entirely by adding back $117.3M of stock-based compensation — 18.8% of revenue. Net of SBC, FY25 owner free cash flow was −$51.5M; the share count has risen ~54% since 2020. Roughly 77% of revenue is still commodity delivery growing only ~11–12% into structural mid-single-digit annual price deflation, with the customer base concentrated (top-10 = 34% of revenue; ByteDance-affiliated units ≈ 10%) and the total logo count flat at ~3,092.

Business quality. No durable structural moat (§4). Fastly’s delivery product is technically excellent but non-defensible — multi-CDN sourcing lets customers reroute traffic instantly, which is precisely why net retention collapsed in 2022–24 when whales optimized their spend. A moat is forming in security and compute (customers build WAF rules and edge code on the platform), but it is nascent, ~23% of revenue, and contested by Cloudflare (NET), a rival ~3.5x Fastly’s size, growing ~30%, near GAAP breakeven, and carrying a market capitalization roughly 28x Fastly’s.

Valuation. At ~$18, EV ≈ $3.0B, ~4.8x trailing / ~4.1x forward EV/sales, ~60x forward non-GAAP EPS, ~1.6% forward FCF yield. That is essentially Akamai’s multiple (~5x EV/sales) for a company with none of Akamai’s profitability, and it carries none of Cloudflare’s ~31x growth premium. On its own five-year history the stock sits mid-range (≈42nd percentile P/S), not cheap. Rule-of-40 is ≈25.

Bottom line (memo, position-free). A real turnaround at a full price, with reported cash generation that is materially an SBC artifact. The embedded expectation is durable low-teens growth with gradual margin expansion — plausible but unproven, and no longer discounted.

2. Business Overview

What Fastly does (FACT). Fastly operates an edge-cloud platform: a globally distributed, software-defined network of points-of-presence (POPs) that sits between end-users and its customers’ origin servers. Off that single network it sells three product families:

1. Network Services (delivery) — the legacy CDN: dynamic site acceleration, caching with ~150ms global instant purge, load balancing, image optimization, video/live streaming, DNS, and origin shielding. This is what makes websites and apps fast.
2. Security — anchored by the Signal Sciences next-generation Web Application Firewall (WAF) acquired in 2020 (~$775M), plus DDoS protection, bot management, API discovery/schema enforcement, TLS, and the newer ContentGuard (AI-bot content protection) and advanced rate-limiting products.
3. Compute / “Other” — Compute@Edge, a WebAssembly-based serverless runtime that executes customer code at the edge; observability; and professional services.

The company was incorporated in 2011 (originally SkyCache), IPO’d in May 2019, is headquartered in San Francisco, and employed ~1,140 people at year-end 2025. It transferred its listing from the NYSE to Nasdaq in December 2025.

Revenue model (FACT / INTERPRETATION). Fastly is fundamentally consumption-based: customers sign contracts with minimum commitments but pay largely on actual usage — bandwidth delivered, requests served, compute executed. This is why management reports trailing-twelve-month metrics and repeatedly cautions that quarterly revenue “can fluctuate frequently.” It is partly recurring (minimum commitments, now reflected in RPO of $354M) but is materially more variable than a seat-based SaaS subscription — a structural reason the stock de-rated so violently when usage from a few whales fell.

Revenue by product line, FY2025 vs FY2024 (FACT, 10-K):

The mix-shift is the whole bull case in one table: ~77% of revenue is commodity delivery growing ~12% (usage intensity from existing large customers), while the stickier, higher-value security and compute lines grow 21% and 64% — but together are only ~23% of the business. The shift is real but early.

Geography (FACT). United States $465.5M (74.6% of FY25 revenue); rest-of-world $158.6M (25.4%); no single non-US country exceeds 10%. Management is investing to expand Asia-Pacific (new Singapore office, regional leadership hires) where it is underpenetrated.

Customers / KPIs, FY2025 (FACT). Total customers 3,092 (up just 31 net — essentially flat); “enterprise” customers (>$100k trailing-twelve-month revenue) 628, generating 94% of revenue (average enterprise customer ≈ $934k). Gross annual-revenue retention 98.7% (very low churn); LTM net retention 110.1% (up from 102.3%); RPO $353.8M (+45%). In Q1-26 the company stopped reporting total customer count (arguing >90% of revenue is from large customers) and disclosed 634 large customers, 113% NRR, and $369M RPO (+63%). Affiliated business units of a single company (ByteDance/TikTok) aggregated 10% of FY2025 revenue — a live single-name concentration risk, sharpened by the January-2026 announcement of a ByteDance US joint-venture restructuring.

Verdict (§2). A subscale, consumption-revenue, whale-concentrated infrastructure business in a genuine but early transition from commodity delivery toward security and edge compute. The NRR recovery and RPO growth are real green shoots; the flat logo count and 94% enterprise-revenue concentration expose the absence of a bottom-up customer-acquisition engine.

3. Industry Dynamics

Fastly competes across three arenas with very different economics, and it is heaviest in the worst one.

Arena 1 — Content delivery (deflationary; where Fastly is ~77% exposed). Raw CDN delivery is a commoditizing utility. The price customers pay per gigabyte delivered falls at a double-digit annual rate as bandwidth, hardware, and interconnect costs decline and competition intensifies; management confirmed mid-single-digit year-over-year price erosion on network services (only) in Q1-26, offset by mid-20s% traffic-volume growth. A pure-delivery CDN must therefore grow volume faster than price falls simply to stay flat — a treadmill. That Network Services grew only +12% in FY25 despite strong traffic growth is the treadmill made visible. The entire legacy CDN cohort is fleeing this arena: Edgio (formerly Limelight/EdgeCast) filed for bankruptcy in 2024; StackPath wound down. Supply is exiting at the bottom of the cycle — a partial positive for survivors like Fastly and Akamai.

Arena 2 — Edge security (better economics; crowded). WAF, DDoS, bot management, and API protection carry higher gross margins, stickier switching costs (customers encode security policy into the platform), and are not subject to the same per-GB deflation. This is where the industry’s profit pool is migrating and where Fastly is growing fastest (+47% in Q1-26). But it is a crowded field — Cloudflare, Akamai, Imperva/Thales, F5, and the cloud hyperscalers all compete — and scale advantages accrue to the largest platforms.

Arena 3 — Edge compute (nascent; high-optionality). Serverless compute at the edge (Fastly Compute@Edge, Cloudflare Workers) is early and small (~3% of Fastly revenue) but is the highest-optionality arena, particularly if AI inference — as opposed to centralized training — migrates toward distributed, low-latency edge nodes. This is genuine call-option value, but it is unproven and dominated in mindshare by Cloudflare’s developer ecosystem.

The hyperscaler / free-tier overhang (FACT). Delivery is a bundled loss-leader for Amazon CloudFront, Google Cloud CDN, and Azure Front Door — and free at Cloudflare’s entry tier. These players do not need CDN to be profitable; they use it to pull higher-margin compute and egress. This structurally caps independent CDN pricing power and is existential for subscale pure-plays.

Capital cycle (Marathon lens, INTERPRETATION). Capital is still flooding into the broad edge/CDN space — hyperscaler edge capex, Cloudflare reinvesting a $2.2B revenue base, and low-cost entrants such as Bunny.net (~$0.005/GB). Capital inflow plus falling incremental returns is the classic setup for competed-away economics. The one countervailing signal is supply exit at the bottom (Edgio, StackPath) plus competitors (Akamai) raising delivery prices to pass through 2026 DRAM/memory-cost inflation — a cyclical, not structural, reprieve. Notably, Fastly is declining to pass memory costs through, betting that its capital efficiency lets it hold price and take share.

Barriers to entry (Greenwald lens, INTERPRETATION). Low and falling for delivery; higher for security and credible edge compute — but those higher barriers accrue to the scale leader (Cloudflare), not to a #3–#4 pure-play.

Verdict (§3): structurally BAD industry for a subscale, delivery-weighted pure-play. The only structurally attractive pockets — security and edge compute — are precisely where Fastly is smallest and where the scale advantage is pulling away from it. The industry rewards platform scale; Fastly does not have it.

4. Competitive Position — The Moat Question

This is the crux of the thesis, and the honest answer is unflattering: Fastly has no durable structural moat over the ~77% of its business that is delivery, and only a nascent, contested one over the ~23% that is security and compute. Working through Greenwald’s taxonomy of the three genuine advantage types:

Supply / cost advantage — ABSENT (arguably negative). Fastly is subscale relative to Akamai (~$4B revenue), the hyperscalers, and Cloudflare ($2.17B). Its architecture — fewer, larger, denser POPs running on modern software-defined infrastructure — drives high cache-hit ratios and a record 57.1% GAAP gross margin, and management argues this makes it “more capital-efficient” per unit of traffic. That is credible and is a genuine mitigation of its scale disadvantage — but “efficient for our size” is not a cost advantage over larger rivals. It does not let Fastly underprice a hyperscaler that treats delivery as a free feature.

Demand / customer captivity (switching costs) — WEAK in delivery, MODERATE in security/compute. Multi-CDN sourcing is standard practice among large enterprises — they deliberately run two or more CDNs and shift traffic between them in real time for resilience and price leverage. A delivery customer can move volume to a competitor essentially overnight. This is not theoretical: it is exactly what happened in 2022–24, when net retention fell below 100% as large customers “optimized” their Fastly spend. Real switching costs exist only where customers build on Fastly — WAF rule sets, VCL/Compute@Edge code, observability integrations — i.e., in the ~23% of revenue that is security and compute, and even there the stickiness is earned account-by-account, not structural.

Economies of scale + captivity (Greenwald’s only durable moat type) — NOT present at Fastly; present and widening at Cloudflare. The durable version of a platform moat is a self-reinforcing flywheel: a large user base (Cloudflare’s millions of free-tier developers) → mindshare → ecosystem → workloads → lower unit costs → still more users. Cloudflare has this; Fastly does not. Fastly’s top-down, enterprise-sales motion generates no bottom-up developer flywheel — and the flat 3,092 total-logo count is the direct evidence. A moat you can prove by a KPI that isn’t growing is not a moat.

Technical differentiation is real but non-defensible (FACT / INTERPRETATION). Fastly is genuinely well-regarded by developers — instant global purge, real-time configuration (VCL), a strong WebAssembly compute runtime, excellent documentation, and a top rating in the 2026 Forrester Wave for Edge Development Platforms (named one of only two Leaders). But elegance that customers can (a) get “good enough” more cheaply or bundled elsewhere and (b) hedge via multi-CDN does not convert into pricing power. The definitive test: the 2022–24 revenue stall happened with the technical edge fully intact.

Reliability / brand (FACT). The 08-June-2021 outage — a single Fastly software bug that briefly took down large parts of the internet (Amazon, Reddit, the UK government, the NYT, Shopify) — simultaneously advertised Fastly’s centrality and its single-point-of-failure risk, reinforcing enterprise multi-CDN behavior. It cuts both ways: on the Q1-26 call, management cited winning a security cross-sell because a competitor had a high-profile outage.

Head-to-head scorecard (FACT):

Cloudflare out-grows Fastly roughly 2:1 and is closer to GAAP profitability — the worst possible combination for a smaller rival — and the market awards it ~28x Fastly’s capitalization on ~3.5x the revenue. Akamai is the low-growth, profitable “melting-ice-cube” legacy comp. Fastly sits awkwardly between them: too subscale and unprofitable to be Cloudflare, too growth-dependent to be valued like Akamai.

Verdict (§4): no durable structural moat. A real but non-defensible technical edge in commodity delivery, and a genuine but early, subscale position in security/compute. The entire investable question is whether the security/compute mix-shift and NRR recovery can compound faster than delivery commoditizes — a race run from behind against a competitor carrying ~28x the market’s confidence.

5. Growth History and Forward Opportunities

History (FACT). Fastly’s revenue trajectory is a near-perfect map of the growth-stock cycle:

The 2020–21 hyper-growth was substantially a whale-driven, COVID-inflated phenomenon; the 2022–24 deceleration was that same concentration reversing (large customers renegotiating and optimizing) plus secular price deflation. The FY25 re-acceleration to +14.8%, extended to +20% in Q1-26, is the genuinely new development — and it is being driven by the right things: security (+47%), compute (+67%), net-retention recovery (100% → 113%), and RPO (+63%). The COVID-era dollar-based net-expansion peaked around 118–147%; today’s 113% is a healthier, less-concentrated number even if lower.

Forward opportunities (FACT / INTERPRETATION).

• Security cross-sell (the primary near-term driver). Fastly has expanded from essentially one security product (the Signal Sciences WAF) to five or six (DDoS, bot management, API discovery/schema enforcement, ContentGuard, advanced rate-limiting). Management reports that in Q1-26 nearly half of security instances sold to new customers were of these newer products, and that customers increasingly land with multiple security products at once — the mechanism by which security can become a genuine second growth leg. Management targets security + compute exceeding a $200M annual run-rate by late 2026, and says >50% of FY26 incremental revenue will come from security + other.
• Edge compute + AI (the optionality). Compute grew +67% in Q1-26 (largest-ever sequential step-up). Management frames AI/agentic traffic as a multi-vector tailwind: volume (agent traffic flowing through the network), security attach (protecting AI workloads, privacy/VPN, API governance), and — the biggest call option — edge inference, where Fastly argues its distributed architecture (ill-suited to centralized training) could be well-suited to low-latency inference. This is real optionality but explicitly early (“design-partner” stage) and unquantified — management could not break out what percentage of traffic growth is AI-driven.
• Geographic expansion — APAC underpenetration, new Singapore office, new CMO (Joan Jenkins) to build brand beyond the technologist audience.

Quality of growth (INTERPRETATION). Improving but mixed. The composition is high-quality (mix-shifting toward stickier, higher-margin security/compute; NRR-driven expansion within a low-churn base; larger upfront RPO commitments). The core is low-quality (commodity delivery on a price-deflation treadmill, ~77% of revenue, growing ~11%). The flat total-logo count means all growth is intensity, not breadth — leaving the model dependent on a small number of large customers spending more.

Verdict (§5): mixed-to-improving quality growth. The re-acceleration is real and led by the right products, but it rests on a narrow, whale-concentrated base with a commoditizing core; the durability of >110% NRR through a full delivery-deflation cycle is unproven.

6. Financial Quality

Revenue and margin trajectory (FACT). Revenue $624.0M FY25 (+14.8%). GAAP gross margin recovered to 57.1% (from 54.4% FY24, and a 48.5% trough in FY22) — genuine operating leverage on traffic, but still below the 59–75% of profitable peers, a reminder that the core is commodity. Non-GAAP gross margin ran ~62% for FY25 and hit a record 65.1% in Q1-26 (though ~190bps of that was a one-time server-useful-life accounting change; FY26 guidance is 64%). GAAP operating margin improved to −19.0% (from −28.3% FY24 and −61.8% FY21) on genuine opex discipline (two rounds of layoffs across 2022–23) — a real, multi-year cost-out story.

The SBC / “adjusted” gap — the central quality-of-earnings issue (FACT / INTERPRETATION). Management’s headline profitability and cash-flow figures are positive; the GAAP figures are not; and the entire bridge is stock-based compensation.

Stock-based compensation was $117.3M in FY25 — 18.8% of revenue (and $107.9M / $136.3M / $145.8M in the three prior years — chronically ~20–30% of revenue). SBC is a real economic cost: it transfers ownership from existing shareholders to employees, and Fastly’s share count has risen from ~104M (2020) to ~160M diluted (2026), roughly +54%. The critical implication is for the celebrated “free cash flow”:

The trend is genuinely improving — the real cash burn has more than halved in two years — but on an owner-economics basis Fastly is still burning ~$50M/year, not generating $66M. Any valuation framing that calls the stock “cheap on ~19x EV/FCF” is relying on an SBC mirage. This is the single most important number in the memo.

Returns on capital (FACT). With a −$1.114B accumulated deficit and persistent GAAP losses, ROE is deeply negative (~−13%) and ROIC is negative. Economics are improving with scale (the margin trajectory is unambiguous) but have not yet crossed into positive returns on capital. State plainly: this is not yet a company that earns its cost of capital.

Balance sheet (FACT). Cash + short-term investments $361.8M at YE25; current ratio 2.6x. Total debt ~$430–437M (convertible notes $368.6M par + $68.3M finance leases). Net debt is near-neutral. Goodwill is $670M — 72% of total equity — almost entirely the 2020 Signal Sciences acquisition, unimpaired. Working capital: DSO is elevated (cash-conversion cycle ~52 days) but manageable; deferred revenue and RPO are growing (RPO +45%), a positive leading indicator.

Runway (FACT). The company is now self-funding on a reported basis (positive CFO $94.4M FY25) even as it burns modestly on an SBC-adjusted basis; with $330M cash (Q1-26) and no near-term maturity crunch, liquidity is not a concern over the forecast horizon.

Verdict (§6): improving but not yet good. Real operating leverage and a halving of true cash burn in two years — but still GAAP-lossmaking, still burning ~$50M net of SBC, and still diluting shareholders ~5–6%/year. Economics improve with scale; they have not yet arrived.

7. Capital Allocation

Fastly’s capital allocation splits cleanly into a competent half (liability management) and an undisciplined half (equity dilution), under a management team that has turned over three times at the top in two years.

Debt / convertible management — competent (FACT). In 2021 Fastly issued $948.8M of 0%-coupon convertible notes (conversion price ~$102.80) at the top of the growth-stock bubble. As the stock collapsed, management opportunistically repurchased the notes well below par — e.g., retiring $235M face for ~$176.4M cash, and $367.3M face for ~$309.1M — shrinking the obligation to $368.6M par by YE25 and capturing a real discount for shareholders. This is genuinely skillful, value-additive liability management. Two caveats: (1) the newer 7.75% 2028 notes carry a meaningful cash coupon (interest expense rose to $12.7M), and (2) the new conversion strikes ($15.26 / $19.74) are far lower than the old $102.80 — i.e., far more dilutive if the stock rises. The $39M current portion due March 2026 was retired on schedule; remaining maturities are 2028 and 2030.

M&A (FACT / INTERPRETATION). The defining deal is Signal Sciences (2020, ~$775M, mostly stock) — expensive, but it bought Fastly its entire security franchise, which is now the growth engine and the best strategic asset in the portfolio; the $670M of resulting goodwill is unimpaired, which the security line’s growth supports. Smaller: Glitch (2022, developer community — modest). No large, value-destructive roll-up; M&A discipline is acceptable.

R&D and S&M intensity (FACT). R&D ~26% of revenue and S&M elevated — appropriate for a platform investing in a mix-shift, but a drag on near-term margins and a reason GAAP profitability remains elusive.

Buybacks vs. dilution (FACT). There is no meaningful buyback (a token $10.7M repurchase in 2022). Against ~$117M/year of SBC and a ~54% rise in share count since 2020, the net equity story is chronic dilution. This is the weakest element of capital allocation.

Insider behavior — SEC Form 4/144 sweep (FACT). Across the trailing 60 months, ~395 Form 4s and ~191 Form 144s show a uniform pattern: routine RSU vesting sell-to-cash, 10b5-1-planned sales (e.g., President Scott Lovett selling into the $17–22 rally), and director stock grants (code A). There are zero discretionary open-market purchases (code P) — no insider conviction buying anywhere in the file, even at the sub-$5 capitulation low. Net insider flow is steadily negative. Additionally, founder Artur Bergman has pledged ~2.27M shares as collateral for personal credit — a modest but noted governance flag. The insider read is neutral-to-slightly-negative: no alarm-bell dumping beyond ordinary equity-comp mechanics, but no conviction signal whatsoever.

Management and incentives (FACT). Current CEO Charles “Kip” Compton was promoted from Chief Product Officer effective 16-June-2025, the same day predecessor Todd Nightingale resigned (his awards forfeited) — the third CEO in roughly two years, a real execution-continuity concern. CFO Rich Wong joined August 2025; a new CMO (Joan Jenkins) joined Q1-26. CEO total compensation (~$5.69M, a reasonable 22.6:1 CEO-to-median ratio) is moderately aligned and improving: PSUs now include a relative-TSR gate versus the Russell 2000, and for 2026 add FCF and committed-monthly-recurring-revenue (CMRR) gates. But the STI/LTI metrics are non-GAAP with no ROIC gate, and the 2025 short-term plan paid out at 135.4% of target in a year the company posted a GAAP net loss — a misalignment between “bonus-worthy” adjusted metrics and actual shareholder economics.

Verdict (§7): mixed. Competent, value-additive on the liability side (convert deleveraging); undisciplined on the equity side (chronic SBC dilution, no offsetting buyback); acceptable on M&A. The revolving-door CEO seat and adjusted-metric-driven bonus payouts in a loss-making year are governance concerns that temper an otherwise-improving execution picture.

8. Changes and Headwinds — Last Two Years

Strategic / operational changes (FACT).

• Leadership overhaul. Three CEOs in ~two years (Bergman → Nightingale → Compton, June 2025); new CFO (Wong, Aug 2025), new CMO (Jenkins, Q1-26), Scott Lovett elevated to President, Go-to-Market. A near-complete C-suite refresh — disruptive, but the new team owns the FY25–26 re-acceleration.
• Product expansion. Security portfolio widened from one product to ~six; ContentGuard (AI-bot content protection) and enhanced API security launched in Q1-26; Compute@Edge language support completed; a “Fastly Agent Toolkit” for AI coding agents released.
• Metric changes. Fastly stopped reporting total customer count (Q1-26), shifted capex reporting to infrastructure-only, and changed server-useful-life accounting (a ~190bps one-time GM benefit) — a cluster of presentation changes that flatter comparability and warrant a skeptical eye.
• NYSE → Nasdaq transfer (Dec-2025) — cosmetic, aligning with a tech-growth index cohort.

Headwinds (FACT / INTERPRETATION).

• Memory-cost inflation. DRAM/memory prices rose 2–3x, raising Fastly’s capex; management is front-loading FY26 infrastructure capex to 10–12% of revenue (~$70–80M) versus 5% in 2025 — the direct reason FY26 free cash flow is guided flat-to-down ($40–50M) despite rising adjusted profit. Competitors are passing memory costs into delivery pricing; Fastly is not, betting on capital efficiency to hold share.
• Price deflation on network services (mid-single-digits YoY) — structural, ongoing.
• Customer concentration — ByteDance-affiliated units ≈10% of revenue, with a January-2026 US-JV restructuring introducing traffic-retention uncertainty; top-10 = 34%.
• Competitive intensity — Cloudflare out-growing and better-capitalized; hyperscaler bundling.

Verdict (§8): net neutral-to-slightly-positive for the thesis, with real risks. The operational changes (product breadth, GTM discipline, cost-out) have driven the re-acceleration and are genuine positives; the leadership churn, capex/memory headwind, and concentration are meaningful offsets. The environment has improved from 2024’s crisis but is not benign.

9. Risk Analysis

Catastrophic-loss risk: low in the near term — the balance sheet is near net-neutral with $330M cash and no maturity crunch before 2028, so bankruptcy risk is remote. The realistic downside is not a zero but a valuation de-rate back toward 2–3x sales (i.e., the high-single digits) if the turnaround stalls, plus ongoing dilution.

10. Valuation Discussion (Embedded Expectations)

Where it trades (FACT). At $18.13, market cap ≈ $2.9B; enterprise value ≈ $3.0B (near net-neutral balance sheet). That is:

• ~4.8x trailing / ~4.1x forward EV/sales (FY26 revenue midpoint $717M).
• ~8.3x EV/gross profit (FY25 GAAP gross profit $356M).
• ~60x forward non-GAAP EPS (FY26 guide $0.27–0.33) — and not meaningful on GAAP (loss-making).
• ~1.6% forward FCF yield (FY26 FCF guide $40–50M) — and negative net of SBC.
• Rule-of-40 ≈ 25 (15% growth + ~9% non-GAAP operating margin) — below the 40 threshold; on a true-FCF-margin basis it is lower still.

Own-history context (FACT). On AZI’s own-history percentiles, FSLY sits at the 42nd percentile on P/S, 57th on P/B, ≈49th composite — i.e., mid-range, not cheap, and a notable departure from the “cheapest-ever multiple” setups elsewhere. The five-year EV/sales range spans ~13–43x (the 2020–21 bubble) down to ~2–3x (the 2022–24 trough); today’s ~4.8x is well off both extremes.

Comp-set bracket (FACT).

• Akamai (AKAM): ~5.0x EV/sales, +5% growth, ~15% GAAP operating margin, profitable, ~16.5x EV/EBITDA. Fastly trades on top of Akamai’s sales multiple while being unprofitable and higher-growth — the market is paying an Akamai multiple for a not-yet-Akamai business, betting the growth justifies the missing profitability.
• Cloudflare (NET): ~30.8x EV/sales, +30% growth, ~75% gross margin. Fastly gets none of this premium — correctly, given the growth and moat gap, but it frames the ceiling: to earn even a fraction of NET’s multiple, Fastly must demonstrate a durable platform, not a delivery utility with a security attach.

Embedded-expectations read (INTERPRETATION). At ~$3.0B EV and ~4.1x forward sales, the market is underwriting durable low-teens revenue growth with gradual margin expansion — neither a Cloudflare-style reacceleration nor a repeat of the 2024 stall. Specifically, the price appears to assume: (a) network-services deflation stays “managed” (volume growth continues to out-run mid-single-digit price erosion); (b) security + compute scale into a genuine second leg (the $200M+ run-rate target lands and keeps compounding ~25–40%); © NRR holds above ~110%; and (d) non-GAAP margins march toward the low-teens. That is a plausible base case — but it is the base case now priced in, leaving little margin of safety and meaningful downside if any leg slips.

Scenario framework (INTERPRETATION; illustrative EV zones, not price targets):

The distribution is genuinely two-sided, and the base case brackets the current price — which is the point: at $18 you are paid roughly fairly for the base case, over-paid if the bear plays out, and under-paid only if the bull (unproven edge-AI monetization) materializes. No price target; embedded-expectations framing only.

Verdict (§10): fully-valued on the base case. The stock has already re-rated to price a successful turnaround; the asymmetry that existed at $5–8 (cheapest-ever, left-for-dead) is gone. What remains is a fairly-priced call on execution, with an SBC-obscured cash-flow profile.

11. Variant Perception

Consensus belief. After a violent 2026 rally, consensus has swung from “left-for-dead subscale CDN” to “credible turnaround / edge-AI beneficiary.” The bulls point to accelerating revenue (+20%), security (+47%), NRR recovery (100%→113%), RPO (+63%), positive FCF, and the AI/edge optionality; they treat the FY25–26 inflection as the start of a durable compounding phase and are willing to pay ~4–5x sales for it.

Strongest bull case. The mix-shift is self-reinforcing: as security and compute grow from ~23% toward ~35–40% of revenue, blended growth re-accelerates, switching costs deepen, gross margin structurally rises, and Fastly earns a platform multiple rather than a delivery-utility multiple. Edge inference for AI is a large, unpriced call option that plays to Fastly’s distributed architecture. Net retention above 110% on a low-churn base compounds attractively, and a near-net-neutral balance sheet plus positive cash generation removes the existential risk that capped the multiple in 2022–24.

Strongest bear case. Fastly is a subscale, moat-light, whale-concentrated commodity CDN whose “profitability” and “free cash flow” are SBC illusions (net of $117M SBC, it still burns ~$50M and dilutes ~5–6%/year), now trading at a profitable-incumbent multiple after a momentum-driven double. The core 77% grows ~11% into structural deflation; the growth engine (security/compute) is real but small and contested by a rival 3.5x its size with a genuine developer flywheel. The FY26 capex ramp into memory inflation is already compressing the very free cash flow the re-rating celebrated. Three CEOs in two years, no insider buying, and a ByteDance-concentration overhang round out the risk.

The 3–5 assumptions that matter most: (1) durability of network-services growth against price deflation; (2) whether security/compute scales into a real second leg (>$250M run-rate, sustained 25%+); (3) NRR holding >110% through a full cycle; (4) whether SBC declines as a share of revenue (i.e., does reported FCF ever become real FCF); (5) whether edge-AI/inference monetizes on any material timeline.

Falsification tests. Bull is falsified if network-services growth rolls back toward low-single-digits, NRR slips below 105%, or security decelerates below ~25% — any of which exposes the core as still-melting under a full multiple. Bear is falsified if security + compute clears a $250M run-rate with NRR >112% and SBC falls toward ~10% of revenue (real owner FCF turns clearly positive) — proving a durable platform rather than a dressed-up delivery utility.

Factor-positioning input (FACT / INTERPRETATION). A quantitative factor/risk model characterizes FSLY as a high-volatility cloud-beta vehicle, not a value stock: dominant Cloud-Computing factor beta 2.58, market beta 1.70, LowVol loading ≈ −1.0 (high-vol), Value loading ≈ 0, SmallSize +0.56, alpha −0.27. The long-run track record is catastrophic (5-year return −21%/year, max drawdown −91.8%, Sharpe −0.27), overlaid with a violent one-year rip (rs_12m 159.7; 1-year return +162% annualized) that has already reversed roughly half (3-month return −91% annualized). The tape’s message: this is a re-rated, no-longer-cheap, high-beta momentum turnaround whose bullish consensus is chasing an FCF-inflection/edge-AI narrative — evidence that consensus may be offsides on price, not on the operational direction. Treat as positioning input, not a call.

12. Fact vs. Interpretation Table

13. Open Questions

1. How much of security’s +47% and compute’s +67% is organic breadth vs. a few large deals? Management called Q1-26 security “broad-based” but Q3-25 was one-deal-driven — durability is unproven.
2. Will ByteDance’s US-JV restructuring retain or reroute the ~10% of revenue it represents? A material single-name swing factor.
3. Does SBC decline as a percentage of revenue? The entire “real FCF” question hinges on this; management has not committed to a trajectory.
4. Can NRR hold >110% through a full delivery-deflation cycle, or is the recovery partly cyclical (a rebound off the 2024 optimization trough that fades)?
5. What is the actual AI-traffic contribution? Management explicitly could not quantify it — a large part of the re-rating rests on a number they won’t (or can’t) provide.
6. Is the record Q1-26 gross margin sustainable ex the one-time server-life change and against the FY26 capex/capacity ramp?
7. What does the September-2026 Investor Day commit to on long-term margin, FCF, and SBC targets — and will the market hold management to them?

14. What Must Be True

For the bull case to work (and its single falsification test):

• Security + compute must scale into a genuine second growth leg — clearing a >$250M run-rate and sustaining 25%+ growth — while network-services volume continues to out-run price deflation, so blended growth holds in the mid-to-high teens.
• Net retention must hold above ~110% through a full cycle, proving the recovery is structural (mix-shift-driven) rather than a cyclical rebound.
• SBC must fall toward ~10% of revenue so that reported free cash flow becomes real owner free cash flow — converting the “adjusted profitability” narrative into cash economics.
• Falsification: if, over the next 12–18 months, security decelerates below ~25%, NRR slips below 105%, or SBC stays stuck near ~19% of revenue while real FCF remains negative, the bull thesis (a durable platform earning a platform multiple) is broken.

For the bear case to work (and its single falsification test):

• Network-services growth must roll back toward low-single-digits as price deflation reasserts and volume growth normalizes, exposing ~77% of revenue as a melting core.
• A large customer (ByteDance ≈10%) must optimize away or the security cross-sell must stall, re-concentrating the risk and puncturing the growth story — while the ~4.8x-sales multiple (an Akamai multiple on an unprofitable business) de-rates toward 2.5–3x.
• Falsification: if network services sustains double-digit growth, security clears the $250M run-rate on schedule, and NRR holds >112%, the bear thesis (subscale commodity CDN dressed up by SBC accounting) is broken and the multiple is defensible.

15. Source Appendix

See Appendix B — Source Appendix for the full primary-source list. Primary sources: FY2025 Form 10-K (filed 2026-02-25), Q1-2026 10-Q, 2026 DEF 14A, the Q1-2026 earnings call (2026-05-06), the trailing-five-year SEC filing corpus, public market data, and Cloudflare/Akamai public results for peer comps. Management commentary is treated as hypothesis throughout and validated against filings, financials, and external evidence.

The analytical body of this note carries no buy/sell recommendation and no price target; the sole position-taking view is the clearly-labeled Author's Take block at the top, which is the author’s own opinion and general information only, not investment advice.

APPENDIX A — Standard Diligence Questionnaire

Supplemental to the research note. Fact / Interpretation / Assumption labeled where it matters. Report date 2026-07-04.

General

What thoughtful questions have other investors asked about this company? The recurring institutional debates are: (1) Is the FY25–26 revenue re-acceleration durable or a cyclical rebound off the 2024 large-customer-optimization trough? (2) Is Fastly’s “profitability” and “free cash flow” real, or an artifact of adding back ~19%-of-revenue stock comp? (3) Can a subscale #3–#4 CDN build a durable security/compute platform from behind against Cloudflare’s developer flywheel? (4) How exposed is the model to ByteDance (~10% of revenue) and top-10 concentration (34%)? (5) Is edge-AI/inference a real revenue driver or a narrative? (6) After a triple-off-the-lows, is the stock still cheap? (Answer: no — mid-range on own history.)

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Neither “earnings” (still GAAP-lossmaking) nor a clean cycle. Revenue growth is recovering off a FY24 cyclical trough (+7.4%) to +15% (FY25) / +20% (Q1-26) — closer to a cyclical low-to-mid on the growth curve, driven partly by internal actions (security cross-sell, GTM discipline) and partly by a rebound in whale usage. Internal vs external: a mix — the security/compute mix-shift is internal; delivery volume and price-deflation are external. How stable are revenues? Below-average stability for software: consumption/usage-based, so quarterly revenue “fluctuates frequently” (management’s words); minimum commitments (RPO $354M) provide a partial floor. Gross retention is high (98.7%), but net revenue swings materially with a few large customers’ usage. Outlook / market size: Global CDN + edge-security + edge-compute is a large, growing TAM, but the delivery portion is deflationary and commoditizing; the growth/profit is migrating to security and compute (international and domestic).

Business Quality & Competitive Moat

More or less competitive industry? More — hyperscaler bundling, Cloudflare’s free tier, and low-cost entrants intensify delivery competition; security is also crowded. Partial relief from supply exit (Edgio bankruptcy, StackPath). How profitable is the business (ROIC/ROE)? Poor — GAAP operating margin −19% FY25; ROE ~−13%; ROIC negative; −$1.114B accumulated deficit. Improving (from −62% operating margin in FY21) but not yet earning its cost of capital. (Fact.) How profitable is the industry / barriers to entry? Delivery: low barriers, thin/negative economics. Security & compute: higher barriers and better economics, but they accrue to the scale leader. Many competitors (Cloudflare, Akamai, hyperscalers, Imperva, F5, Bunny). Can the business be easily understood? Yes — a distributed network selling delivery, security, and compute on usage-based pricing. Undermined by foreign low-cost labor? Not labor — but undermined by low-cost capital/infrastructure competition (hyperscalers, Bunny.net at ~$0.005/GB). Do brands matter? Modestly — developer reputation and reliability matter (Forrester Wave Leader; but the 2021 outage cut the other way). Not a consumer brand. Nature of competition / switching costs? Delivery: price/performance, with low switching costs (multi-CDN sourcing lets customers reroute instantly). Security/compute: moderate switching costs (customers encode WAF rules and edge code). (Interpretation — this asymmetry is the whole moat story.)

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The developer/security platform and customer relationships are partly capitalized as $670M goodwill (Signal Sciences); no large hidden asset. Off-balance-sheet liabilities? Operating/finance leases for network capacity ($68.3M finance-lease balance on-sheet; operating-lease commitments in the footnotes). Convertible-note dilution is the key contingent claim (new strikes $15.26/$19.74). How conservative is the accounting? Mixed/aggressive on presentation: heavy reliance on non-GAAP metrics that add back ~$117M SBC; a FY25 server-useful-life change adding ~190bps to gross margin; and the Q1-26 discontinuation of total-customer-count disclosure. Revenue recognition itself is standard. (Interpretation — treat the “adjusted” figures skeptically.) How capex-hungry? Historically light (~5% of revenue FY25) but rising — FY26 infrastructure capex guided to 10–12% of revenue as memory prices rose 2–3x and capacity ramps; management argues its software-defined network is more capital-efficient than legacy peers.

Capital Allocation & Management

How much FCF, and how is it used? Reported FCF +$65.8M FY25, but −$51.5M net of SBC — real owner cash is still negative. Used to fund growth and retire debt; no buyback of consequence. Significant acquisitions recently? No recent large deal; the defining deal was Signal Sciences (2020, ~$775M, mostly stock) — the security franchise, now the growth engine. Glitch (2022, small). Buying back shares? No — a token $10.7M in 2022 against ~$117M/year SBC. Net effect is chronic dilution (~+54% shares since 2020). Issuing large amounts of stock to insiders? Yes — SBC is 18.8% of revenue; the model is equity-heavy. Insider Form 4 activity is routine sell-to-cover / 10b5-1; zero open-market purchases. Compensation policy / director & management motivation? CEO ~$5.69M, 22.6:1 ratio; PSUs use relative-TSR (vs Russell 2000) and, for 2026, FCF/CMRR gates — improving alignment, but no ROIC gate and a 135.4%-of-target 2025 STI payout in a GAAP-loss year. Founder Bergman pledged ~2.27M shares. Third CEO in ~two years (execution-continuity risk). (Interpretation — alignment improving but imperfect.)

Valuation & Market Data

ADR / MLP / K-1? No — US Class A common stock (NASDAQ: FSLY); no K-1. Dividend policy? None (no dividend; reinvesting/deleveraging). How profitable is the business? Not yet, on GAAP — see above. Net income vs cash from operations diverging? Yes, and instructively: GAAP net loss −$121.7M vs CFO +$94.4M FY25 — a ~$216M gap bridged mainly by $117.3M SBC and $77.9M D&A. This is the classic “positive cash flow, negative earnings” software profile — legitimate as far as it goes, but the SBC portion is a real economic cost to shareholders. (Fact + Interpretation.)

Risks & Downside

What would cause the stock to decline? A network-services growth rollback into deflation; a ByteDance/large-customer optimization; a security-cross-sell stall; SBC staying ~19% of revenue (real FCF stays negative); a high-beta/momentum de-rate (beta 1.70, already −47% off the 52-week high); disappointment on the unquantified AI narrative; multiple compression from ~4.8x toward 2.5–3x sales. Risk of catastrophic loss? Low near-term — near net-neutral balance sheet, $330M cash, no maturity crunch before 2028. Bankruptcy risk remote. Chance of total loss? Very low over the forecast horizon. The realistic downside is a valuation de-rate plus ongoing dilution, not a zero.

Recent News & Events

Has the business environment changed recently? Yes, favorably on execution: revenue re-accelerated to +20% (Q1-26), security to +47%, NRR to 113%, RPO +63%; FY26 guidance raised. Offsetting: memory-cost inflation (capex ramp), a ByteDance US-JV restructuring (Jan-2026), and a fully-valued stock after a triple-and-halve round trip. Significant acquisitions / accounting-policy changes? No acquisitions; yes on accounting — server-useful-life change (GM benefit), infrastructure-only capex reporting, and discontinued total-customer-count disclosure. Recent changes — markets, facilities, management? New APAC office (Singapore); near-complete C-suite refresh (CEO Compton Jun-2025, CFO Wong Aug-2025, CMO Jenkins Q1-26, President-GTM Lovett); NYSE→Nasdaq transfer (Dec-2025); Investor Day scheduled 23-Sep-2026.

APPENDIX B — Source Appendix

Compiled 2026-07-04. Primary sources first. Every non-obvious fact in this note traces to an entry here.

Primary — SEC filings (EDGAR, CIK 0001517413)

1. Form 10-K, FY2025 (filed 2026-02-25) — revenue by product line & geography; customer/NRR/RPO KPIs; SBC; convertible-note footnote; goodwill; risk factors; ByteDance-affiliated 10% concentration disclosure. https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001517413&type=10-K
2. Form 10-Q, Q1-2026 — quarterly product-line revenue, KPIs, cash flow, debt schedule.
3. Prior 10-Ks (FY2021–FY2024) — multi-year revenue, margin, SBC, and restructuring history.
4. Form 8-K earnings releases — Q3-25 (2025-11-05), Q4/FY25 (2026-02-11), Q1-26 (2026-05-06); guidance, non-GAAP reconciliations. Q4/FY25 press release: https://www.businesswire.com/news/home/20260211703867/en/Fastly-Announces-Both-Record-Fourth-Quarter-and-Full-Year-2025-Financial-Results
5. DEF 14A proxy, 2026 (filed 2026-04-22) — executive compensation, PSU/rTSR metrics, CEO transition (Compton/Nightingale), founder share pledge.
6. Form 4 / Form 144 corpus (5 years; ~395 Form 4, ~191 Form 144, via EDGAR) — insider transaction pattern (routine sell-to-cover / 10b5-1; zero code-P purchases).
7. 8-K, CEO transition (June 2025) — Compton promoted from CPO; Nightingale resignation.
8. 8-A12B / Form 25 — NYSE → Nasdaq listing transfer (Dec-2025).

Primary — Earnings call transcript

9. Q1-2026 earnings call (2026-05-06). CEO Kip Compton, CFO Rich Wong. Source of: $173M/+20% revenue, security $38.8M/+47%, compute $8M/+67%, NRR 113%, RPO $369M/+63%, 634 large customers, non-GAAP GM 65.1% (190bps one-time), non-GAAP op income $19.1M, adj EBITDA $29.5M, FY26 guide ($710–725M / op profit $58–68M / EPS $0.27–0.33 / FCF $40–50M), capex 10–12% of revenue, memory prices +2–3x, 7.75% 2028 convert, edge-AI/inference commentary, ContentGuard launch, Forrester Wave Leader.

Market and quantitative data (reconciled to filings)

10. Company financial statements (FY2019–FY2025) — income statement, balance sheet, cash flow, and derived ratios (ROE/ROIC/margins), enterprise value, and valuation multiples, reconciled to the 10-K.
11. Public market price history — 5-year daily prices; own-history valuation percentiles (P/S ≈42nd, P/B ≈57th, composite ≈49th).
12. Factor/risk model data — factor loadings (Cloud-Computing beta 2.58, market beta 1.70, LowVol ≈ −1.0, Value ≈ 0, SmallSize +0.56), risk-adjusted track record (5yr −21%/yr, max DD −91.8%, Sharpe −0.27; 1yr +162% annualized; 3mo −91% annualized), beta/alpha and relative strength.

Peer / industry sources

13. Cloudflare (NET) FY2025 results and investor materials — revenue ~$2,168M (+30%), ~75% gross margin, near GAAP breakeven, ~$81B market cap. https://www.cloudflare.com/press/
14. Akamai (AKAM) FY2025 results — revenue ~$4B, ~+5% growth, ~15% GAAP operating margin, profitable, ~16.5x EV/EBITDA.
15. CDN / edge market structure — trade coverage of Edgio (2024 bankruptcy), StackPath wind-down, hyperscaler CDN (CloudFront/Cloud CDN/Front Door), Bunny.net low-cost pricing; CDN market-share data (accessed 2026-07-04).
16. Forrester Wave: Edge Development Platforms, 2026 — Fastly named one of two Leaders (per company disclosure / Q1-26 call).

Framework references

17. Greenwald & Kahn, Competition Demystified — barriers-to-entry / three-advantage-types moat taxonomy (applied §3–§4).
18. Edward Chancellor (ed.), Capital Returns (Marathon) — supply-side capital-cycle lens (applied §3).

Management commentary is treated throughout as hypothesis, validated against filings, financials, and external evidence. This note reflects no position in the securities discussed.