Riot Platforms, Inc. (NASDAQ: RIOT) — Landlord Price, Miner Risk: Pricing the Unsigned Campus as a Done Deal

Independent equity research. Report date: 2026-06-14.

⚡ Claude’s Take

This block is the author’s own independent opinion. It is general information only and not investment advice. Everything below it (the analysis) is deliberately position-free and carries no price target.

Verdict: AVOID at $26.61 / accumulate-on-deep-weakness only / NOT a short. Fair-value zone ~$10–15 per share on a sum-of-the-parts that gives full credit for the BTC treasury, the one signed AMD lease, and the mining/power base — but refuses to pay, in advance, for the unsigned Corsicana campus that is ~$6–8B of the current ~$10.7B enterprise value. Conviction: medium.

Riot is a well-run operator of a fundamentally moat-less business. Its Bitcoin-mining engine (89% of FY2025 revenue) is a textbook commodity — undifferentiated output sold at a globally-set price, no barriers to entry, a self-defeating capital treadmill where the company mined more coins in FY2025 (5,686 vs. 4,828) yet watched its cost-to-mine jump +54% to ~$49,600/BTC excluding depreciation (~$91,400 including it) after the April-2024 halving. The genuine assets are not the mining — they are ~1.7 GW of scarce, energized, now fee-simple ERCOT power and land (Corsicana ~1 GW, Rockdale ~700 MW), an in-house switchgear vertical (ESS Metron) that is a real construction-speed edge, and a signed, investment-grade AMD colocation lease at Rockdale (50 MW, ~$636M contracted revenue, ~$51M stabilized NOI) that proves the AI-pivot model can work. Activist Starboard is pushing the value-unlock. This is not a fraud or a zombie; it is a real attempt to escape a bad industry into a good one.

But the market has already paid the landlord multiple while the stock still carries the miner’s risk profile — and the landlord’s cash flows are mostly unsigned. At $26.61 (≈$10.1B market cap, ≈$10.7B EV, ~16x EV/sales, 3.9x P/B at the 83rd percentile of its own ~10-year history, +152% over twelve months), strip out the real parts — BTC treasury (~$1.0B and shrinking, because Riot is now a net seller funding its build), the AMD lease (~$0.5B net), and the mining power/fleet (~$1–3B) — and roughly $6–8B of EV is left over, paid for an HPC pipeline that has signed exactly 50 MW. Capitalized at the cohort’s own $/contracted-MW, that residual implies the market is paying Riot for ~200–550 MW of signed-equivalent hyperscale capacity it has not closed — essentially the entire 756 MW Corsicana campus priced as a done deal after sixteen months of marketing with zero signed tenant. My base-case SOTP lands near $12 — about half the quote. The framing, grounded in the tape, is a crowded ~2.9-beta crypto-momentum trade at its richest-ever valuation: it still clusters with crypto-miner ETFs (WGMI/BKCH/MARA), not data-center REITs; it carries positive Momentum on negative Quality (the classic mean-reversion setup); and its empirical lifetime max drawdown is −99.98% — this equity has gone to near-zero before and the beta says it can again. I would not short it — 50 MW is real, Corsicana could sign, Starboard is live, and a Bitcoin melt-up plus a crowded high-beta tape make the squeeze risk acute — but I would not pay up here. Flips bullish: a named investment-grade hyperscaler signing 100+ MW at Corsicana with a closed non-recourse project-finance facility on AMD (optionality becomes contracted, financeable NOI). Flips bearish: two or three more quarters with no Corsicana lease and BTC breaking below ~$50k — which simultaneously craters the treasury, the funding model, and the 2.9-beta equity. Tag: “They built a beautiful substation — and sold you the whole power campus before signing a tenant.”

1. Executive Summary

Riot Platforms, Inc. (NASDAQ: RIOT) is a Castle Rock, Colorado–headquartered, vertically-integrated industrial-scale Bitcoin self-miner in the early stages of a deliberate pivot toward leasing its secured power and land as AI/HPC data-center capacity. FY2025 revenue was $647.4M (+72% YoY), of which Bitcoin Mining was $576.3M (89%), Engineering (ESS Metron/E4A switchgear and power-distribution) $64.7M (10%), and residual legacy hosting under 1%. The company operates ~1,292 MW of developed power across Rockdale and Corsicana (Texas) and Kentucky, ran a deployed hash rate of 42.5 EH/s in Q1 2026, and holds a Bitcoin treasury of ~15,680 BTC.

The investment debate is unusually clean and unusually prospective. What is real: scarce, energized, fee-simple ~1.7 GW of ERCOT power and interconnect in a power-constrained AI-capex supercycle; an in-house engineering vertical that demonstrably speeds and de-risks data-center fit-out; and one signed, investment-grade AI colocation lease — AMD at Rockdale, 50 MW, ~$636M of contracted revenue over a 10-year primary term, ~$51M average annual NOI. What is not yet real: the prize the equity is priced for — a hyperscaler lease at the 756 MW Corsicana campus, which after roughly sixteen months of marketing (advisor Altman Solon engaged since January 2025) has zero signed tenants as of the Q1 2026 call.

The financial reality beneath the narrative is poor. GAAP operating income has been negative every year from FY2022 to FY2025 (FY2025 operating margin −53%). Reported net income is dominated by non-cash Bitcoin fair-value marks under ASU 2023-08 (a +$457.4M gain in FY2024 manufactured a +$109.4M net profit; a −$115.9M FY2025 mark and −$326.7M Q1’26 mark drove large net losses) — GAAP EPS is noise. The business is funded not by operating cash (negative free cash flow every year) but by serial equity dilution (share count ~4x in 4.5 years, ~$3B raised via ATM, ~$376M of stock-based compensation) and, increasingly, by selling down the Bitcoin treasury (5,363 BTC sold in FY2025, 3,778 in Q1’26).

Competitively, Bitcoin mining confers no durable advantage — no barriers to entry, no customer captivity, output undifferentiated, share structurally eroded by rising network hashrate. The only genuine edges are operational: ultra-cheap power (~3.7 c/kWh all-in) and the ESS Metron vertical. Both are real but modest and replicable. The land-and-power optionality is valuable but unmonetized.

On valuation, Riot is expensive on everything contracted and cheap only on optionality it has not closed. At ~$10.7B EV the market embeds ~$6–8B of unsigned HPC pipeline — roughly 200–550 MW of signed-equivalent capacity against 50 MW actually signed — while peers (HUT, CIFR, WULF, IREN, APLD) carry 500–907 MW of contracted/operating capacity with hyperscaler or neocloud anchors. On the metric that drives the entire re-rating (contracted HPC MW), Riot is the cohort laggard yet trades at a cohort-comparable EV. The stock is best understood as a fat-tailed real option: a small “hard” floor (~$2.5–4.5B of BTC + AMD + mining), enormous unsigned upside, and a bear tail that is a near-total loss given the 2.9 beta and the −99.98% empirical drawdown history.

This report takes no position and sets no price target. It frames the embedded expectations, the scenarios, and the falsification tests on both sides.

2. Business Overview

Riot Platforms (formerly Riot Blockchain, renamed January 2023) is a Nevada-incorporated, Colorado-headquartered company that earns money three ways, in sharply descending order of materiality:

(1) Bitcoin Mining — $576.3M, 89% of FY2025 revenue. Riot operates company-owned ASIC fleets that compete to validate Bitcoin blocks in exchange for the block subsidy (3.125 BTC post the April-2024 halving) plus transaction fees. This is self-mining only — a crucial nuance, because the legacy “Data Center Hosting” segment (hosting other miners’ machines for a fee) was wound down: Riot stopped accepting new hosting in 2023 and terminated all legacy hosting customers during FY2024. The flagship is the Rockdale, Texas facility (~700 MW, now 100%-owned and consolidated after the 2025 Rhodium asset acquisition); secondary sites are Corsicana (Texas, in development) and two Kentucky facilities acquired via Block Mining (December 2024). Mining revenue is recognized at the fair value of Bitcoin earned on the day of award.

(2) Engineering (ESS Metron + E4A Solutions) — $64.7M, 10% of revenue. Acquired in 2021 (Whinstone/ESS Metron) and December 2024 (E4A), this unit designs and manufactures low- and medium-voltage electrical switchgear and power-distribution equipment — among the most supply-constrained components in data-center construction — for both Riot’s own buildout and external commercial, governmental, utility, power-generation, and data-center customers. Engineering backlog (remaining performance obligations) was $224.6M at FY2025-end (+302% YoY), ~90% data-center-driven, dipping to $193.4M in Q1’26 because Riot is reserving capacity for its own build. This vertical integration is the company’s single most defensible operating edge.

(3) Data-Center (AI/HPC colocation leasing) — nascent, first revenue Q1 2026. The new third leg: leasing built-to-suit data-center capacity to AI/high-performance-computing tenants under long-term contracts. The only signed lease is AMD at Rockdale (50 MW after an April 2026 expansion). This is the segment the equity is priced on, yet it contributed only a trickle of operating-lease revenue in the most recent quarter.

A fourth, non-operating but financially central feature is the Bitcoin treasury: 18,005 BTC at 12/31/2025 (carrying $1.575B), declining to ~15,680 BTC by Q1’26 (~$1.0–1.1B at ~$64k spot) as Riot turned from holder to net seller to fund its data-center capital expenditure.

Revenue quality is mixed-to-poor. Mining revenue is recurring in form but is a price-taking commodity stream with no contractual durability and a deteriorating unit economic. Engineering is genuine industrial revenue with a backlog. The data-center leasing — once scaled and contracted — would be the highest-quality stream (long-dated, triple-net, financeable), but today it is a rounding error in the P&L and a giant call option in the multiple.

Verdict: A commodity-mining business (the overwhelming majority of revenue) bolted to a genuinely useful engineering vertical, with an embryonic but strategically central data-center leasing arm. The reported business and the priced business are different companies: investors are buying the unsigned future, not the mining present.

3. Industry Dynamics

Riot sits at the intersection of two very different industries, and the entire bull thesis is a bet that it can migrate from the bad one to the good one.

Bitcoin mining — structurally bad. Mining is a textbook commodity industry with a self-defeating capital cycle. The output (Bitcoin) is perfectly fungible and priced globally; a miner has zero pricing power. Every input — MicroBT/Bitmain ASICs, ERCOT/MISO power, land, capital — is available to all bidders, and costs converge on the same efficiency frontier (sub-3.5 c/kWh power, the latest-generation sub-15 J/TH rigs). The April-2024 halving cut the block subsidy to 3.125 BTC, structurally doubling unit production cost overnight. Global network hashrate sits near a record ~1 ZH/s and difficulty at an all-time high; “hashprice” (revenue per unit of hash) is structurally pressured at roughly $37–38/PH/day. This is Marathon Asset Management’s capital cycle in its purest form: high Bitcoin prices draw in capital, the network’s provisioned hash roughly doubled across 2024–25, difficulty rose to an all-time high, and incumbent margins compressed — Riot’s own cost-to-mine excluding depreciation rose to $49,645/BTC in FY2025 (49% of the $101,350 production value), up +54% from $32,216 in FY2024, and ~$91,400 including depreciation (~90% of production value). A miner must reinvest perpetually just to hold share, because share = own hashrate ÷ global hashrate, and the denominator only grows. The 10-K calls it an “arms race.” In Greenwald’s framework, an industry with no barriers to entry is one where strategy is irrelevant and only operational efficiency matters — a bad-industry base case.

AI/HPC data-center colocation — structurally good. The pivot thesis rests on a real dislocation: AI compute is bottlenecked not on chips but on power and interconnect, which carry multi-year lead times. Miners that secured large blocks of energized grid capacity — Riot’s ~1.7 GW in ERCOT — hold a genuinely scarce asset. Converting that capacity into long-dated, triple-net, take-or-pay leases to creditworthy AI/hyperscale tenants transforms a commodity cash flow into a contracted, recurring, financeable one — the basis on which the “converted miner” cohort (HUT, CIFR, WULF, IREN, APLD) re-rated from ~book value to data-center-infrastructure multiples. This is a structurally better business: barriers to entry (power scarcity, interconnect queues), customer captivity (multi-year leases with switching costs), and recurring revenue.

The catch is that the AI/HPC re-rating is itself a Marathon capital cycle in formation — the entire converted-miner cohort plus the hyperscalers plus the neoclouds are racing to build the same capacity, which over time compresses lease rates and $/MW valuations. Riot is a late, sub-scale entrant to this better industry: secured power but, so far, only 50 contracted MW.

The ERCOT power-market angle deserves emphasis, because it is simultaneously Riot’s best asset and a source of the confusion in how the stock is valued. Riot’s Texas sites sit in ERCOT, a deregulated, energy-only market with no capacity payments and extreme price volatility — which a large, flexible, interruptible load can monetize. Riot procures power under fixed-price PPAs and then voluntarily curtails during scarcity-price events, selling its hedged power back into the grid and participating in demand-response (4CP) programs; this generated \~$56.7M of power-curtailment credits in FY2025 (up from $33.7M), worth roughly $10k per Bitcoin mined and pulling net all-in power cost toward \~3.0 c/kWh after credits. This flexibility is genuinely valuable and is part of why ERCOT miners can claim sub-3.5 c/kWh power. But two cautions: (i) the same flexibility that makes a Bitcoin load grid-friendly is exactly what a 24/7 hyperscale AI tenant does not want — AI training/inference demands firm, uninterruptible power, so converting curtailable mining megawatts into HPC leases sacrifices the demand-response revenue and requires firmer (more expensive) supply arrangements; and (ii) the curtailment-credit stream is policy-dependent and could compress if ERCOT rules or large-load interconnection standards change. The power edge is real, but it does not transfer one-for-one into the AI-landlord economics the multiple assumes.

Verdict: Bitcoin mining is a structurally bad industry — commodity output, no entry barriers, a perpetual-reinvestment treadmill. AI/HPC colocation is structurally better but increasingly crowded. Riot is attempting the right migration but is not yet there: it remains, by revenue and by risk profile, a Bitcoin miner with an AI option.

4. Competitive Position

The central competitive question is binary: does Riot have a durable moat? In mining, no. In AI/HPC, not yet — and it is behind.

Pressure-testing the claimed advantages (Greenwald moat test):

• Low-cost power (~3.0–3.7 c/kWh, ERCOT fixed-price PPAs, curtailment credits). Real and best-in-class — Rockdale 3.7 c, Corsicana 3.5 c, Kentucky 4.8 c, with $56.7M of demand-response/curtailment credits in FY2025 (~$10k per BTC mined). But this is a shared, replicable advantage of any well-capitalized operator that secures a site and a PPA in ERCOT — not a proprietary, privileged resource. It is table stakes that Riot executes well, not a barrier.
• Scale (42.5 EH/s deployed, one of the largest North American miners). Mining scale does not create Greenwald’s economies-of-scale barrier, which requires customer captivity — and mining has none (output sold to an anonymous global market). A bigger miner is just more rigs on the same three-year depreciation curve. Size ≠ scale advantage here.
• Vertical integration (ESS Metron/E4A switchgear). The closest thing to a real edge. In-house manufacturing of the most supply-constrained data-center components (switchgear, PDUs) plus electrical engineering speeds buildout, lowers cost (~$23–24M cumulative capex savings since 2021), and directly de-risks AI/HPC fit-out — it is cited as the enabler of Riot’s unusually fast AMD delivery. But it is a modest cost/capability advantage, not a wide moat: rivals can hire EPC firms and order switchgear (with longer lead times).
• Land + power optionality (Corsicana ~1 GW, ~900 acres now owned; Rockdale ~700 MW). The most valuable asset and the entire re-rating case — secured, energized, large-scale grid capacity is genuinely scarce. But optionality is not a moat until it is monetized with contracts.

The decisive comparison — Riot is the cohort laggard on signed HPC MW. Riot has contracted 50 MW (AMD only, at Rockdale), with an A-rated chip vendor (not a hyperscaler). Corsicana’s 600+ MW has been under “formal evaluation” since January 2025 with zero signed tenant or LOI sixteen months later. Against the AI-pivot peers (a cohort that has broadly re-rated to rich valuations at current prices):

What the AMD signing does and does not prove. It is tempting to read the AMD lease as evidence of a moat. It is not — but it is a meaningful credibility signal. The fact that AMD, an A-rated semiconductor company that designs the very GPUs that fill AI data centers, chose to put compute in Riot’s Rockdale facility validates that Riot’s sites and engineering meet high-density AI specifications, and that the ESS Metron switchgear vertical can deliver power infrastructure on a timeline tenants care about (first 5 MW rent-commenced within months of signing). That de-risks the execution question — can Riot physically build and operate AI-grade colocation? — which several skeptics had raised. But credibility is not captivity. AMD is one tenant at 50 MW; the lease is a normal commercial colocation contract, not a structural lock on a scarce resource that competitors cannot replicate. CIFR (AWS), WULF (FluidStack/Google), HUT (an AA hyperscaler), and APLD (CoreWeave) have all signed larger anchors at better-known credits, which tells you the “we can sign AI tenants” capability is shared across the cohort, not proprietary to Riot. The signing moves Riot from “unproven developer” to “proven small developer” — worth something, but a long way from the contracted, financeable, multi-hundred-MW landlord book that the cohort multiple presumes.

Both Greenwald advantage tests fail in mining: share erodes structurally with network hashrate, and the financials show no franchise (FY2025 operating margin −53%, net loss −$663M, deeply negative ROIC, breakeven EBITDA). A real moat must tie to a financial outcome that would deteriorate without it; Riot’s “moat” claims do not.

Verdict: No durable moat in mining (commodity, no barriers, no captivity, a share-eroding treadmill). The single defensible edge is a modest cost/capability advantage — cheap ERCOT power plus the ESS Metron engineering vertical — alongside valuable but unmonetized land-and-power optionality. Riot is a well-run operator of a moat-less commodity business that happens to hold scarce power assets, and it is trailing its AI-pivot peers on the one metric (contracted HPC MW) that would create a genuine landlord moat. The bull case rests entirely on optionality conversion, not on any existing franchise.

5. Growth History and Forward Opportunities

Historical growth. Revenue compounded from $213.2M (FY2021) → $259.2M (FY2022) → $280.7M (FY2023) → $376.7M (FY2024) → $647.4M (FY2025). The FY2025 +72% jump, however, is largely inorganic and price-driven, not a sign of improving economics: it reflects the Rockdale/Rhodium consolidation (bringing 125 MW and all ASICs in-house), the Block Mining/Kentucky consolidation, the growing engineering backlog, and a higher average Bitcoin price — not better mining margins. Indeed unit economics deteriorated over the same window: cost-to-mine rose +54% post-halving even as BTC mined grew +18% (5,686 vs. 4,828). This is the capital-cycle treadmill made visible — more output, worse economics.

Hash-rate growth has been steady (31.5 → 38.5 EH/s across FY2025; 42.5 EH/s deployed by Q1’26), funded by continuous capex and ATM issuance. But hashrate growth that merely tracks the network is value-neutral: it preserves share without improving returns.

Forward opportunities. The growth case has decisively shifted from “mine more Bitcoin” to “convert power into AI/HPC leases.” The credible, contracted growth is the AMD ramp: management guides exiting 2026 at ~$37.8M annualized operating-lease revenue run-rate, scaling to ~$55.6M exiting 2027 once the full 50 MW is online. That is real but small relative to a $10B market cap.

The uncontracted opportunity is the entire bull thesis:

• Corsicana — 756 MW of planned critical-IT capacity on 1 GW of approved power; core-and-shell construction initiated; management targets ~160 MW of completed shell by Q2 2027 and “additional announcements in 2026.” Zero leases signed.
• AMD option ladder — AMD holds options for +50 MW more (of the original tranche) plus a new 100 MW option (conditional on using the 50 MW and on power availability), a potential pathway to 200 MW at Rockdale.
• Engineering scale-up — plans to grow ESS Metron capacity ~25% in 2026 to serve both internal and external data-center demand.

The quality of this growth is the crux. The contracted portion (AMD) is high-quality but immaterial to the valuation. The uncontracted portion (Corsicana + options) would be high-quality if signed — but it is, today, a slide-deck pipeline, and management itself hedges (“I can’t tell you when our next lease will be signed… peers have seen deals start and stop or fall through”).

Verdict: Low-quality historical growth (inorganic, price-driven, on deteriorating unit economics) pivoting toward a potentially high-quality but almost-entirely-unsigned forward opportunity. The growth that matters for the thesis has not happened yet.

6. Financial Quality

This is the section where the narrative and the numbers diverge most sharply.

Operating income has been negative every year FY2022–FY2025: −$82.6M, −$255.1M, −$331.4M, −$343.3M (FY2025 operating margin −53%). The only profitable operating year in the five-year set was FY2021 (+$23.9M, at the peak of the prior Bitcoin bull). EBITDA is a rounding error: ~+$3.5M FY2025, −$119.3M FY2024, −$22.3M on a TTM basis through Q1’26. There is no through-cycle cash earnings power.

GAAP net income is a Bitcoin price artifact. Under ASU 2023-08, Riot marks its Bitcoin treasury to market, and — unusually — books the “change in fair value of bitcoin” inside operating income: −$184.7M (FY2023), +$457.4M (FY2024), −$115.9M (FY2025). That single line explains the headline swings. FY2024’s +$109.4M net income was manufactured by the +$457.4M mark; absent it, FY2024 was a large operating loss. FY2025’s −$663.2M net loss = −$343.3M operating loss − $115.9M mark − ~$158.1M one-time Rhodium settlement loss − $29.7M impairment − $5.8M convertible-investment loss. Q1’26’s ~−$500M net loss was driven by a −$326.7M mark as Bitcoin fell. GAAP EPS is noise; read adjusted EBITDA and segment economics instead.

The true cost of production is the tell. Cost-to-mine excluding miner depreciation was $49,645/BTC in FY2025 (49% of production value), and ~$91,400 including depreciation (~90%). Fully loaded (with SG&A and SBC), per-BTC economics are thin-to-negative and depend on Bitcoin holding well above ~$90–100k. The +54% YoY cost jump is the halving plus a +47% rise in network hashrate. The one genuine financial edge — ultra-cheap power at ~3.7 c/kWh all-in, aided by ~$56.7M of curtailment credits — does not offset the halving and hardware depreciation.

The funding model is serial dilution plus treasury liquidation. Free cash flow has been negative every year (FY2025 FCF −$567M; TTM FCFF −$163M). The gap is plugged by:

• Equity: diluted share count rose from ~93.5M (FY2021) to ~379M now — roughly 4.0x in 4.5 years (~305% dilution) — with ~$2.96B of stock sold via ATM over five years (a new $500M ATM launched December 2025), plus ~$376M of cumulative stock-based compensation (~$125.7M in FY2025 alone, ~19% of revenue).
• Bitcoin: sold 5,363 BTC ($535.5M) in FY2025 and 3,778 BTC ($289.5M) in Q1’26 to fund the data-center build — the treasury fell from 19,287 BTC (Q3’25) to ~15,680 BTC (Q1’26).

Liquidity (Q1’26): cash $205.7M; total debt $877.2M, comprising 0.75% Convertible Senior Notes due 2030 (~$594M principal, conversion price ~$14.86 — deeply in-the-money at $26.61, i.e. dilutive), the $200M Coinbase Credit Bitcoin-collateralized term loan (fully drawn; second amended-and-restated agreement signed April 2026, fixed-rate, extended maturity), and ~$35M of leases. Net debt $636.5M; total debt/capital ~27%; current ratio 1.08. Runway is not supported by operating cash — it rests on the BTC treasury (~$1.0B), the $500M ATM, the Coinbase facility, and an as-yet-unclosed non-recourse project-finance facility on the AMD lease.

Segment economics underline the point. Within FY2025’s 37.9% reported gross margin (which excludes miner depreciation, parked in “other operating expenses”), the mining segment’s all-in cash margin is thin and entirely Bitcoin-price-dependent; the engineering segment carries genuine but modest industrial margins on its $224.6M backlog; and the new data-center operating lease — once stabilized — targets an ~80–90% NOI margin, the highest-quality stream but immaterial in size today (~$37.8M annualized run-rate exiting 2026). The mix tells the story: ~89% of revenue comes from the lowest-quality, price-taking source, and the highest-quality source is a rounding error. A useful sanity check is the cost-to-mine “breakeven” — at ~$91,400/BTC fully loaded against a ~$64k spot, current production is cash-negative on a fully-burdened basis and only the cheapest-power, lowest-depreciation marginal coins clear. This is why the company is deliberately characterizing mining as “a means to an end” and redirecting power toward HPC.

A note on the Bitcoin-as-funding flywheel. Management’s stated financing sequence is: Phase 1, fund initial development off the balance sheet and the Bitcoin treasury; Phase 2, raise tenant-backed non-recourse project finance (targeting ~80% loan-to-cost on the AMD lease, citing the investment-grade tenant); Phase 3, recycle recovered equity into the next build. The logic is sound if Phase 2 executes — but no project-finance facility had closed as of the Q1’26 call, so the build is currently funded by the dilutive sources (ATM, SBC) and treasury liquidation. The “no equity issued in Q1’26” boast is accurate but came at the cost of selling 3,778 BTC; it is treasury depletion dressed as discipline. The flywheel only becomes virtuous when non-recourse debt replaces coin sales — an unproven step.

Verdict: Poor financial quality. Economics do not improve with scale — unit cost rose +54% post-halving while the business stayed operating-loss-making across four years. Reported net income is a leveraged bet on Bitcoin spot, not an earnings stream. Negative FCF every year, self-funded by ~4x share dilution and a now-shrinking coin stack. The only genuine financial edge is cheap power. This is a commodity producer with negative through-cycle ROIC; the equity is a levered option on the data-center pivot, funded by selling Bitcoin.

7. Capital Allocation

M&A has been a series of capacity/power/land grabs rather than demonstrably returns-accretive deals: Whinstone/ESS Metron (2021, the Rockdale flagship plus the switchgear vertical — the one acquisition with a clear strategic payoff); Block Mining (December 2024, ~$92.5M up to $375M with earnout, Kentucky, +~1 EH/s); and the Rhodium Encore assets at Rockdale (2025, $185M total — ~$129.9M cash + $49.0M Riot stock — reclaiming 125 MW and all ASICs, ending a loss-making legacy hosting arrangement, and dismissing litigation, at the cost of a one-time $158.1M settlement loss). Goodwill stands at $122.5M and intangibles at $313.0M. The acquisitions were bought for megawatts, not for measured returns, and are integrated into a negative-ROIC operation; the Rhodium deal in particular was defensive.

The Bitcoin treasury strategy has flipped from HODL to net seller. From 19,287 BTC (Q3’25) to ~15,680 BTC (Q1’26), actively liquidated to fund capex. The treasury (~$1.0B) is now ~10% of market cap — meaningful but a minority of equity value — and is a depleting funding bridge, not a permanent store.

There is no real buyback. Despite headlines, no board-authorized repurchase with a dollar cap and execution is disclosed; management has referenced buybacks only as a possible future use of proceeds. Riot was a net issuer (+$212.7M of stock sold in FY2025). Do not credit any capital return.

Executive compensation is, surprisingly, better-aligned than the sector norm. The 2025 annual incentive plan weighted relative Bitcoin production (25%), relative adjusted EBITDA (25%), data-center revenue (15%), data-center NOI (15%), and strategic data-center execution (20%). The long-term plan vests on relative total shareholder return, with a negative-TSR cap (if absolute TSR is negative, vesting is capped at 100% of target) — a genuine guardrail, and the data-center NOI weighting discourages “grow hash at any cost.” Caveats: the AIP still rewards production/EBITDA growth that is funded by dilution, and part of the AIP is paid in Bitcoin (CEO Les took 12.5 BTC for 2025), entangling pay with the commodity bet. CEO Les’s FY2025 total comp was $3.46M (down from an anomalous $83.5M in FY2024 inflated by a one-time ~$79.3M equity mega-grant). Insiders hold meaningful stakes (Les ~9.4M shares; Executive Chairman Yi ~10.6M).

The insider signal is mildly bearish-to-neutral: programmatic selling, zero open-market buys. CEO Jason Les sold 175,000 shares at ~$25.19 (~$4.4M) via a 10b5-1 plan (adopted August 2025); CAO Werner sold 37,616 shares at $26.50 via a plan. The January 2026 Form 4 cluster was routine annual equity grants. No discretionary purchases (code P) appear in the trailing window. Management is monetizing equity into strength, not accumulating.

Verdict: Weak-to-mixed, and pivot-dependent. The dominant historical pattern is value-destructive — ~$3B of dilutive ATM plus ~$376M of SBC funding a negative-FCF, negative-ROIC commodity operation, with acquisitions bought for megawatts and a treasury now liquidated to plug the capex hole. What keeps this from “egregious”: genuinely cheap, scarce power; an unusually shareholder-aligned comp structure; the ESS Metron payoff; and the AMD lease as a real proof-point that the optionality could convert. But absent the pivot working and project finance closing, capital allocation has been a serial wealth transfer from new shareholders (ATM) and from the coin stack to fund a structurally unprofitable miner.

8. Changes and Headwinds — Last Two Years

The strategic transformation (2024–2026):

• 2024: Acquired Block Mining (Kentucky, +~1 EH/s). Starboard Value built a stake; CNBC reported (December 2024) Starboard pushing Riot to pivot power toward AI/HPC.
• February 2025: Board changes and a formal evaluation of AI/HPC uses of Riot’s power — the formal pivot inflection.
• April 2025: Rhodium settlement and $185M asset acquisition — 100% control of Rockdale, legacy hosting contracts terminated, all litigation dismissed (a one-time $158.1M P&L loss).
• Q3–Q4 2025: Corsicana land roll-up (~900–925 acres) and conversion of the Rockdale ground lease to fee-simple (~$96M, eliminating ~$130M of future rent); Corsicana core-and-shell construction initiated.
• January 2026: First AMD lease (25 MW, Rockdale). April 2026: AMD expanded to 50 MW; Coinbase facility amended and extended.
• Leadership: CFO transition (Colin Yee → senior adviser; Jason Chung, ex-Head of Corporate Development, promoted to CFO, March 2026); new data-center leadership hired from Oracle/Digital Realty and Google/Meta; three new independent directors and phased board declassification (all directors annually elected by 2029).

Starboard status: the activist campaign is active and escalated. Following the December 2024 stake, Starboard sent a public letter dated February 18, 2026 urging Riot to accelerate the AI/HPC transition (arguing the 1.7 GW could generate >$1.6B of annual EBITDA if monetized, naming Corsicana and Rockdale “premier” sites); shares rose ~7–9% on the news. No signed standstill/board-seat settlement has surfaced — the letter reads as public pressure. (Note: the “waiver and irrevocable proxy” that appears in searches relates to Riot’s stake in Bitfarms, a separate matter — not Starboard.)

Accounting: adoption of ASU 2023-08 (Bitcoin fair-value/mark-to-market) drives the GAAP volatility discussed in the Financial Quality section above.

Headwinds: the April-2024 halving (structurally doubling unit cost); record network hashrate/difficulty; a Bitcoin price that has fallen from the ~$87.5k FY2025-end mark to ~$64k; the unclosed project finance; and the simple fact that the central catalyst — a Corsicana hyperscaler lease — has not arrived after sixteen months.

Verdict: Mixed, tilting to a higher-variance, higher-optionality thesis. Strengthening: the signed investment-grade AMD lease, the Rhodium clean-up, fee-simple control of scarce energized power, and Starboard’s alignment pressure. Weakening/unproven: the unsigned Corsicana prize, a funding model dependent on a depleting treasury and unclosed project finance, a structurally pressured mining core, heavy SBC, and a stock that has already re-rated +152% partly on the narrative. The changes move Riot from “pure crypto-beta miner” to “optionality-rich, partially-contracted AI-infrastructure developer” — a better business if Corsicana lands a hyperscaler, but the thesis hinges on an event that has not occurred and a funding bridge that is not secured.

9. Risk Analysis

Catastrophic-loss risk is real, not merely tail. Risks 1 and 2 are correlated: a crypto winter that takes Bitcoin down would simultaneously starve the funding model, depress the treasury, and freeze HPC demand — while the 2.9-beta equity falls hardest. The −99.98% empirical drawdown (twice in the company’s history the equity has gone to near-zero) is the anchor for the bear case. This is not a business with a margin of safety; it is a leveraged option.

10. Valuation Discussion (Embedded Expectations)

Riot has no earnings to multiply and a Bitcoin-distorted book, so the honest frame is a sum-of-the-parts and an embedded-expectations read of what the ~$10.7B enterprise value is paying for. (Note: ROIC’s reported EV of ~$5.0B is stale — built on the Q1’26 quarter-end share price of ~$12.36, before the rally to $26.61. The live EV, on ~379M shares at $26.61 plus $636.5M net debt, is ~$10.7B.)

The four parts:

1. Bitcoin mining (as mining): negative-to-breakeven through-cycle economics; value sits in the power assets and ASIC fleet, not an earnings multiple. Crediting energized ERCOT capacity not earmarked for HPC plus the fleet at scrap-to-modest values gives ~$1.0–3.0B. Peer cross-check: MARA — a larger-hashrate, weak-HPC pure miner — carries only a ~$3.6B market cap, framing the ceiling.
2. Bitcoin treasury: ~15,680 BTC × ~$64k ≈ $1.0B (~10% of cap). This is a depleting asset (net seller), worth spot, not a HODL premium — and at the Q1’26 burn pace it is roughly a four-quarter funding bridge absent appreciation or project finance.
3. AMD contracted HPC: ~$51M stabilized NOI capitalized at a 7–9% data-center cap rate = ~$0.57–0.73B gross, less ~$0.17B remaining build capex = net ~$0.4–0.56B. The only contracted HPC value.
4. Unsigned HPC pipeline (Corsicana 756 MW + AMD options): the entire re-rating case — and uncontracted.

The embedded-expectations crux. Stripping the real parts:

EV $10.72B − BTC $1.0B − AMD net $0.5B − mining $1.0–3.0B = $6.3–8.3B residual paid for the unsigned pipeline.

Dividing that residual by the cohort’s own EV-per-contracted-MW range ($15–30M/MW) implies the market is paying Riot for ~200–550 MW of signed-equivalent hyperscale capacity — against 50 MW actually signed. The stock embeds roughly 4x–11x the contracted HPC capacity it has closed; in effect, essentially the entire 756 MW Corsicana campus (or the full AMD option ladder) is priced as a done deal before a single Corsicana lease exists. That is the crux: ~$6–8B of the ~$10.7B EV is unsigned optionality.

Peer comparison — EV per contracted MW (the cleanest cross-sectional anchor). The converted-miner cohort is valued primarily on contracted/operating HPC megawatts and tenant credit. On that metric Riot is the laggard, yet it carries a cohort-comparable EV:

The entire cohort already trades expensively on contracted MW (the converted-miner names broadly carry rich valuations at current prices), which is itself a Marathon capital-cycle warning: a wall of capital is racing to build the same capacity, and $/MW valuations are likely to compress as supply arrives. Riot is the one name in the group being asked to pay a cohort-grade EV with essentially none of the cohort’s contracted book. It screens “cheap” only if one assumes the unsigned Corsicana campus converts at the high end of these multiples — precisely the assumption the capital cycle argues against.

Scenario analysis (illustrative equity value per share on ~379M shares; no price target — embedded-expectations framing only):

The current $26.61 sits above the base (~$12) and well below the bull (~$46) — the market is underwriting Corsicana lease-up as more-likely-than-not and Bitcoin holding or rising. The risk is asymmetric and fat-tailed: the bear is a near-total loss (a high-beta crypto equity that has twice approached zero), while the bull requires multiple things to break right at once.

Multiples context: ~16x EV/sales, ~15.5x P/S, 3.9x P/B (book/share $6.89), with AZI own-history percentiles at the 83rd (P/B, richest-ever zone), 66th (P/S), and 74th (composite). GAAP P/E is meaningless (Bitcoin marks).

Verdict: Expensive on everything contracted; cheap only on optionality the company has not monetized. The honest read is a real option — a ~$2.5–4.5B “hard” floor (BTC + AMD + mining) against ~$10.7B of EV, with the remaining ~$6–8B a bet on Corsicana conversion and Bitcoin. Base-case SOTP (~$12) is roughly half the quote. The stock is not cheap; it is a levered call option on an event that has not occurred, funded by a shrinking coin stack.

11. Variant Perception

Consensus view. The bull narrative — voiced by Starboard, much of the sell side, and the tape — is that Riot is mid-transition from a moat-less Bitcoin miner into a high-value “AI power landlord”: its 1.7 GW of secured, energized ERCOT power plus in-house engineering plus the signed investment-grade AMD lease prove the model, and Corsicana’s 756 MW will convert to hyperscaler leases worth $1.6–2.1B of NOI, re-rating the equity toward the HUT/WULF/CIFR landlord multiple. The +152%/12m move and 83rd-percentile P/B say consensus prices this conversion as more-likely-than-not.

Strongest bull case. Scarce, energized, fee-simple 1.7 GW in power-constrained ERCOT is genuinely valuable in an AI-capex supercycle where power/interconnect is the binding constraint. The AMD lease (A-rated, already expanded once, with a 150 MW option ladder) is a real proof-point. Vertical integration is a demonstrated speed/cost edge. If Corsicana signs one hyperscaler/neocloud, Bitcoin stays firm, and Starboard forces a value-unlock, the SOTP bull is ~$40–46. Because the cohort already trades at $25–30M/contracted-MW, a single large signing could re-rate the whole stub quickly.

Strongest bear case. Strip the narrative and Riot is a structurally unprofitable commodity miner (negative through-cycle ROIC, cost-to-mine $45–50k/BTC, −53% operating margin) that has funded itself with ~$3B of ATM dilution, ~$376M of SBC, and a now-shrinking treasury, and whose only signed HPC tenant is 50 MW versus peers’ 500–907 MW. Corsicana has had zero signed tenants in sixteen months; management itself hedges. At $26.61 the market pays ~$6–8B for ~200–550 MW of signed-equivalent HPC that does not exist. If Corsicana never converts, the options lapse, project finance does not close, and a crypto winter takes the 2.9-beta equity down, the SOTP bear is ~$1–3. The funding model — selling a depleting coin stack into capex — breaks precisely when Bitcoin falls, the worst possible correlation.

The 3–5 assumptions that matter most:

1. Does Corsicana sign a creditworthy hyperscaler/neocloud — and when, at what $/kW/month, with who bearing build capex? This single event is ~$6–8B of the EV. Unsigned for sixteen months.
2. Bitcoin’s price path — drives the treasury mark, the funding model, and (via 2.9 beta) the equity. A crypto winter hits all three at once.
3. Does non-recourse project finance close on AMD (de-dilutive) or does the build stay ATM/BTC-funded (dilutive)? None closed as of 2026-06-14.
4. The dilution path — ~379M shares already, ITM converts at $14.86, a live $500M ATM. How much more equity before Corsicana (if ever) converts?
5. The cap rate / lease economics the market applies — cohort $25–30M/MW versus a normalizing $12–18M/MW as the capital flood compresses HPC rents.

What falsifies each side. Falsifies the bear (turns bullish): a signed Corsicana hyperscaler lease (named IG tenant, 100+ MW, build-capex clarity) and a closed non-recourse project-finance facility on AMD. Falsifies the bull (turns bearish): two or three more quarters with no Corsicana lease, plus Bitcoin breaking below ~$50k (forcing accelerated treasury sales or a dilutive raise), plus AMD declining further options.

Factor-positioning read. The tape says consensus is offside in how it is positioned. Riot still trades as a crowded, ~2.3–2.9-beta crypto-momentum name — it clusters by factor similarity with crypto-miner ETFs (WGMI, BKCH, BITQ) and miner single-names (MARA, HIVE, Bitfarms), not with data-center REITs or AI-infra names; it carries positive Momentum (+0.41 to +0.72) on negative Quality (−0.33), the textbook mean-reversion setup; and it runs ~61% idiosyncratic volatility. The market has paid the landlord multiple while the stock retains the miner risk profile (crypto beta plus a history of ~100% drawdowns), with the landlord cash flows still unsigned. That mismatch — landlord price, miner risk — is the variant-perception core, and it tilts the asymmetry toward the bear tail.

Verdict: Consensus prices Corsicana conversion plus a Bitcoin bull as the base case; the evidence supports neither as more-likely-than-not. The honest read is a fat-tailed real option: a small hard floor, enormous unsigned upside, and a bear tail that is a near-total loss. Fifty signed MW is not 756, and “landlord price / miner risk” resolves against the longs unless Corsicana actually signs.

12. Fact vs. Interpretation

13. Open Questions

1. Will Corsicana sign a creditworthy hyperscaler/neocloud, when, and on what terms (lease rate per kW/month, tenant credit, who funds build capex)? This single event is the bulk of the equity value and has been unsigned for sixteen months.
2. Has any non-recourse project-finance facility on the AMD lease closed post-Q1’26? (None seen through 2026-06-14.) This determines whether the next build phase is dilutive (ATM/BTC) or debt-funded.
3. Does Bitcoin selling continue at the ~3,778 BTC/quarter Q1’26 pace? At that rate the ~15,680 BTC treasury depletes in ~4 quarters absent appreciation or project finance.
4. Exact current Starboard ownership and whether any board-seat/standstill exists (the February 2026 letter reads as public pressure, not a signed settlement).
5. What is the latest (post-Q1’26) monthly Bitcoin production and treasury count, and is the cost-to-mine improvement (−26% QoQ to ~$44,629 in Q1’26) sustaining?
6. Will AMD exercise its remaining options (+150 MW pathway), or do they lapse?

14. What Must Be True

For the bull case to be right (equity worth materially more than ~$26.61):

• Corsicana must sign at least one large, creditworthy hyperscaler/neocloud lease at cohort-grade economics within the next several quarters.
• Non-recourse project finance must close (proving the build can be debt-funded, de-risking dilution).
• Bitcoin must hold or rise (supporting the treasury, the funding model, and the high-beta equity).
• The AI/HPC capital cycle must not compress $/MW valuations faster than Riot signs capacity.
• Falsification test: two or three more quarters with no Corsicana lease, and/or Bitcoin below ~$50k forcing a dilutive raise — would confirm Riot is a cash-burning miner whose “optionality” does not convert.

For the bear case to be right (equity worth a fraction of ~$26.61):

• Corsicana stays unsigned; the AMD options lapse; project finance does not close.
• Bitcoin weakens, simultaneously hitting the treasury, the funding model, and the 2.9-beta equity.
• Continued dilution erodes per-share value as the build burns cash.
• Falsification test: a named investment-grade hyperscaler signing 100+ MW at Corsicana with build-capex clarity, plus a closed non-recourse facility on AMD — would convert the optionality to contracted, financeable NOI and break the bear.

The single highest-information event for either side is identical: a signed (or conspicuously absent) Corsicana hyperscaler lease. Everything else is secondary.

15. Source Appendix

Primary sources: Riot Platforms FY2025 Form 10-K (filed 2026-03-02), Q1’26 Form 10-Q (2026-04-30), DEF 14A (2026-04-30), material 8-Ks (AMD lease 2026-01-16; AMD expansion + Coinbase facility 2026-04), Forms 4/144; Q1’26 / Q4’25 / Q3’25 earnings-call transcripts; Starboard Value letter (2026-02-18). Quantitative data: ROIC.ai (statements, ratios, EV, multiples), third-party valuation-percentile and factor-model data; public price history. Peer cohort for comparison: MARA, CIFR, WULF, IREN, APLD, HUT. Bitcoin spot and network data via public sources (CoinGecko, Hashrate Index), accessed 2026-06-14.

The analysis above is deliberately position-free and contains no buy/sell recommendation and no price target. The only stated position appears in the clearly-labeled “Claude’s Take” block at the top, which is the author’s own independent opinion.

APPENDIX A — Standard Diligence Questionnaire

Riot Platforms, Inc. (NASDAQ: RIOT) — as of 2026-06-14

Supplemental to the research memo. Fact / Interpretation / Assumption labels applied where it matters.

General

What thoughtful questions have other investors asked about this company? The dominant question is whether Riot is still a Bitcoin miner or has become an AI/HPC data-center landlord — and what fraction of the ~$10B market cap is justified by signed contracts versus unsigned optionality. Related: Will Corsicana (756 MW) sign a hyperscaler, and when? (Fact: zero signed after ~16 months.) Can the company fund its buildout without crushing dilution now that it is selling its Bitcoin treasury? (Fact: net seller; ~$3B ATM already.) Is Starboard’s value-unlock thesis (1.7 GW → >$1.6B EBITDA) achievable? (Interpretation: only if power converts to leases at cohort economics.) How correlated is the equity to Bitcoin in a downturn? (Fact: ~2.9 beta, −99.98% lifetime max drawdown.)

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? There are no real earnings — GAAP net income is a Bitcoin fair-value artifact. Operating income has been negative every year FY2022–FY2025. On the Bitcoin cycle, the mining business sits post-halving (unit cost +54% YoY) with Bitcoin (~$64k) below its FY2025-end mark (~$87.5k) — closer to a margin trough than a peak. (Fact.)

Driven by the external environment or internal actions? Predominantly external — Bitcoin price and network difficulty dwarf internal levers. Internal actions (cheap-power procurement, ESS Metron efficiency, the AMD lease) matter at the margin and for the pivot, but mining profitability is a price-taker’s fate. (Interpretation.)

How stable are revenues? Unstable. ~89% is commodity mining revenue (Bitcoin price × coins mined, both volatile); ~10% is engineering (a backlog-supported but lumpy industrial stream); the new data-center leasing is contracted and stable but immaterial today. (Fact.)

Outlook for products/services? Mining: structurally pressured (treadmill). Engineering: growing with data-center demand (backlog +302% YoY). Data-center leasing: the growth vector — but contracted run-rate is only ~$37.8M (exiting 2026) scaling to ~$55.6M (exiting 2027) from AMD; the rest is unsigned. (Fact.)

How big will this market be? The AI data-center power market is large and growing (power/interconnect-constrained). Bitcoin mining is a mature, zero-sum commodity market. Riot’s addressable opportunity is its own 1.7 GW; monetizing it as HPC is the prize. (Interpretation.)

Business Quality & Competitive Moat

Is the industry getting more or less competitive? Mining: more competitive (network hashrate +47% YoY, difficulty at all-time high). AI/HPC: intensely and increasingly competitive (the whole converted-miner cohort plus hyperscalers plus neoclouds building simultaneously). (Fact/Interpretation.)

How profitable is the business (ROIC, ROE)? Deeply negative through-cycle. FY2025 ROE −22.1%, ROA −16.8%, ROIC negative. The one positive net-income year (FY2024, ROE +4.3%) was a Bitcoin mark. (Fact, ROIC ratios.)

How profitable is the industry — competitors, barriers? Mining has no barriers to entry and converging, thin-to-negative economics. The AI-landlord business has real barriers (power scarcity, interconnect queues) but is being competed down. (Interpretation, Greenwald.)

Can the business be easily understood? The mining mechanics, yes. The valuation requires understanding Bitcoin fair-value accounting (ASU 2023-08), the SOTP, and the unsigned-optionality premium — moderately complex. (Interpretation.)

Can it be undermined by foreign low-cost labor? Not labor — but by foreign/lower-cost power. Riot’s edge is cheap ERCOT power, which is geographically specific; the AI-landlord business is anchored to US power and tenants. (Interpretation.)

Do brands matter? Not for mining (commodity output). For AI/HPC leasing, reputation/track record matters — the AMD signing is a credibility signal — but it is not a consumer brand. (Interpretation.)

Nature of competition? Mining: pure cost competition on power and fleet efficiency. AI/HPC: competition on speed-to-power, site quality, tenant credit, and lease economics. (Interpretation.)

Customers’ switching costs? Mining: none (no customers — output sold to a global market). AI/HPC: high once a tenant signs a multi-year lease and installs equipment — but Riot has only one such tenant (AMD). (Fact/Interpretation.)

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The land-and-power optionality (1.7 GW, fee-simple Corsicana/Rockdale) is carried at cost, well below its potential AI-monetized value — the entire bull case. The ESS Metron vertical’s strategic value is understated. (Interpretation.)

Off-balance-sheet liabilities? No major hidden liabilities identified; leases are largely capitalized. The in-the-money convertible notes ($14.86 strike) imply future share dilution not yet in the basic count. (Fact.)

How conservative is the accounting? Mark-to-market Bitcoin accounting is volatile, not aggressive — it injects large non-cash swings but is GAAP-compliant and arguably more transparent than the old impairment model. Mining revenue recognized at fair value on award day. SBC is large (~19% of revenue) and real. (Interpretation.)

How CapEx-hungry is the business? Extremely. Capex was $567M (FY2025) and $690M (FY2024); the Corsicana/HPC buildout is a multi-year, multi-billion-dollar sink. FCF is negative every year. (Fact.)

Capital Allocation & Management

How much FCF, and how is it used? None — FCF is negative every year (FY2025 −$567M). The company consumes capital, funded by ATM equity, SBC, Bitcoin sales, and debt. (Fact.)

Significant acquisitions recently? Yes — Rhodium Encore assets ($185M, 2025, defensive), Block Mining (Dec 2024, Kentucky), E4A (Dec 2024). Bought for megawatts, not measured returns. (Fact/Interpretation.)

Buying back shares? No — despite headlines, no authorized repurchase program; Riot was a net issuer (+$212.7M in FY2025). (Fact.)

Issuing large amounts of stock to insiders? Heavy SBC (~$125.7M FY2025, ~$376M cumulative); a one-time ~$79.3M CEO mega-grant inflated FY2024. (Fact.)

Compensation policy of directors/management? Better-aligned than the sector norm: LTIP vests on relative TSR with a negative-TSR cap; AIP weights data-center NOI/revenue and relative production. Caveat: part of AIP paid in Bitcoin. (Fact/Interpretation.)

Motivations of management? Insiders hold meaningful stakes (Les ~9.4M, Yi ~10.6M shares) but are selling (10b5-1 plans, no open-market buys). Comp structure encourages the value-accretive pivot, but pay is still entangled with the commodity bet. (Fact/Interpretation.)

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? No — a US C-corp common stock (Nevada-incorporated), NASDAQ-listed, standard 1099 treatment. (Fact.)

Dividend policy? None — no dividend; all capital reinvested (and more). (Fact.)

How profitable is the business? Through-cycle unprofitable on a cash/operating basis; “profit” appears only when Bitcoin marks turn positive. (Fact.)

Is net income diverging from cash from operations? Yes, massively and in both directions — net income is dominated by non-cash Bitcoin marks while operating cash flow is negative. Read adjusted EBITDA and segment economics, not net income. (Fact.)

Risks & Downside

What factors would cause the stock to decline? A Bitcoin selloff (2.9 beta); continued Corsicana silence; failed project finance; a dilutive raise; mining-margin compression; HPC $/MW compression; a momentum unwind (positive Momentum on negative Quality). (Interpretation.)

Risk of a catastrophic loss? Real, not merely tail. The equity has approached zero twice (−99.98% lifetime max drawdown); a crypto winter would hit the treasury, funding model, and equity simultaneously. (Fact/Interpretation.)

Chance of a total loss? Low in the near term given ~$1B treasury + ~$200M cash + scarce power assets that retain liquidation value — but a prolonged crypto winter combined with a failed pivot and heavy dilution could impair the equity by the large majority. The hard-asset floor (power/land) limits a literal zero. (Interpretation.)

Recent News & Events

Has the business environment changed recently? Yes — the formal AI/HPC pivot (Feb 2025), the AMD lease + expansion (Jan/Apr 2026), Starboard’s escalated public letter (Feb 2026), CFO transition (Mar 2026), and a Bitcoin pullback from ~$87.5k to ~$64k. (Fact.)

Significant acquisitions / divestitures? Rhodium asset acquisition (2025) consolidated Rockdale; legacy hosting wound down. (Fact.)

Change in accounting policies? Adopted ASU 2023-08 (Bitcoin fair-value/mark-to-market), driving GAAP volatility. (Fact.)

Recent changes — new markets, facilities, management? New data-center leadership (ex-Oracle/Digital Realty, ex-Google/Meta); Corsicana core-and-shell construction initiated; three new independent directors and phased board declassification; new $500M ATM (Dec 2025). (Fact.)

APPENDIX B — Source Appendix

Riot Platforms, Inc. (NASDAQ: RIOT) — Research date 2026-06-14

Primary sources first. All URLs accessed 2026-06-14 unless noted. Facts are reconciled to primary filings; third-party aggregated data (ROIC.ai, AZI, FactorsToday) is used for ratios/percentiles/factor loadings and cross-checked to filings.

Primary — SEC Filings (Riot Platforms, CIK 0001167419)

1. Form 10-K, FY2025 (filed 2026-03-02). Segment revenue, cost-to-mine, power cost, hash rate, BTC treasury rollforward, Bitcoin fair-value accounting (ASU 2023-08), debt stack, risk factors. https://www.sec.gov/Archives/edgar/data/1167419/000110465926022322/riot-20251231x10k.htm
2. Form 10-Q, Q1 2026 (filed 2026-04-30). Q1’26 balance sheet (cash $205.7M, debt $877.2M), BTC marks, treasury count, share count ~379M. https://www.sec.gov/Archives/edgar/data/1167419/000110465926053120/riot-20260331x10q.htm
3. DEF 14A proxy (filed 2026-04-30). Executive compensation (AIP/LTIP metrics, relative TSR with negative-TSR cap), beneficial ownership, board declassification, new directors. https://www.sec.gov/Archives/edgar/data/1167419/000110465926053098/riot-20260609xdef14a.htm
4. Form 8-K (2026-01-16) — first AMD data-center lease at Rockdale (25 MW). https://www.sec.gov/Archives/edgar/data/1167419/000110465926004551/riot-20260116x8k.htm
5. Form 8-K (2026-04-27/30) — AMD expansion to 50 MW; Q1’26 results. https://www.sec.gov/Archives/edgar/data/1167419/000110465926049432/riot-20260421x8k.htm and …/000110465926052943/riot-20260430x8k.htm
6. Form 8-K (2026-04-21) — Second Amended & Restated Coinbase Credit Agreement ($200M BTC-collateralized, fixed-rate, extended maturity).
7. Forms 4 / 144 — CEO Jason Les (Form 144 2026-05-11; Form 4 2026-05-13, 175,000 sh @ ~$25.19, 10b5-1); CAO Ryan Werner (Form 4 2026-05-29, 37,616 sh @ $26.50, 10b5-1); Jan-2026 grant cluster (code A). Via EDGAR filing index.
8. Form 10-K (FY2024) and prior — multi-year P&L/BS/CF (FY2021–FY2024) for trend analysis.

Primary — Earnings-Call Transcripts (via ROIC.ai MCP)

9. Q1 2026 earnings call (2026-04-30) — AMD expansion, Corsicana status (“I can’t tell you when our next lease will be signed”), BTC-sale funding model, cost-to-mine $44,629.
10. Q4/FY2025 earnings call (2026-03-02) — FY2025 production 5,686 BTC, cost-to-mine $49,645, CFO transition, treasury strategy.
11. Q3 2025 earnings call (2025-10-30) — Corsicana core-and-shell initiation, engineering backlog.

Primary — Company & Activist

12. Riot Platforms Investor Relations / press releases — AMD lease (2026-01-16), Q1’26 results (2026-04-30), $500M ATM (2025-12-30), Rhodium closing (April 2025). https://www.riotplatforms.com
13. Starboard Value LP — public letter to Riot Platforms, 2026-02-18 (1.7 GW → >$1.6B EBITDA argument). https://www.starboardvalue.com ; CNBC (2024-12-21) initial Starboard stake report.
14. Businesswire (2025-02-12) — board changes + formal AI/HPC evaluation announcement.

Quantitative Aggregators (cross-checked to filings)

15. ROIC.ai — income statement, balance sheet, cash flow, profitability ratios (ROE/ROA/ROIC/margins), per-share data, enterprise value, valuation multiples (FY2020–FY2025 + quarterly). Pull 2026-06-14.
16. AZI valuation-index — own-history percentiles: P/B 83rd, P/S 66th, composite 74th; P/B 3.86x, P/S 15.3x. Pull 2026-06-14.
17. FactorsToday — factor loadings (Market beta 2.27–2.49; Industry:Fintech +2.25; Momentum +0.41/+0.72; Quality −0.33), leaderboard (lifetime Sharpe −0.16, max DD −99.98%; y3 +38.5%/yr), stock-info (beta 2.94, RS_12m +152), related-stocks (WGMI/BKCH/MARA cluster), specific vol (~61%). Pull 2026-06-14.
18. AZI price history CSV — close $26.61 (2026-06-12), EMAs 25.03/22.28/17.44, 52-wk range $8.87–$28.94. https://azitrading.com/controls/download-data.php?t=RIOT

Market / Industry Data

19. Bitcoin spot ~$64k (2026-06-14, CoinGecko; ~$63.5k 2026-06-12 per Fortune/CoinDesk). Network hashrate ~1 ZH/s, difficulty ATH, hashprice ~$37–38/PH/day (Hashrate Index, June 2026).

Peer Cohort (for comparison)

20. Public filings and disclosures of AI-pivot Bitcoin-mining peers MARA, CIFR, WULF, IREN, APLD, and HUT — used for the EV/contracted-MW comparison and cohort framing.

Management commentary (transcripts, IR, activist letters) is treated as hypothesis and validated against filings, financials, and external data. Third-party aggregated figures are starting points reconciled to primary SEC filings; where they disagree, the filing governs.