Riot Platforms
Vertically integrated: self-mine Bitcoin using owned ASIC fleet at company-owned Texas sites, hold mined BTC on the balance sheet as a treasury, and increasingly lease power/data-center capacity to AI/HPC tenants under multi-year operating leases. Revenue = Bitcoin mining (dominant) + data-center/engineering + a small power-credit/demand-response stream from curtailing during ERCOT peaks.
Earnings, margins, COGS & capex
FY2025 was a record top line ($647.4M, +72% YoY) driven by more Bitcoin mined (5,686 BTC vs 4,828) and higher BTC prices, yet the company posted a $663.2M net loss (vs $109.4M net income in FY2024) dominated by non-cash items — $346.8M D&A, a contract/legal-settlement loss, stock-based comp, and unrealized BTC mark-to-market. Q1 2026 revenue of $167.2M beat estimates (~$122M) as data-center revenue scaled, but a ~$500.5M net loss (-$1.44/sh) landed as a falling BTC price marked down the treasury (~$326.7M fair-value decline plus a ~$51.9M derivatives loss) and pushed fully-loaded mining economics underwater. The story is a deliberate transition from pure-play miner to AI/HPC landlord using the same low-cost Texas power.
Income statement — where each revenue dollar goes
% of revenueOf every $1 of revenue, ~0¢ is cost of goods and ~0¢ operating expense, leaving ~100¢ of operating profit (~-102¢ net).
Revenue trend
Margins
negative, non-cash-driven
up from ~$32,216 in 2024 (post-halving difficulty)
underwater on fully-loaded basis
COGS structure
Dominated by electricity/power cost, hosting/site opex, and heavy depreciation of the ASIC fleet and infrastructure ($346.8M D&A in FY2025). Post-April-2024-halving difficulty growth (a ~47% rise in global network hash rate cited for the higher unit cost) has structurally raised cost-per-coin; Riot offsets via low ERCOT power prices and power-curtailment credits.
Capex
Growth capex concentrated on the Corsicana, TX campus (~756 MW planned critical IT capacity), a 600 MW substation (400 MW targeted Q1 2026, remaining 200 MW H2 2026), building core-and-shell (a single 168 MW standard-design building, up from a planned 112 MW across two, targeted 2027), plus ongoing miner purchases. Core-and-shell capex guided flat while delivering ~50% more critical IT capacity per building.
Latest earnings
Revenue beat (~$167.2M vs ~$122M est); larger-than-expected net loss (~-$1.44/sh, ~$500.5M) driven by non-cash BTC markdown; stock rose on the data-center momentum and AMD 50 MW news
Operating-lease revenue ~$37.8M run-rate exiting 2026, ~$55.6M in 2027 after full AMD delivery; 600 MW substation completion through H2 2026; continued evaluation of ~600 MW for additional AI/HPC leases
- BTC produced Q1 2026
- 1,473 (~16.6/day)
- Deployed hash rate
- 42.5 EH/s (~4.3% of network)
- All-in cost to mine
- ~$96,283/BTC
- BTC held
- ~15,679 (~$1.1B at Q1-end mark)
- Cash
- ~$233.5M
Growth drivers
- AI/HPC data-center leasing — AMD anchor tenant doubled contracted capacity from 25 MW to 50 MW; formal evaluation underway for the remaining ~600 MW at Corsicana
- Operating-lease revenue ramp: guided ~$37.8M exiting 2026, ~$55.6M after full AMD 50 MW delivery in 2027
- Hash-rate growth: 31.5 EH/s (YE2024) to 38.5 EH/s (YE2025) to 42.5 EH/s deployed (Q1 2026)
- Bitcoin price appreciation on ~15,700 BTC treasury (operating leverage to BTC upside)
- Low-cost Texas power + demand-response credits improving mining unit economics vs peers
Reported financials — SEC EDGAR
Audited GAAP figures pulled from SEC filings · latest filing 2026-03-02. The audited primary-source spine — not financial advice.
Revenue — annual (GAAP)
Margins & balance sheet — FY’25
Bull & bear
RIOT is a cheap call option on two secular tailwinds at once — Bitcoin and AI power scarcity — sitting on a rare ~756 MW powered land bank in Texas with an AMD anchor already proving the AI/HPC lease model.
- The Corsicana pivot re-rates the stock: contracted AI/HPC lease revenue (guided ~$37.8M exiting 2026, ~$55.6M 2027) is recurring, high-margin, and valued at infrastructure multiples, not miner multiples
- AMD doubling to 50 MW is external validation; the remaining ~600 MW under formal evaluation is unpriced optionality
- Balance sheet is a fortress for the sector: ~15,700 BTC (~$1.1B) + ~$233.5M cash, modest debt (mainly $594.4M converts due 2030)
- Operating leverage to BTC: a rising Bitcoin price lifts mining revenue and marks the treasury up simultaneously
- Best-in-class power strategy (owned sites, curtailment credits) keeps Riot lower on the mining cost curve than most peers
This is still, today, a money-losing Bitcoin miner whose economics went underwater at Q1 2026 BTC prices, funding an ambitious AI/HPC build with dilution and debt while only ~50 of ~756 MW is actually leased.
- Q1 2026 fully-loaded cost to mine (~$96k) exceeded per-coin value (~$76k) — the core business lost money on each coin
- $663.2M FY2025 net loss and ongoing negative FCF (~-$774M FY2025); the AI story is mostly a 2027+ promise, not current cash flow
- AI/HPC pivot faces fierce competition (CoreWeave, established data-center REITs, hyperscalers self-building) for tenants and interconnects
- Chronic dilution — ATM equity issuance plus $594.4M converts — steadily erodes per-share value
- GAAP earnings are hostage to BTC mark-to-market; a crypto bear market craters both revenue and the treasury at once
- Valuation already embeds substantial AI-optionality that may not convert on the guided timeline
What it is worth
Sum-of-the-parts: (1) mining segment on EV/hash or normalized mining cash flow, (2) BTC treasury marked to market (~$1.1B), (3) AI/HPC data-center capacity on a per-MW / contracted-lease-multiple basis. Blend with a scenario on how much of the ~756 MW converts to leases.
AI/HPC conversion slips, BTC drawdown marks down the treasury and pushes mining further underwater, dilution continues — the stock de-rates toward a distressed-miner multiple.
AMD 50 MW plus incremental leases ramp on the 2026-2027 guided path (~$37.8M to $55.6M lease revenue) while mining stays margin-thin — value roughly supported around the current cap, driven by execution and BTC price.
Most of Corsicana's ~756 MW converts to AI/HPC leases at data-center economics + BTC rallies — the stock re-rates to an AI-infrastructure multiple, well above the current ~$8-9B cap.
Market cap (~$8-9B mid-2026) already prices in meaningful AI/HPC optionality well beyond current mining cash flow; the debate is how much of the ~600 MW un-leased capacity converts and at what economics. Miner-multiple vs infrastructure-multiple is the core re-rating question.
SWOT
Strengths
- One of the largest US-listed self-mining fleets (42.5 EH/s deployed) with company-owned Texas sites and low-cost power
- Large ~15,700 BTC treasury (~$1.1B) plus cash — real optionality and balance-sheet strength
- Corsicana campus (~756 MW planned) is a scarce, powered land bank exactly where AI/HPC demand is bottlenecked on power
- AMD anchor tenant validates the AI/HPC pivot — leases convert volatile mining revenue into contracted recurring revenue
- Sophisticated power-market strategy (ERCOT curtailment/demand-response credits) lowers effective power cost
Weaknesses
- Fully-loaded mining economics went underwater in Q1 2026 (~$96k cost vs ~$76k value/coin)
- Persistent large net losses ($663.2M FY2025) and negative free cash flow (~-$774M FY2025) during heavy build-out
- Earnings dominated by non-cash BTC mark-to-market — GAAP results swing wildly with crypto price
- AI/HPC pivot is early: only ~50 MW leased vs ~756 MW planned; execution + tenant conversion unproven at scale
- Heavy shareholder dilution history (equity/ATM issuance and $594.4M converts) funding the build
Opportunities
- Convert the remaining ~600 MW at Corsicana into AI/HPC leases at data-center economics far above mining margins
- Ride structural AI power scarcity as a powered-shell landlord to hyperscalers/neoclouds
- BTC price appreciation levered through the treasury and rising hash rate
- Potential to monetize/spin the data-center segment at infrastructure multiples vs miner multiples
Threats
- Bitcoin price drawdowns hit both mining revenue and the treasury simultaneously
- Post-halving network difficulty keeps compressing mining margins industry-wide
- Competition for AI tenants and power interconnects from better-capitalized data-center players
- Regulatory/energy-policy risk (ERCOT rules, mining scrutiny, potential BTC tax/accounting changes)
- Rising rates / capital-market access risk for a capital-intensive, cash-burning model
Moats, dependencies & bottlenecks
Moats
Moderate-to-strong ~756 MW of secured power at Corsicana in ERCOT is genuinely scarce; new interconnects take years. This is Riot's most durable advantage.
42.5 EH/s and curtailment credits keep costs low, but hash rate is commoditized and difficulty erodes any lead.
~$1.1B BTC + cash funds the build without desperate dilution, but it is a war chest, not a structural moat.
First AI/HPC lease creates a reference customer, but not yet switching-cost lock-in across a tenant base.
Dependencies
Revenue + asset value Both mining revenue and the ~15,700 BTC treasury move with BTC; a drawdown is a double hit.
Input + regulatory Effective power cost and curtailment credits depend on ERCOT pricing and rules; policy shifts change the model.
Hardware supply chain Mining fleet renewal depends on China-based ASIC makers (context only); tariffs/export controls are a supply risk.
Customer concentration The AI pivot's near-term contracted revenue is concentrated in one 50 MW tenant.
Tenant demand for the remaining ~600 MW depends on the broader AI-GPU buildout continuing.
A cash-burning, capex-heavy model relies on continued equity/convert access.
Advantages
- Scarce, secured Texas power capacity positioned for AI demand
- Vertically integrated owned sites (not a hosting tenant) — controls power, land, and buildout
- Strong sector balance sheet (large BTC + cash, modest leverage)
- Proven power-market/curtailment sophistication lowering cost
- Early-mover, tenant-validated AI/HPC optionality (AMD)
Weaknesses
- Core mining unit economics currently underwater on a fully-loaded basis
- Large GAAP losses and negative free cash flow
- Heavy dependence on volatile BTC for both revenue and asset value
- Ongoing shareholder dilution
- AI/HPC revenue still nascent vs. the scale of the promise
Bottlenecks
- Converting the remaining ~600 MW at Corsicana into signed, revenue-generating AI/HPC leases
- Substation/power delivery timeline (400 MW Q1 2026, remaining 200 MW H2 2026) gating both mining and data-center growth
- Post-halving mining difficulty squeezing per-coin economics until either BTC price rises or the fleet is upgraded
- Capital availability to fund the build without excessive dilution
- Data-center construction/commissioning execution to hit the 168 MW core-and-shell 2027 target
Top signals & trends
Top signals
Concrete demand signal for the AI/HPC pivot.
Data-center segment scaling ahead of expectations.
Mining core lost money at prevailing BTC prices.
Mostly non-cash BTC markdown, but headline pressure persists.
Unlocks capacity for both mining and leasing.
Dilution funding the build.
Trends
Powered sites re-rated as AI-infrastructure; Riot is squarely in this move (peer to CORZ/CoreWeave, IREN, WULF).
Structurally higher cost-per-coin pressures pure mining.
Riot's secured Texas MW become more valuable as interconnect queues lengthen.
Amplifies both upside and drawdown through mining + treasury.
Makes GAAP earnings swing with crypto prices each quarter.
Ecosystem & competitor graph
Suppliers feed the company; customers pull from it. Line thickness shows the strength of each tie (supply-chain dependency, customer earnings contribution). Hover to isolate a tie.
Primary ASIC supplier (China-based; context only, not a buy/own call).
Alternative ASIC supplier (China-based; context only).
GPU supply underpinning AI-tenant demand for Riot's HPC capacity.
AI-accelerator maker and Riot's anchor data-center tenant.
Power supplier and demand-response counterparty central to Riot's cost structure.
Anchor AI/HPC lease tenant (50 MW contracted).
Mining rewards + fees are effectively the 'customer' for the mining segment.
Targeted for the remaining ~600 MW at Corsicana under evaluation.
Largest US-listed miner by hash rate and BTC treasury; direct scale rival.
Efficient US pure-play miner focused on low-cost power; strong operator.
Miner-turned-AI-host; CoreWeave relationship made it the template for the HPC pivot Riot is chasing.
Miner aggressively expanding into AI cloud/HPC with owned power.
Low-cost power miner leaning hard into AI/HPC hosting deals.
Texas-based miner also pursuing HPC/AI data-center conversions.
Diversified miner/infrastructure operator with power + compute assets.
Vertically integrated miner + ASIC designer; global scale.
Pure-play AI neocloud — the AI/HPC tenant/landlord competition Riot's data-center segment must win against.