NextFin News - Riot Platforms, long one of Wall Street's most visible "hodl" miners, has quietly become a net seller of its own Bitcoin. In the first quarter of 2026, the company sold 3,778 bitcoin for $289.5 million — roughly 2.6 times the 1,473 coins it mined — as its treasury fell from 18,005 BTC at the end of 2025 to 15,679 BTC on March 31. By June 30 the balance had dropped again, to 11,380 BTC, even as Riot signed a 20-year, $9.1 billion data-center lease with a frontier AI laboratory. The two facts together are the story. Riot is not merely raising cash to survive a weak Bitcoin quarter; it is converting a volatile, concentrated crypto treasury into contracted, long-duration AI infrastructure revenue, and in the process rewriting the playbook for what a public Bitcoin miner is supposed to be. Shares have noticed: Riot's stock is up more than 80% year-to-date through late September, far outpacing the S&P 500's roughly 13% gain, as investors price the company less like a mining operation and more like a data-center landlord.
The Selling Is Real, and It Is Bigger Than Production
The filings leave little room for ambiguity. In the three months ended March 31, 2026, Riot sold 3,778 BTC for proceeds of approximately $289.5 million — an average of about $76,600 per coin — against quarterly production of just 1,473 BTC. The company was liquidating previously accumulated coins, not simply harvesting freshly mined rewards. Its treasury fell 13% in the quarter, and the pressure continued into the second quarter: Riot ended June with 11,380 BTC, valued at approximately $666 million at the June 30 Bitcoin price of $58,527, alongside $548.9 million in cash and more than $1.2 billion in total liquid assets.
Riot has been candid about the mechanics. In its March 10-Q, the company disclosed that "the recent volatility and downward trend in the market price of Bitcoin" had reduced "the purchasing power of our bitcoin holdings," and warned the decline "may necessitate the sale of a greater volume of our bitcoin than previously anticipated to fund our ongoing operations and working capital needs." That is the cyclical leg of the story in one sentence: when Bitcoin fell from $68,222 at the end of the first quarter to $58,527 at the end of the second, a miner with large capital plans suddenly needed to sell more coins to fund the same dollar amount of construction.
But the same filing describes a deliberate policy rather than a panic. Riot wrote that its Bitcoin Treasury Strategy is "designed to balance long-term value appreciation with operational flexibility and liquidity management," and that it "selectively sell[s] or leverage[s] portions of our bitcoin holdings... to fund operational needs, capital expenditures, and strategic initiatives, particularly when market conditions present opportunistic pricing above predetermined thresholds." The company has formalized what used to be heresy in mining circles: the coin pile is working capital, not a sacred reserve.
The Other Half of the Trade: $9.8 Billion of Contracted AI Revenue
The selling would be one story if Riot were simply burning cash. It is not. The proceeds are flowing into a data-center build-out that is already producing contracted revenue. In the second quarter, Riot reported total revenue of $174.2 million, up 14% from $153.0 million a year earlier. The composition is what matters: Bitcoin Mining revenue fell to $113.7 million from $140.9 million, while Engineering revenue rose to $37.3 million from $10.6 million, and the new Data Center segment contributed $23.2 million in its second full quarter of revenue. Revenue is growing even as the mining line shrinks.
The strategic anchor is the Rockdale, Texas campus. Riot completed delivery of the initial 25 megawatts to Advanced Micro Devices on time and on budget, converting that lease to recurring revenue, with a 25 MW expansion under construction — 10 MW due in November 2026 and 15 MW in May 2027, bringing AMD's total to 50 MW. Then, in August, the company announced a 191 MW lease with a frontier AI laboratory, a 20-year contract running to June 2048 worth approximately $9.1 billion in initial revenue, or as much as $16.1 billion if both five-year extension options are exercised. Together with AMD, Riot has now contracted 241 MW of critical IT capacity with two of the largest names in the AI ecosystem, representing roughly $9.8 billion of long-term contracted revenue.
"Today's announcement of a landmark 20-year, 191-megawatt data center lease with a leading frontier AI lab marks a defining moment in our evolution into a leading developer of large-scale data centers," CEO Jason Les said in the second-quarter release. "In just over six months, Riot has now executed leases totaling 241 megawatts of capacity, representing approximately $9.8 billion of long-term, contracted revenue with two of the most important companies in the AI ecosystem."
The filing language is even more explicit about the destination: Riot describes its "strategic evolution from a bitcoin mining-focused enterprise to a diversified data center and digital infrastructure company." The Bitcoin sales are the funding leg; the AI leases are the asset leg. One cannot be understood without the other.
The Mechanism: From Hashprice Volatility to Contracted Cash Flows
The deeper question is why this trade makes economic sense at all. The answer lies in the difference between two kinds of revenue. Bitcoin mining revenue is a function of hashprice — the dollar value of block rewards net of energy cost — which swings with Bitcoin's price, network difficulty, and power rates. It is inherently volatile, difficult to hedge at scale, and priced by equity markets at a steep discount. A mining stock trades not on its coin pile but on the discounted, uncertain stream of future block rewards.
Data-center lease revenue is different. Once a hyperscaler or AI lab signs a long-term contract, Riot earns a contracted stream of rent and services revenue backed by large tenants, with the power capacity already fully approved and, increasingly, already energized. The company's filings describe a fully approved power pipeline of roughly 2.0 GW across more than 1,300 acres. Permitting is the binding constraint in U.S. data-center development, and Riot has already cleared it. The margin on that revenue is far more predictable than the margin on a block reward, and the market assigns it a different — typically higher — multiple.
So the mechanism is an asset-class arbitrage conducted inside a single balance sheet. Riot sells an asset whose future cash flows are volatile and discounted, and reinvests the proceeds into an asset whose cash flows are contracted and re-rated. Each coin sold funds kilowatts; each kilowatt leased to an AI tenant replaces an uncertain future block reward with a certain contractual payment. The trade only works if the return on the deployed capital exceeds the expected appreciation of the coin, adjusted for risk. Riot's management is effectively arguing that $9.1 billion of contracted revenue — with estimated cumulative net operating income of $7.3 billion to $8.2 billion over the base term and an estimated $365 million to $411 million in average annual NOI — is worth more than the 11,380 BTC still sitting in the treasury.
There is also a timing element, and it explains the urgency. The AI lease is back-loaded: the initial 96 MW is not expected until December 2027, with full 191 MW deployment by June 2028. That gap between spending today and receiving revenue years from now is exactly why the Bitcoin sales were necessary — and why the company also raised more than $600 million in follow-on equity in late 2025 and secured a $573 million interim financing facility from Morgan Stanley for the AI build. Riot is bridging a multi-year capital valley with a volatile asset before the contracted revenue begins to compound.
This Is Not Just Riot: The Industry Is Rotating
Riot is the cleanest public example, but it is not acting alone. MARA Holdings, the largest publicly traded Bitcoin miner, sold 23,093 BTC for approximately $1.63 billion in the first half of 2026 and then pledged more than half of its remaining stash as collateral for $600 million in loans, directing the capital toward a $1.5 billion acquisition of the Long Ridge energy plant in Ohio and a Texas AI data-center development. CleanSpark sold roughly 553 BTC for $36.6 million in February 2026. Industry researcher CoinShares estimated in the first quarter that public miners already derived around 30% of revenue from AI and high-performance-computing workloads, with the share projected to reach roughly 70% by year-end as more contracts come online.
The rotation is structural, not coincidental. The 2021 miner playbook — mine, hold, repeat — made sense when Bitcoin's price trajectory was the dominant return driver and when miners' power infrastructure had no higher-value use. Neither condition holds today. Bitcoin has matured into a macro asset whose quarterly moves are driven by ETF flows, rate expectations, and liquidity cycles rather than miner accumulation. Meanwhile, the AI build-out has created a competing, contracted bidder for the same scarce input: low-cost, high-capacity power. When a miner can lock in a 20-year lease with an AI lab, the opportunity cost of simply holding coins rises sharply. The "hodl" ethos was a product of its time; the time has changed.
The Counter-Thesis: Selling at the Bottom to Fund a Promise
The strongest argument against this strategy is also the simplest: Riot is liquidating a hard asset at cyclical lows to fund a promise. Bitcoin traded at $68,222 at the end of March 2026 and $58,527 at the end of June — well below the prior cycle's highs — and by late September had recovered to roughly $84,000. A miner that sells coins to pay for construction today is making an implicit bet that the coins' future appreciation will be lower than the return on the built infrastructure. If Bitcoin instead enters a sustained bull phase, Riot will have swapped a leveraged call on the asset for a fixed lease rate, and shareholders may look back on the sales as value destruction at the worst possible moment.
The critique has deep roots in how mining equities are traditionally underwritten: selling inventory caps upside participation while leaving full downside exposure to the asset. Riot's own filing concedes the pressure — the warning that falling Bitcoin prices "may necessitate the sale of a greater volume" is an admission that the strategy is, in part, defensive rather than purely opportunistic.
There is also execution risk on the AI side. The $9.1 billion is contracted revenue spread over 20 years, not cash in hand. Delivery is phased through 2028; any delay in construction, interconnection, or tenant fit-out pushes revenue further out while capital costs continue to accrue. The $573 million Morgan Stanley facility is interim financing pending an "investment-grade credit backstop" — if that backstop fails to materialize on acceptable terms, Riot's funding cost rises. And the AI demand boom itself is not guaranteed: if hyperscaler capital expenditure slows or if power-rich competitors flood the market with new capacity, contracted pricing power could erode faster than the long-term terms suggest.
The falsifying signal is quantifiable. If Bitcoin reclaims and holds above $90,000 through the second quarter of 2027 while Riot's treasury continues to draw down toward single-digit-thousands of BTC, the "rational reallocation" narrative fails: the company will have sold an appreciating asset at a discount to fund revenue that has not yet arrived. Conversely, if the treasury stabilizes above roughly 10,000 BTC and the Rockdale deliveries come online on schedule in late 2027 with the AI segment contributing a majority of revenue, the pivot is validated.
What to Watch: Three Time Horizons
Short term (next one to two quarters): liquidity and the pace of sales. Watch Riot's monthly Bitcoin production-and-sales disclosures and the treasury balance. The company ended the second quarter with 11,380 BTC and $548.9 million in cash; a continued steep drawdown would signal that Bitcoin's price is still dictating the funding plan rather than strategic choice. Also watch Bitcoin's price path — a sustained move back above $85,000 reduces the volume of coins Riot must sell to fund the same capital program.
Medium term (2027): delivery execution at Rockdale. The initial 96 MW is due in December 2027. Any slippage there is the first concrete sign that the AI revenue story is running behind the capital spend. Also watch the conversion of the AMD expansion — the 10 MW phase in November 2026 and 15 MW in May 2027 — as a leading indicator of Riot's ability to deliver on time and on budget.
Long term (2028 and beyond): revenue mix and multiple re-rating. The structural question is whether Riot trades as a mining stock or an infrastructure stock. If by 2028 the AI and data-center segments contribute the majority of revenue and net operating income — as the company's estimated $365 million to $411 million average annual NOI contribution from the AI lease alone implies — the equity should re-rate toward a contracted-infrastructure multiple rather than a hashprice multiple. If mining remains the dominant profit pool, the Bitcoin sales will look like a costly detour.
Base case: Riot delivers the Rockdale capacity on schedule, the treasury stabilizes in the low-to-mid five-figure BTC range, and the stock gradually re-rates as an infrastructure name with a mining option attached. Upside case: AI demand stays strong, contracted pricing holds, and Riot's permitted 2.0 GW pipeline becomes a scarce, re-rated asset — the Bitcoin sales then look like the cheapest possible funding source in hindsight. Downside case: Bitcoin rips higher while AI capex slows, leaving Riot short coins and long construction.
Riot's Bitcoin sales are not a retreat from crypto. They are a bet that permitted power, delivered on time to AI tenants, will compound faster and more safely than a coin held in a vault. The market is being asked to judge the company not on how many coins it owns, but on how many megawatts it can deliver.
Explore more exclusive insights at nextfin.ai.
