Bitcoin miners become AI data centers: the power-landlord pivot reshaping crypto

From hash rate to gigawatts

Bitcoin mining was always an energy arbitrage business dressed in cryptography. The winning formula: secure cheap power, build substations and cooling faster than anyone else, and run fixed-function ASICs that turn electrons into hashes. Margins were thin, halving cycles were brutal, and stock prices tracked Bitcoin beta — but the physical infrastructure was real.

Artificial intelligence changed what that infrastructure is worth. Training and inference clusters need the same things miners already built: high-voltage interconnections, liquid or immersion cooling, acreage near transmission lines, and operational teams that understand 24/7 uptime at megawatt scale. The bottleneck for AI data centers is not GPUs alone; it is time to power. Grid connections that take hyperscalers three to five years can be standing ready at a converted mining site in months.

According to Bernstein research cited by Decrypt, miners have signed 17 deals worth more than $110 billion over the past two years, contracting roughly 6 gigawatts of power to AI tenants — about 10% of all U.S. AI data center capacity currently under construction. The firm initiated coverage on TeraWulf and Cipher Digital with Outperform ratings, projecting aggregate AI revenue across its coverage to grow from $1.2 billion in 2026 to $10.7 billion by 2030.

CoinShares' 2026 outlook, summarized by ETF Trends, offers a starker revenue split: for companies that secured AI contracts, mining could fall from roughly 85% of revenue in early 2025 to less than 20% by year-end 2026. AI colocation deals reportedly generate about three times the revenue per megawatt compared to commodity mining, with operating margins in the 80–93% range versus the single-digit economics of hashing at scale.

What actually gets converted — and what does not

Headlines about miners "abandoning Bitcoin" often confuse two different asset classes. A SHA-256 ASIC cannot run a transformer model; it has no tensor cores, no high-bandwidth memory, and no general-purpose compute. When Core Scientific, IREN, Cipher, or Hut 8 announce a conversion, they are re-leasing the building shell — power capacity, substations, cooling, permits — not repurposing the silicon.

The displaced ASICs flood secondary markets, depressing hardware prices and pushing hashrate toward smaller operators and home miners — a second-order decentralization effect explored in detail by D-Central's Hashcenter migration analysis. Corporate megasites shrink their mining footprint; the network's hash power does not disappear, but it migrates downmarket.

Q1 2026 earnings season made the shift visible in filings. TheEnergyMag reported that Core Scientific accelerated colocation for CoreWeave, Cipher shut mining at portions of its Black Pearl site after securing a hyperscale lease, and IREN repositioned as an AI cloud operator with multi-billion-dollar processing agreements while scaling back mining. Hut 8 signed a 15-year, $9.8 billion lease for its 352-megawatt Beacon Point campus in Texas, lifting contracted AI capacity to roughly 597 megawatts. On earnings calls, leadership at several names stated plainly that Bitcoin is no longer the long-term strategic focus — integrated power and compute is.

This connects to the broader infrastructure squeeze we covered in our gaming memory shortage analysis: the same AI buildout competing for DRAM and NAND is also competing for the electrons and square footage that used to mine blocks.

Selling the treasury to build the future

For years, public miners held Bitcoin on balance sheets as a conviction trade — accumulated coins as strategic reserves, mirroring Michael Saylor's corporate treasury playbook. That culture broke in 2026.

Bitget News reported that listed miners collectively reduced treasuries by more than 15,000 BTC from peak levels. Core Scientific sold roughly $175 million worth (about 1,992 BTC) in March 2026 to fund operational transitions. The sales are not distressed liquidations in the traditional sense; they are balance-sheet reallocation — exchanging an asset the market is de-rating for concrete AI infrastructure with contracted hyperscaler tenants.

That selling pressure lands on a spot market already weakened by ETF outflows and leverage liquidations, as we detailed in our Bitcoin ETF outflows and AI rotation piece. The miner treasury unwind is a separate channel with the same macro driver: capital is leaving Bitcoin because AI infrastructure offers higher expected returns and more predictable cash flows. Strategy's 32-BTC sale for preferred dividends grabbed attention for symbolism; miners selling thousands of coins for data center capex is a volume story.

Our Bitcoin fundamentals guide explains why supply dynamics matter independently of narrative — and miner treasury sales are a supply dynamic, even when the sellers still believe in the asset long term.

Warm shells and the time-to-compute race

Bernstein coined the term warm powered shells — facilities where electricity is already flowing, transformers are commissioned, and cooling is operational, waiting only for GPU racks to arrive. Hyperscalers racing to deploy inference capacity ahead of competitors will pay premium rents for months saved on interconnection queues.

Miners' planned 30-gigawatt power portfolio, in Bernstein's framing, positions them as the fastest path from land to compute. TeraWulf's partnership with Fluidstack and Google is projected to reach $1.7 billion in AI revenue by 2030 with EBITDA margins near 84%. Cipher Digital, with a majority-hyperscaler client base, could hit $1.2 billion at margins approaching 93%.

The irreversibility argument matters for Bitcoin holders. Once a site is wired for high-density GPU power — different voltage rails, different cooling, multi-year tenant contracts — flipping back to ASIC mining is economically irrational even if Bitcoin rallies. Mining may continue as a flexible back-of-house workload that absorbs curtailed power, but the corporate identity of these companies has already changed on investor calls and in capital allocation.

What this means for June 2026 and beyond

The miner pivot does not invalidate Bitcoin's settlement layer or its scarcity schedule. It does reshape who holds price risk and where marginal supply originates. Three implications stand out for anyone reading charts this week:

First, corporate treasury selling from miners adds a persistent spot overhang distinct from ETF redemptions. Both channels reflect the same rotation trade, but they do not turn off simultaneously — miners fund construction on multi-quarter timelines.

Second, network hashrate may stay elevated even as public miners shrink their share, because secondary-market ASICs find new homes. Security of the chain and economics of listed equities are diverging.

Third, the AI buildout is not purely a stock-market story. Physical power, land, and cooling — the hard assets miners spent a decade accumulating — are now strategic inputs for the same boom draining liquidity from crypto markets ahead of SpaceX, OpenAI, and Anthropic IPOs. Understanding one without the other misreads June 2026.

For position sizing through a choppy summer, our risk management guide treats structural industry shifts the same as macro shocks: reduce size when multiple independent sell channels align, and avoid conflating "the network works" with "my entry price was right."

Bitcoin miners are not leaving crypto. They are becoming something larger — infrastructure landlords for the most capital-intensive technology cycle since the internet. The pivot explains part of why Bitcoin feels orphaned at $60,000 even as on-chain settlement keeps clearing blocks every ten minutes.

Sources: Decrypt — Bernstein power landlords note; ETF Trends — CoinShares miner outlook; TheEnergyMag — Q1 2026 miner shifts; Bitget — treasury liquidations; D-Central — Hashcenter migration. Related on Solana Garden: ETF outflows and AI rotation, AI data center memory squeeze, Bitcoin fundamentals, risk management.