TeraWulf has signed a roughly $19 billion AI hosting agreement with Anthropic, and with it, a company that started life mining Bitcoin has effectively become an AI infrastructure landlord. The deal is the clearest data point yet in a sector-wide migration: crypto miners are converting their power contracts, land, and grid interconnects into hosting capacity for AI labs desperate for somewhere to plug in. Our take is that this is the trade of the year for miners with real energy assets, and it quietly repositions TeraWulf from a Bitcoin-price bet into a landlord for the AI boom.

  • The ~$19B agreement has TeraWulf hosting AI compute for Anthropic, a multi-year infrastructure commitment dwarfing its mining revenue.
  • It completes a pivot from Bitcoin mining to AI and high-performance-computing hosting.
  • The scarce assets are not chips but power, land, cooling, and grid interconnects, which miners already hold.
  • TeraWulf, ticker WULF, joins a wave of miners chasing AI hosting contracts to escape Bitcoin's boom-bust cycle.
Repurposing a mining site into AI hostingA Bitcoin miner already holds power contracts, land, cooling, and grid interconnects; TeraWulf converts these into AI compute hosting for Anthropic under a roughly 19 billion dollar deal.SAME ASSETS, NEW TENANTMiner already haspower contractsland + coolinggrid interconnectthe scarce partWas: Bitcoinrigs hashingNow: AI computeAnthropic tenant~$19B, multi-yeargenztech.blog
Fig 1 The rigs change; the power, land, and interconnect that took years to secure do not.

What did TeraWulf actually agree to?

TeraWulf committed to host AI compute for Anthropic under a multi-year deal valued around $19 billion, a figure that dwarfs anything its Bitcoin mining operation produces. In practice, TeraWulf provides the data-center shell, power, and cooling, and Anthropic runs its AI workloads inside. This is the same building-and-power business a miner already runs, redirected from hashing Bitcoin blocks to serving AI training and inference. The contract turns TeraWulf's revenue from something that swings with the Bitcoin price into something closer to a long-term real-estate-and-power lease.

RelatedBitcoin ETFs Snap a 10-Day Outflow Streak With $221M Day

Why are miners the ones landing these deals?

Because they own the bottleneck. The constraint on AI expansion in 2026 is not GPUs alone; it is finding a site with hundreds of megawatts of contracted power, a grid interconnect that took years to permit, land, and cooling. Bitcoin miners spent the last cycle assembling exactly those assets to run rigs cheaply. An AI lab that needs to deploy immediately cannot wait years for a new substation, so it rents a miner's ready-made footprint. The miners, in turn, get a tenant that pays in dollars on a long contract instead of volatile block rewards. It is a near-perfect swap of what each side lacks.

What does it mean for the stock?

The signal for investors is that a miner with a marquee AI hosting contract is no longer valued like a Bitcoin proxy. TeraWulf, trading as WULF, joins peers that have re-rated as the market began pricing them on contracted AI revenue rather than mining margins, a path other miners pivoting to AI and HPC hosting have walked. The read is not that Bitcoin exposure vanishes, but that the durable, dollar-denominated hosting cash flow becomes the anchor of the valuation. This is factual analysis, not investment advice: the thing to watch is contract execution and buildout timelines, because a signed deal and a powered-on data center are years apart.

Where is the risk?

Execution and concentration. Building gigawatt-scale AI capacity is capital-intensive and slow, and a $19 billion headline depends on TeraWulf actually delivering the power and facilities on schedule, often requiring significant financing. There is also customer concentration: leaning heavily on a single tenant means the company's fortunes track that tenant's demand and creditworthiness. And the broader miner-to-AI trade assumes AI compute demand keeps outrunning supply; if the buildout overshoots, the same power assets that look scarce today could look ordinary. The assets are real, but so is the construction risk between announcement and cash flow.

RelatedEmpery Digital Dumps Half Its Bitcoin for AI Compute

How big is the miner-to-AI trade?

Bigger than TeraWulf alone, which is what makes this deal a marker rather than a one-off. Across the sector, Bitcoin miners have spent the past year re-pitching themselves to Wall Street as AI and high-performance-computing hosts, and the ones with genuine, contracted power capacity have re-rated sharply as investors started valuing them on multi-year hosting revenue instead of hash rate. The logic is that the AI buildout is fundamentally a power problem, and miners spent the last cycle accumulating exactly the megawatts, land, and grid interconnects that now bottleneck everyone else. TeraWulf's Anthropic agreement is the largest, cleanest example of the thesis paying off, and it will pull more miners to attempt the same conversion, though not all of them hold assets a hyperscaler actually wants.

What to watch
  • Buildout milestones. Whether TeraWulf brings the contracted capacity online on the promised timeline.
  • Financing. How the multi-billion-dollar construction gets funded without diluting shareholders heavily.
  • Sector spread. Which other miners convert power assets into AI hosting next.

Our take

This deal is the moment the miner-to-AI pivot stops being a narrative and becomes a balance sheet. TeraWulf is no longer primarily a wager on Bitcoin's price; it is a landlord renting scarce power and interconnect to one of the most compute-hungry companies alive. That is a fundamentally more durable business, and it explains why the market rewards miners who land these contracts. The discipline required is patience: a $19 billion number is a promise about facilities that do not yet exist, and the value is realized in megawatts delivered, not press releases issued. For miners sitting on real energy assets, though, the strategic question of the year has an answer, and it is to become the AI boom's landlord before the window closes.

Primary sources

Original analysis by GenZTech. Reporting informed by TeraWulf.