Anthropic has struck a roughly $9 billion agreement with cloud computing firm Riot Platforms to secure the massive AI compute capacity its Claude models demand, according to reporting from Bloomberg and StreetInsider published on August 11, 2026. The deal, tied to a long-term data center lease disclosed alongside Riot's second-quarter results, marks one of the largest infrastructure commitments yet by an AI lab and highlights a fast-emerging trend: former cryptocurrency miners retooling their facilities to power artificial intelligence.

The agreement is the latest sign that the AI industry coverage of late 2026 is increasingly a story about concrete, steel, and electricity as much as algorithms. As frontier models grow larger, the companies building them are locking in decades-long contracts for power and cooling.

The terms, as reported

While the companies have not released a single unified press statement, financial outlets have painted a consistent picture of the arrangement:

  • Value: Bloomberg and StreetInsider report the Anthropic-Riot agreement at roughly $9 billion. Investing.com characterized the lease Riot secured at approximately $9.1 billion.
  • Duration: StreetInsider reported it as a 20-year commitment.
  • Scale: The deal covers approximately 191 megawatts of AI data center capacity, according to StreetInsider.
  • Timeline: Stock Titan reported that the lease is expected to begin delivering capacity with 96 megawatts coming online in December 2027.

Riot announced the lease as part of its Q2 2026 financial results, where the company also beat analyst expectations on revenue. Shares jumped on the results and the landmark deal, as reported by Benzinga and Seeking Alpha.

Why a Bitcoin miner is hosting AI workloads

Riot Platforms is best known as one of the largest publicly traded Bitcoin mining companies in North America. But miners like Riot possess exactly what AI companies need: large facilities already wired for enormous power consumption, access to cheap electricity, and the cooling infrastructure required to run hardware around the clock.

As cryptocurrency mining margins have come under pressure and AI demand has exploded, a growing number of miners have begun converting portions of their sites into AI data centers. AI hosting typically commands far higher revenue per megawatt than crypto mining, making the pivot economically attractive — provided the facilities can meet the stricter reliability and latency standards that enterprise AI customers demand.

For Anthropic, the appeal is straightforward. Training and running frontier models requires staggering amounts of computing power, and the supply of advanced AI chips and the facilities to house them is tight. Locking in nearly 200 megawatts of dedicated capacity for two decades gives the company a measure of certainty in a market where compute has become a strategic bottleneck.

A defining theme of 2026: compute as a strategic asset

The Anthropic-Riot deal fits a broader pattern that has dominated this year. Across the industry, AI labs, cloud providers, and financial institutions are pouring capital into infrastructure. Just a day earlier, Nvidia announced a partnership with Wall Street giants including Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR to establish AI compute financing platforms aimed at mobilizing over $500 billion in third-party capital.

OpenAI, meanwhile, has been pursuing its own multibillion-dollar data center build-out, including reported plans in Ohio. The common thread is that frontier AI development is now gated as much by access to power and facilities as by research breakthroughs.

These infrastructure deals also carry political weight. Energy-hungry data centers have drawn scrutiny from lawmakers and communities concerned about grid strain and environmental impact, and companies have begun framing their build-outs in terms of national competitiveness and job creation.

What it means for Anthropic

For Anthropic specifically, securing long-term compute reinforces its position as a serious challenger to OpenAI and Google in the frontier-model race. The company has been on a fundraising and expansion tear, and tying up dedicated capacity reduces its dependence on a single cloud provider for the GPUs that underpin Claude.

It also signals confidence: a 20-year commitment implies Anthropic expects AI compute demand to remain elevated for the foreseeable future, not just through a short-term boom.

The road ahead

The deal's impact will hinge on execution. Capacity is not expected to begin arriving until late 2027, meaning the compute crunch Anthropic faces today will not be directly relieved by this agreement for more than a year. In the interim, the company will continue to rely on existing cloud partners.

Still, the size and duration of the commitment send a clear message. In the current AI business landscape, whoever controls the compute controls the pace of progress — and Anthropic is paying handsomely to make sure it controls a meaningful share of its own.

For ongoing reporting on the infrastructure race reshaping artificial intelligence, keep up with our breaking AI news.

Stay Ahead of AI

The compute land grab is reshaping who can build frontier models. Follow our full coverage of model releases, funding rounds, and infrastructure deals, and read more AI news →