Anthropic secures 3.5 gigawatts of AI compute capacity as Bitcoin miners lease to host data centers

The artificial intelligence industry has crossed an energy threshold that is reshaping the economics of Bitcoin mining. On April 6, Anthropic, the company behind the Claude model, announced an agreement to secure 3.5 gigawatts of next-generation Google TPU compute capacity manufactured by Broadcom. The contract is described as the largest infrastructure deployment in Anthropic’s history.

At the same time, large Bitcoin miners are increasingly shifting away from mining and toward hosting and AI-related infrastructure. Core Scientific, one of the world’s largest data center operators for mining, is preparing to liquidate practically all of its Bitcoin reserves during this year to fund a major conversion of capacity toward artificial intelligence.

Key deals and capacity shifts

Core Scientific’s planned conversion involves 1.2 gigawatts of capacity toward hosting hardware for artificial intelligence. Hut 8, meanwhile, secured a 15-year lease contract valued at 7 billion dollars, with Anthropic as the main tenant and Google as the financial backer.

The scale of Anthropic’s deal is presented in energy terms: one gigawatt of electrical consumption is described as roughly equivalent to the demand of one million U.S. households. Anthropic has reserved energy equivalent to 3.5 million homes to train and serve language models.

Broadcom confirmed in an SEC filing that most of the new capacity will be located in the United States and will begin operations starting in 2027. The allocation is said to add to an additional gigawatt of capacity Anthropic already receives from Google during 2026.

Financial momentum behind the AI bet

Anthropic’s revenue growth supports the scale of the commitment. The company’s annualized revenue is described as having crossed 30 billion dollars, more than triple the 9 billion dollars reported at last year’s close.

Customer demand for Claude is also highlighted: the number of corporate customers spending over one million dollars annually on Claude doubled in under two months, rising from 500 to more than 1,000 companies.

Why miners are pivoting

The article frames the shift as a move from volatile mining rewards to more predictable, contract-backed cash flows. It states that public miners currently lose close to 19,000 dollars for every Bitcoin they produce, with production costs around 80,000 dollars per unit while the market price is near 68,000 dollars—down nearly 47% from the all-time high set in October.

It also compares the capital intensity of AI hosting versus mining. Preparing a megawatt for high-performance computing workloads is described as requiring between 8 million and 15 million dollars in capital expenditures, compared with 700,000 to 1 million dollars for a Bitcoin mining facility.

Revenue mix and deal volume

Despite the higher entry cost, the article says public mining companies are embracing the shift because AI hosting can provide stable, long-term revenue. It cites TeraWulf as having secured 12.8 billion dollars in contracted high-performance computing hosting revenue.

According to CoinShares analysis referenced in the article, publicly traded mining firms could derive up to 70% of their total revenue from AI hosting by the end of this year. For companies that have already closed binding agreements, mining revenue is described as falling from 85% of total revenue to less than 20%.

The sector is also described as having announced more than 70 billion dollars in cumulative deals related to AI and high-performance computing.

Grid constraints and the timing of capacity

The article links the deals to structural electricity constraints. PJM Interconnection, described as the nation’s largest grid operator, projects a 6-gigawatt shortfall by 2027. It also states that U.S. data center electricity demand is expected to rise from under 15 gigawatts today to 134.4 gigawatts by 2030—an increase of nearly nine times in seven years.

It further notes that five AI data centers are expected to reach 1 gigawatt of capacity this year. Up to 11 gigawatts of announced capacity for 2026 is described as not yet breaking ground due to bottlenecks in transformer and grid equipment supply.

In this context, the article characterizes Anthropic’s 3.5 gigawatts as a major anchor on the system.

Potential effects on Bitcoin

The article describes consequences for the Bitcoin ecosystem on two fronts: market pressure and network security.

First, it says the liquidation of reserves by miners such as Core Scientific to fund AI conversions adds direct selling pressure to a market already described as weak.

Second, it points to changes in network metrics. It states that mining difficulty—an automatic adjustment mechanism reflecting active hash on the network—registered a drop of 7.76%. It suggests that as more operators redirect gigawatts away from mining and toward AI hosting, the network strength metric could contract further in the short term.

Hut 8’s 15-year deal, with Google’s financial backing, is presented as evidence that long-term lease agreements can shift miners’ balance sheets toward fixed-income-like cash flows.

What to watch next

The article highlights a timeline for monitoring the transition. Anthropic’s new TPU capacity is described as coming online in 2027. The first data hall of Hut 8’s River Bend complex is described as opening in the second quarter of 2027. Core Scientific’s conversion of 1.2 gigawatts is described as accelerating throughout 2026.

It concludes that the key question is how much additional Bitcoin will flow into spot markets during the pivot and how quickly network difficulty adjusts as computing power shifts from mining to AI hosting.