Anthropic signs $19 billion, 20‑year AI data‑center lease with TeraWulf

Anthropic, the leading large‑language‑model developer, has entered a 20‑year lease agreement with data‑center specialist TeraWulf worth $19 billion, according to multiple industry reports. The deal, announced in early July 2026, will convert TeraWulf’s existing power‑rich facilities into a dedicated AI‑compute hub, marking a decisive shift for the former crypto‑miner into the mainstream artificial‑intelligence infrastructure market.

Deal details

SiliconANGLE reported that the lease totals $19 billion and spans two decades, a scale that rivals the biggest cloud‑provider contracts disclosed in recent years. Constellation Research confirmed the 20‑year term and added that the agreement covers multiple data‑center sites across the United States, each equipped with high‑density power delivery to meet Anthropic’s GPU‑heavy workloads. Yahoo Finance echoed both the monetary value and the duration, describing the arrangement as a “massive AI data‑center lease” that will anchor Anthropic’s next generation of models.

TeraWulf, listed on NASDAQ under the ticker WULF, will receive a steady, long‑term revenue stream that offsets the volatility of its legacy cryptocurrency mining business. Stock Titan noted that the lease coincides with a $530 million sale of the company’s Abernathy assets, suggesting a broader strategic pivot toward AI‑focused infrastructure. The same source highlighted that the lease will likely increase the firm’s cash flow stability and improve its balance sheet, although exact financial projections were not disclosed.

Shares of TeraWulf surged following the announcement, with Yahoo Finance reporting a notable price jump on the day of the news. While the precise percentage rise was not detailed in the available sources, the market reaction underscores investor optimism about the company’s transition from mining to AI services.

Why it matters

The agreement arrives at a moment when demand for AI compute is outpacing supply, and major cloud providers are scrambling to expand capacity. By locking in a $19 billion commitment, Anthropic secures a dedicated, power‑intensive campus that can host the massive GPU clusters required for training and inference of next‑generation language models. This reduces reliance on shared cloud resources, potentially lowering latency and cost for the company’s customers.

For TeraWulf, the lease represents a strategic re‑orientation. The firm originally built its infrastructure to support proof‑of‑work cryptocurrency mining, which requires abundant, cheap electricity—an asset that is equally valuable for AI workloads. Converting existing sites into AI‑grade facilities allows TeraWulf to leverage its real‑estate and power contracts without the need for massive new construction. Analysts cited by Constellation Research argue that this could set a precedent for other crypto‑mining operators seeking relevance as the sector contracts under tighter environmental regulations.

On a macro level, the deal illustrates the growing convergence of the AI and energy sectors. As AI models become more compute‑hungry, the industry is turning to providers that can guarantee both electrical capacity and cooling efficiency. The $19 billion figure, highlighted by multiple outlets, signals that AI developers are prepared to invest heavily in dedicated infrastructure rather than relying solely on public cloud offerings.

Reactions and viewpoints

Industry observers have praised the partnership for its potential to accelerate Anthropic’s roadmap. A senior analyst quoted by SiliconANGLE said the lease “provides Anthropic with the predictability needed to plan multi‑year research cycles without the risk of cloud capacity constraints.”

Conversely, some critics caution that the long‑term nature of the deal could lock Anthropic into fixed pricing at a time when hardware costs are rapidly falling. Constellation Research noted that “if GPU prices continue to decline, Anthropic may end up paying a premium compared with a flexible cloud‑based approach.” The analyst did not quantify the risk, but the comment reflects a broader debate about capital‑intensive versus operational‑expenditure models for AI compute.

Investors reacted positively to TeraWulf’s pivot. Stock Titan reported that the $530 million Abernathy sale, combined with the Anthropic lease, “positions TeraWulf as a serious contender in the AI‑infrastructure arena.” The article suggested that the company could become an attractive acquisition target for larger data‑center operators seeking to expand their AI footprint.

What’s next

Implementation of the lease will begin with the retrofitting of TeraWulf’s existing campuses to meet Anthropic’s specifications for power density, cooling, and network latency. Both companies have indicated that the first phase of the project could be operational by early 2027, though exact timelines were not disclosed.

Anthropic is expected to integrate the new capacity into its training pipeline for upcoming model releases, potentially accelerating the rollout of more capable language models. The firm has not announced any specific product milestones tied to the new data‑center resources.

For TeraWulf, the lease is just the first of what management hopes will be a series of AI‑focused agreements. The company’s leadership, referenced in the Yahoo Finance coverage, has signaled an intent to market its power‑rich sites to other AI developers, leveraging the Anthropic deal as a proof‑of‑concept.

Regulators and environmental groups will likely monitor the partnership closely, given the high electricity consumption associated with AI training. The companies have not released details on the energy mix or sustainability commitments tied to the lease, leaving an open question about how the new compute will align with broader decarbonization goals.

In the broader competitive landscape, the Anthropic‑TeraWulf agreement adds another layer to the rivalry among cloud giants, specialized AI infrastructure firms, and emerging players. As AI applications become more pervasive, the race to secure reliable, low‑cost power for compute will intensify, and the $19 billion lease could be a bellwether for future deals of similar magnitude.