The AI data-center boom is reshaping the industry as crypto miners, Anthropic, Cerebras, and hyperscalers fuel billions in infrastructure spending through 2030.
Projected spending of up to $7 trillion by 2030 has set off a wave of long-term lease deals and pulled in an unlikely cast of players, including bitcoin miners and a former cleaning-products company
The scramble to build the physical backbone of artificial intelligence has become one of the largest capital stories in technology, and it is reshaping who owns the industry's most valuable real estate.
Behind every chatbot query and cloud application sits a data center packed with servers and the power to run them. As AI demand climbs, the companies that build and own these facilities have moved from the background of the tech industry to its financial center.
The projected sums are large. Real estate firm JLL estimates that global data-center capacity could roughly double, from about 103 gigawatts today to around 200 gigawatts by 2030, an expansion it says may require as much as $3 trillion in new spending. McKinsey puts the figure higher, suggesting worldwide data-center outlays could approach $7 trillion by the end of the decade.
Those numbers are forecasts, not guarantees, and they hinge on AI demand holding up. The money already moving, though, is real.
The clearest sign of the boom is a run of multi-billion-dollar lease deals in which AI labs commit to a decade or two of capacity.
On July 6, TeraWulf, a former bitcoin miner, said it had signed a 20-year lease with Anthropic at its Justified Data campus in Hawesville, Kentucky. The agreement covers about 401 megawatts of computing capacity and is expected to generate roughly $19 billion in contracted revenue over its initial term, a figure larger than TeraWulf's own stock market value at the time. Initial capacity is due online in the second half of 2027, with the site fully operational by early 2028.
TeraWulf paired that announcement with the sale of its 50.1% stake in the Abernathy joint venture to an investor group led by its partner Fluidstack, monetizing an investment of about $450 million and freeing up cash for wholly owned projects.
The deal captures a broader shift. Companies that spent the last cycle mining cryptocurrency are repurposing their power contracts and sites to become, in effect, landlords for AI.
Hut 8 has followed a similar path. The energy-infrastructure company said it had fully commercialized its one-gigawatt Beacon Point campus in Nueces County, Texas, through a second 15-year lease worth $9.8 billion for 352 megawatts of capacity. The tenant, an investment-grade company that signed the first phase, has now doubled its contracted capacity at the site to 704 megawatts.
The rush has also drawn in companies with no history in the business.
CleanCore Solutions, which until recently sold aqueous-ozone cleaning products and had pursued a Dogecoin treasury strategy, has remade itself as a data-center developer. In June it named a new chief executive, Tyler Hassen, and turned toward AI infrastructure.
On July 23 the company said it had signed a 10-year colocation agreement with AI chipmaker Cerebras Systems for a campus in Minnesota. CleanCore values the deal at about $800 million over its initial term, with renewal options it says could lift the total past $3 billion. The site is designed for roughly 55 megawatts of utility power and 40 megawatts of critical IT load, and the company expects first revenue in early 2027 through a joint venture in which it holds about 80%.
The ambitions are large for a company of CleanCore's size. In its own regulatory filings, CleanCore points to its limited track record in data centers and its need to secure heavy financing as reasons for caution. About 20 megawatts of power at the Minnesota site is already energized, which the company says lowers some construction risk.
The spending stretches well past land and power. Each hyperscale facility needs servers, cooling systems, chips and networking hardware before a single workload runs.
That has spurred partnerships along the supply chain. AMD and Cerebras used the Advancing AI 2026 event to unveil a joint inference system that pairs AMD's Helios rackscale hardware with Cerebras's Wafer-Scale Engine, aiming at the ultra-low latency that demanding AI applications require.
Industry forecasters expect the hyperscale segment alone to grow from about $31.4 billion in 2026 to more than $52.5 billion by 2030. The wider AI-infrastructure market is projected to expand from roughly $75.9 billion to about $223.5 billion over the same span.
Much of the deal information above originated in company announcements, which present management's own projections and contract values. Long-dated lease figures assume tenants stay solvent and keep paying for 10 or 20 years, and construction timelines can slip. Anyone weighing these companies should treat headline contract values as potential revenue rather than money in hand, and consult a licensed financial professional. This article is informational and is not investment advice.
OpenAI pauses Astra after safety tests suggest the unreleased AI model may have reached a critical l...
SaferAI's report reveals China's GLM-5.2 rivals frontier AI models but lacks meaningful safety guard...
Apple signals that heavy Siri AI users may need iCloud+ to unlock higher usage limits. Here's what T...
Discussion
Join the discussion and share your perspective.