Summary: The Wall Street Journal reported on May 1, 2026 that Coatue formed Next Frontier to buy land near major power sources and develop it into data centers. The vehicle has already formed a joint venture with Fluidstack, whose prior Anthropic contract was reported at $50 billion. The article warns that interconnection delays, demand shifts, thin valuation comparables, and local opposition could undermine the land bet.
The Wall Street Journal broke the news on May 1, 2026: Coatue, the firm that runs both hedge fund and venture capital operations, has stood up a new vehicle named Next Frontier. Its strategy fits on a single line — acquire parcels sitting close to major power sources and turn them into data centers. Coatue hardly needed a fresh way into AI; Anthropic, OpenAI, xAI, CoreWeave, and Singapore’s DayOne were already among its heavy positions. What is new is the asset class: rather than another equity stake, the firm is chasing the inputs the AI build-out genuinely runs short of — land and power.
The timing has its logic. Pew Research’s April tally puts the United States at roughly 3,000 data centers today, with upwards of 1,500 more somewhere in the construction pipeline, and the bulk of the new sites are landing in rural areas. Compute demand is simply outrunning how fast the grid can expand and how fast qualified land can come to market.
Inside the Next Frontier Playbook
As the Journal tells it, the model could not be simpler: identify land beside power infrastructure, buy it, and let it mature into a data center. Scarcity does the rest — parcels that can check every box (electricity, water, fiber, a workable grid interconnection) exist in finite supply, so whoever holds them negotiates from the top of the market. A hedge fund, for its part, can act faster than a hyperscaler’s procurement machinery, and once a site is qualified it can cash in that optionality through leases, joint ventures, or an outright sale. The move also lifts Coatue’s AI exposure up a level, from passive shareholding to active control of a critical input, and raises the ceiling on returns accordingly. TechCrunch reached out for comment; Coatue declined.
Fluidstack, the JV, and the Line Back to Anthropic
The most tangible step in the report: Next Frontier has already put its name to a joint venture with Fluidstack, a cloud infrastructure startup. Fluidstack needs little introduction — the company previously signed a $50 billion contract to construct data centers for Anthropic, and in mid-April it was reported to be negotiating a $1 billion round at an $18 billion valuation, a few months on from the $7.5 billion set in its previous round.
Follow that chain and the “possibly for Anthropic” headline acquires its footing: the joint venture partner’s anchor customer is Anthropic, and Anthropic itself is reported to be negotiating a raise at roughly a $900 billion valuation that could be sealed within two weeks (we dug into those negotiations earlier). Should Next Frontier’s parcels end up feeding Fluidstack’s Anthropic projects, Coatue would be carrying exposure at three layers of the stack at once — the model company, the operator, and the ground beneath them.
Power-Adjacent Land, the Newest Speculative Trade
Coatue has plenty of company in reading the map this way. Blackstone is part of the financing behind Related Digital’s $16 billion Oracle data center project in Saline Township, Michigan, and Shark Tank investor Kevin O’Leary is shopping a data center land financing scheme of his own. The rush began earlier still: in January, Nvidia put $2 billion into CoreWeave — an operator carrying heavy debt — so it could add 5 gigawatts of AI compute.
Chipmakers, private equity, venture funds, and celebrity capital all piling into land-plus-power sends an unmistakable signal: the bottleneck has migrated upstream to siting and interconnection. Whoever controls qualified land sets the pace at which the next tranche of compute actually switches on. For data center developers the read-through is blunt — three years out, the cheapest compute may belong to whoever secured qualified sites back in 2026, rather than to whoever placed the largest GPU orders.
What Could Sink the Land Trade
This is not riskless arbitrage. Interconnection and permitting run on year-long clocks, and carrying costs gnaw at returns the entire time. Should AI demand growth cool, or model efficiency gains arrive sooner than forecast, land bought at frenzy prices ends up stranded. Pricing, meanwhile, still leans on a short list of comparable deals, so valuations stand on thin footing. Politics weighs in as well: rural communities are already balancing the jobs and tax revenue data centers deliver against grid, water, and land-use impacts, and enough local resistance can stall even a fully qualified site indefinitely. Coatue’s thesis only pays off if the compute shortage outlasts the development timeline of these parcels — precisely where consensus sits today, and precisely the reason to keep questioning it.