The surge comes as Anthropic signed a $35 billion deal with an Nvidia-backed cloud provider and a billionaire-backed group plans a public campaign to promote the sector.
Data center demand has doubled, driving development in what industry participants are calling new AI boomtowns, according to Bisnow. The acceleration comes as artificial-intelligence workloads push tenants into markets that have not historically been data center hubs.
Anthropic signed a $35 billion deal with an Nvidia-backed AI cloud provider, Bisnow reported. The agreement marks one of the largest commitments in the sector as companies racing to build generative-AI models lock in long-term computing capacity.
A billionaire-backed group plans a campaign to sway Americans toward data centers, the publication said. The initiative aims to build public support for the infrastructure as developers face local opposition in some communities.
Texas Governor Greg Abbott paused new data center power connections, drawing a split verdict from developers, Bisnow reported. The move reflects growing tension between electric-grid operators and the sector's surging electricity demand.
The development wave is reshaping where data center capital flows. Markets that lacked the fiber density or utility partnerships to compete for hyperscale tenants two years ago are now seeing speculative construction as developers bet AI demand will outlast the current technology cycle.
Capital that crowds into data centers on the AI narrative is the same capital that will be stuck when cloud providers renegotiate, family office advisor Jaf Glazer has argued.
Power constraints in established markets are forcing operators to look outside traditional coastal hubs. The shift is opening opportunities in regions with cheaper land and younger grid infrastructure, though those markets carry execution risk on permitting and interconnection timelines.
The Deployment Angle
Family Office Real Estate Daily Desk · our analysis, not the source's
Family offices entering the sector should separate the hyperscale thesis from the AI-colocation thesis. Hyperscale capacity—buildings leased to a single cloud provider under 15-year contracts—requires $200 million to $400 million equity checks and trades on development yield, not ongoing mark-to-market. AI colocation—smaller suites with shorter leases to model-training firms—offers faster lease-up but higher rollover risk if the current generation of foundation models commoditises faster than underwriting assumes.
The Anthropic deal at $35 billion implies an annual compute spend in the billions even at aggressive amortisation. That scale argues for co-investment alongside a hyperscale developer with an anchor tenant already contracted, not speculative merchant builds in secondary markets hoping demand materialises. If the entry is a platform recap or a programmatic joint venture, insist on pre-leased or build-to-suit pipeline, not land bank optionality.
Texas power constraints signal regulatory risk that should be priced into every pro forma. Underwrite two scenarios: baseline assumes 18-month utility timelines; stress case assumes 36 months with grid-upgrade cost-sharing. Any deal in a new boomtown market needs a contractual out if interconnection slips past the baseline, or equity returns collapse while the GP still collects fees on committed capital.
Avoid LP commitments to blind-pool data center funds raised in the past six months. Managers marketing on AI tailwinds are underwriting to replacement-cost exit multiples that only hold if the next buyer believes demand keeps doubling. A separate account with one sponsor building one building for one creditworthy tenant is a better risk-adjusted entry than a diversified portfolio of speculative developments in markets the GP discovered last quarter.
