- A JPMorgan analysis found that more than 60% of data center capacity planned for completion in 2027 is not yet under construction, with another 7% already delayed — even as the four hyperscalers collectively spent $410B on capex last year and plan $670B+ this year
- The primary bottleneck is grid connection: power companies and grid operators require lengthy reviews before connecting facilities that can consume as much electricity as a mid-size city; supply-chain delays in gas turbines and electrical transformers compound the problem
- Google’s edge: it acquired power company Intersect Power for $4.75B earlier this year — the only hyperscaler to own a power developer — giving it direct control over gigawatts of wind and solar generation dedicated to data center supply
- Google also launched a demand-response partnership with Voltus in PJM (the nation’s largest power market) to create up to 100 megawatts of grid relief capacity — a tactic that positions it for faster grid connection approvals in regions offering fast-track treatment to data centers that reduce consumption during peak demand
What Happened?
Even as Alphabet raised $80 billion in equity specifically to fund AI infrastructure, a new JPMorgan analysis reveals the uncomfortable reality: money is not the binding constraint. More than 60% of data center capacity scheduled to be complete by 2027 hasn’t broken ground, and another 7% is already delayed. The culprits are structural — grid connection queues that can take years, permitting battles, and supply-chain backlogs for critical hardware like gas turbines and electrical transformers. Google’s response has been to build a vertically integrated power strategy that no competitor has matched: its $4.75B acquisition of Intersect Power gives it direct ownership of generation assets, bypassing the grid interconnection queue that strangles competitors. Its demand response programs, now formalized through a three-year Voltus partnership in PJM, create an additional lever for grid access.
Why It Matters?
The data center race is not about who raises the most money — it’s about who can actually build. Google’s power strategy gives it a structural advantage that takes years to replicate: owning a power developer means Intersect’s projects can be sited specifically to co-locate with Google’s planned data centers, then connected to the grid via the faster pathway many regulators are offering to facilities that generate their own power. Alphabet’s $340 billion drop in market cap over three trading sessions following the $80B raise announcement reflects the market’s discomfort with the pace and scale of capex — MoffettNathanson analyst Michael Nathanson said the equity raise raises questions about “the intensity of the capex needs over the next couple of years.” But the paradox is that competitors who raise less capital but can’t deploy it are no better positioned. The bottleneck is construction capacity and power access, not financial capital.
What’s Next?
Watch for regulatory movement on two fronts: grid operators’ fast-track policies for self-generating data centers are still being developed and vary by region; federal action to streamline permitting would be transformative but remains politically difficult. The nuclear pathway is the other major variable: Microsoft’s Three Mile Island restart just received partial federal greenlight, and multiple hyperscalers are backing small modular reactor developers. If SMR technology proves viable at commercial scale, it would represent a step-change in hyperscalers’ ability to self-supply reliable baseload power — and reopen the construction calculus entirely. In the near term, Google’s Intersect acquisition and Voltus partnership give it a 12-18 month head start on deploying its $80B raise into actual steel-in-ground data center capacity.
Source: The Wall Street Journal
