Big tech companies are spending unprecedented sums on AI data centers, but construction delays, power grid constraints, and regulatory hurdles are preventing them from turning capital into functioning facilities. A JPMorgan analysis found that more than 60 percent of data center capacity planned for completion in 2027 is not yet under construction, with another 7 percent already delayed. Microsoft, Alphabet, Meta Platforms, and Amazon spent $410 billion on capital expenditure last year and are expected to spend more than $670 billion this year, yet the physical infrastructure is struggling to keep pace with the investment.

What's Slowing Down the AI Data Center Boom?

The bottleneck isn't money. Hyperscalers can raise billions in capital, but they're running into the unglamorous realities of construction: supply chain delays, permit approvals, and a power grid that's already strained. The most critical constraint is securing grid connections. Some of these facilities consume as much electricity as a midsize city, which creates friction with grid operators and power companies already managing peak demand during extreme weather.

University of Texas at Austin energy expert Josh Rhodes explained the regulatory paralysis: "Because of how much uncertainty there is about how many data centres are real, about how much load is going to be connected, it has kind of paralysed a lot of the processes," he said. Grid operators don't know which projects will actually be built, making it difficult to plan infrastructure upgrades and approve new connections.

Beyond grid approval delays, JPMorgan identified additional supply chain constraints affecting timelines. Delays in gas turbines and electrical transformers are pushing data center projects further behind schedule, compounding the construction bottleneck.

How Are Tech Giants Solving the Power Problem?

Rather than waiting for grid operators to approve connections, major tech companies are taking control of their own power supply. This strategy is becoming a competitive advantage in the race to build AI infrastructure:

• Securing dedicated power sources: Google became the only tech giant to own a power company after buying Intersect for $4.75 billion, a wind and solar developer with projects under development to supply multiple gigawatts of electricity.
• On-site generation: xAI, OpenAI, and Meta have built or proposed data centers powered by on-site gas generation, though this approach has drawn criticism from residents and policymakers concerned about air pollution.
• Nuclear power partnerships: Microsoft struck a 2024 deal with Constellation Energy to restart the undamaged reactor at Three Mile Island, and federal regulators cleared part of the plan this week.
• Demand response programs: Google announced a three-year deal with demand response company Voltus to create more capacity in PJM, the US's largest power market, with potential to create up to 100 megawatts of capacity, roughly the size of a small power plant.

Regulators are now considering whether data centers built beside new power sources should receive faster grid connections, effectively rewarding companies that solve their own power problems.

Why Is Google Raising $80 Billion in Equity?

Alphabet's decision to raise $80 billion in equity capital, including a $10 billion investment from Berkshire Hathaway, signals the intensity of upcoming capital needs. The company had already revised its 2026 capital expenditure forecast to between $180 billion and $190 billion in April, but the equity raise suggests even those projections may be conservative. Alphabet shares fell 3.9 percent on the announcement, and the company lost $340 billion in market value over three trading sessions, its biggest three-day loss on record.

"The fact that they had to raise equity really makes you wonder about the intensity of the capex needs over the next couple of years," said Michael Nathanson, analyst at MoffettNathanson.

Michael Nathanson, Analyst at MoffettNathanson

The equity raise is notable because Alphabet typically issues debt to fund capital projects. The shift to equity financing suggests the company is concerned about debt levels or wants to preserve financial flexibility for an extended period of heavy spending.

How Are Tech Companies Adapting to Grid Constraints?

Google has spent years testing how data centers can reduce power consumption when the grid is strained. The company now operates utility pilot programs that pay it to reduce demand during peak periods, turning data centers into flexible loads that help stabilize the grid rather than destabilize it. This approach creates a win-win: Google reduces its power costs while grid operators gain capacity without building new power plants.

The broader trend reflects a fundamental shift in how hyperscalers approach infrastructure. Rather than waiting for grid operators to accommodate their needs, tech companies are investing in power generation, demand response systems, and load-shifting technologies that make them partners in grid stability rather than just consumers of electricity.

What Does This Mean for the AI Infrastructure Race?

The delays and constraints are reshaping competitive dynamics. Companies that secure their own power sources and solve grid connection problems faster will be able to deploy AI infrastructure ahead of rivals. Google's investments in power generation and demand response, combined with its utility partnerships, position it to connect data centers to the grid faster than competitors facing regulatory delays.

Meanwhile, the broader AI infrastructure market continues to expand. SoftBank announced plans to invest 75 billion euros, approximately $87 billion, to build AI infrastructure in France, including 5 gigawatts of AI data center capacity, marking the company's largest AI infrastructure investment in Europe. Three sites in the northern Hauts-de-France region, including one in Dunkirk, are expected to come online by 2031, with SoftBank partnering with French engineering firm Schneider Electric and state-owned nuclear energy giant EDF.

The fundamental challenge remains: hyperscalers can raise mountains of cash, but they cannot always turn it into buildings with blinking lights as quickly as Wall Street expects. Until supply chains improve, permits accelerate, and grid operators approve new connections, the gap between planned capacity and actual construction will continue to widen.