Microsoft plans to invest approximately $80 billion in the infrastructure for artificial intelligence in fiscal year 2025, with more than half targeted for the U.S. market. The massive investment, primarily in AI-enabled data centers, comes at a pivotal moment, as major tech companies struggle to balance aggressive expansion with uncertain revenue timelines and power constraints.
It also comes as major tech companies collectively spent more than $200 billion on data centers last year, according to Bloomberg.
“Not since the invention of electricity has the United States had the opportunity it has today to harness new technology to invigorate the nation’s economy,” wrote Microsoft president Brad Smith in a blog post announcing the investment. “Artificial intelligence is the electricity of our age.”
In a quarterly earnings call last fall, Microsoft reported delays in building data center infrastructure. “This demand all showed up pretty fast,” CEO Satya Nadella told investors, adding that data centers “don’t get built overnight.”
Investment numbers are also rising quickly across Microsoft’s peers and competitors. Last fall, Amazon said it expected capital spending to reach $75 billion for 2024 — most of it going to data centers. Meta’s capital spending last year was expected to reach $40 billion. And Alphabet told investors to expect “substantial” increases this year.
“I think pretty much everyone today has less capacity than they have demand for,” Amazon CEO Andy Jassy said on an investor call last October. “The faster we grow, the faster we have to invest capital in data centers, and networking gear and hardware.”
The scale of these commitments reflects dramatic shifts in data center economics. “The cost of a data center today, in terms of the full stack CapEx deployment all the way through GPUs, is about $25 million a megawatt, or $25 billion a gigawatt,” explained Brian Janous, Microsoft’s former vice president of energy on Latitude Media’s Catalyst podcast. “So if you go to a utility and say, ‘hey, I want a gigawatt of power,’ that’s going to be a billion dollars of infrastructure.”
Microsoft, a member of the $30-billion Global AI Infrastructure fund, did not respond to a request for comment on how much it plans to invest in energy infrastructure to serve its data center build-out.
An evolution in strategy
The rapid increase in capital spending on data centers illustrates a shift in how tech companies view infrastructure investment.
Janous told Latitude Media there are two competing approaches borrowed from manufacturing: lean manufacturing, which minimizes inventory and accepts longer lead times, and the “theory of constraints,” which focuses on maximizing throughput by building extra capacity at potential bottlenecks. Tech giants are embracing the latter.
“If I am downstream in the system — so I’m data center operators, I’m cloud providers, I’m Nvidia, I’m OpenAI — I’m very much thinking about the world in terms of theory of constraints,” Janous explained. This helps explain why companies like Microsoft are willing to commit such enormous sums — even if the returns from AI are not yet clear.
The announcement positions Microsoft at the forefront of what Smith described as a “three-part vision for America’s technology success.” This includes advancing world-leading AI technology and infrastructure, championing AI skilling programs, and exporting American AI to allies.
However, executing this vision faces significant challenges. Just this week, for instance, the company paused construction on portions of its $3.3 billion data center campus in Wisconsin to “evaluate scope and recent changes in technology.”
The other fundamental challenge is power availability. Microsoft’s recent $16-billion deal to revive a reactor at Three Mile Island is just one example of the lengths companies will go to secure reliable power for AI operations. In the months since, a number of tech companies have also unveiled plans to invest in nuclear fusion, including newer, advanced technologies.
But if Microsoft’s $80 billion investment is any indication, the company seems confident that the industry’s many hurdles can be overcome.
“For any one individual player, the ability to commit billions of dollars to electric utilities to build out more infrastructure is a hard pill to swallow if you’re not fully convinced that you have a customer on the other side,” said Janous.
The question now is whether Microsoft can secure both the power and AI revenue to justify such historic levels of infrastructure spending.
Some analysts, though, are optimistic. In a note to investors, for instance, JPMorgan’s team wrote that Microsoft’s investments are “planting the longer-term seeds for success,” according to Bloomberg.
