Why AI Data Centers Will Keep Spending Billions Through 2025 – Is the Boom Sustainable?

AI data‑center spending is expected to continue booming through 2025. Microsoft alone announced an $80 billion investment this year to build infrastructure for training and deploying AI models.

Microsoft President and Vice Chair Brad Smith described AI as a "golden opportunity" for U.S. economic competitiveness, urging the nation to equip its workforce with AI skills.

Smith also emphasized the need for new alliances and massive infrastructure investment, confirming that over half of Microsoft's AI‑related spending will be in the United States.

Microsoft is not alone: Amazon reported plans to spend $75 billion on capital expenditures in 2024, with even higher spending projected for 2025, largely driven by generative AI workloads on AWS.

According to TrendForce, AI servers are expected to account for more than 70% of the total server market value by 2025, reaching roughly $298 billion. AI‑related server shipments are projected to grow 46% in 2024 and nearly 28% year‑over‑year in 2025, representing over 15% of the overall server market.

Omdia estimates that data‑center capital expenditures will rise over 40% in 2024, with servers comprising nearly half of that spend, valued at $2250‑$2300 billion. AI systems are forecast to make up 20% of server shipments but 73% of server capital spending, driven by high‑cost GPUs such as Nvidia’s, which can cost tens of thousands of dollars each.

Omdia also predicts Nvidia’s hardware will represent 43% of server capital expenditures in 2024, though its share may be hard to increase further in 2025.

Specialized AI cloud providers like CoreWeave are expected to increase spending, potentially surpassing Oracle this year.

Growth Can't Last Forever

The longevity of this trend depends on whether AI becomes a bubble. Critics question how much investment AI is actually consuming, noting limited tangible results so far.

Omdia reports that heavily regulated industries lack the frameworks to deploy AI at scale in 2024, with many companies still in pilot phases and planning industry‑specific AI model deployments for the coming year.

While GPU availability constrained some 2024 investments, supply is expected to improve in 2025.

Despite warnings that AI data‑center expansion could dramatically increase energy demand and greenhouse‑gas emissions—potentially tripling emissions by century’s end compared to a world without generative AI—financial firms like Morgan Stanley see the opportunity as once‑in‑a‑generation.

In summary, the AI data‑center boom appears set to continue in the near term, but its sustainability remains uncertain.