A new wave of capital is pouring into artificial intelligence infrastructure this year. Global spending on AI data centers, specialized chips, and networking hardware has already exceeded $150 billion in 2025. That figure marks a 40 percent increase over the same period in 2024. The numbers come from a consolidated market report released this week by the International Data Corporation.
nvidia and microsoft dominate the spending surge
Nvidia and Microsoft account for nearly half of all AI infrastructure investments this year. Nvidia has secured contracts to supply its H200 and B200 GPUs to cloud providers and enterprise clients worth more than $45 billion. Microsoft has committed $35 billion to expand its Azure AI regions across North America, Europe, and Asia. The two companies are building what amounts to a global fabric of high performance computing clusters designed specifically for large language model training and inference workloads.
The pace of construction has accelerated sharply. Microsoft alone has broken ground on 12 new data center campuses in the first half of 2025. Each campus is designed to support liquid cooled racks of Nvidia GPUs and custom Azure Maia accelerators. Nvidia has also announced partnerships with Equinix and Digital Realty to co-locate its DGX SuperPOD systems in 20 additional locations worldwide.
hyperscalers race to secure energy and supply chains
The scale of these projects has created new bottlenecks. Power availability has become the primary constraint for new data center builds. Both Nvidia and Microsoft have signed long term power purchase agreements with renewable energy developers. Microsoft has secured 8 gigawatts of solar and wind capacity across Texas, Ireland, and Singapore. Nvidia has allocated $2 billion to direct investments in small modular nuclear reactor startups to guarantee stable baseload power for its next generation clusters.
Supply chain logistics have also forced creative solutions. The typical lead time for a high end GPU server rack has stretched to 48 weeks. Nvidia has responded by establishing its own logistics hubs in Malaysia and Mexico. These hubs pre configure and test complete server racks before shipping them to data center sites. The approach reduces on site installation time by 60 percent according to company documentation provided during a recent investor briefing.
The investment wave is reshaping how AI companies think about compute. Several large AI startups including Anthropic and Cohere have shifted from renting cloud instances to signing multi year leases for dedicated clusters. Those deals are valued at $500 million to $2 billion each. The shift indicates that the market is moving toward a model where compute capacity is treated as a long term strategic asset rather than an on demand utility.
Analysts expect the total addressable market for AI infrastructure to exceed $400 billion by 2027. The projections cover GPU sales, networking gear, cooling systems, and data center construction. Nvidia and Microsoft are positioned to capture the largest share of that growth. Smaller players such as AMD, Intel, and Google are also increasing their capital expenditure plans. The race to build the underlying physical layer of the AI economy shows no signs of slowing down. For more on how this infrastructure arms race is affecting smaller startups and investors, read our analysis {$link_text}.
