Tech giants battle in data centers. Microsoft plans to invest $80 billion in fiscal year 2025.

$Microsoft(MSFT.US) $NVIDIA(NVDA.US)

Not only large models, but the entire AI industry is an extremely capital-intensive game.
On January 3, 2025, Eastern Time, Microsoft Vice Chairman and President Brad Smith announced in a blog post on the company's official website: "In fiscal year 2025, Microsoft is expected to invest approximately $80 billion in building AI data centers to train AI models and deploy AI and cloud-based applications globally. More than half of the total investment will be in the United States."
Microsoft's fiscal year 2025 refers to the period from July 1, 2024, to June 30, 2025. This means that Microsoft will invest an additional $40 billion in this field in the next six months.
At the end of June 2024, Amazon announced that it would invest $100 billion in data centers over the next 10 years. Compared to Microsoft's plan, this now seems too conservative.
According to financial reports, Microsoft's capital expenditure in the first quarter of fiscal year 2025 (i.e., the third quarter of the calendar year 2024) increased by 5.3% year-on-year to $20 billion. These expenditures were mainly used to support the company's cloud and AI products, including infrastructure investments such as leases and real estate.
During the earnings call on October 30, 2024, Microsoft CFO Amy Hood stated that capital expenditure in the second quarter of fiscal year 2025 (i.e., the fourth quarter of 2024) would increase sequentially.
According to data from financial platform Visible Alpha, Wall Street analysts expect Microsoft's capital expenditure in fiscal year 2025 to reach $84.24 billion.
Brad Smith's announcement of an $80 billion investment in data centers within a year implies that 95% of Microsoft's fiscal year 2025 capital expenditure is directed toward AI. It remains unclear whether the $80 billion investment is a result of Microsoft's plan to expand capital expenditure in the next six months.
Microsoft's investment in AI can be described as an all-in bet.
In fact, Microsoft's saturation investment in AI is already yielding tangible returns. According to financial reports, Microsoft's AI business primarily includes cloud services and AI assistant Copilot services. In the third quarter of this year, Microsoft's cloud revenue, including products like Office and Azure, grew 22% year-on-year to $38.9 billion, with Azure revenue contributing 12 percentage points from AI.
Microsoft CEO Satya Nadella revealed that by the end of 2024, Microsoft's AI business is expected to achieve $10 billion in annual revenue, making it the fastest-growing business segment to reach this milestone.
In addition to building its own data centers, Microsoft has accelerated external investments in AI. For example, it has committed a total of $13 billion to OpenAI, including a $750 million investment in OpenAI's $6.6 billion funding round in early October 2024. Microsoft has integrated OpenAI's models into its software products and cloud platform, generating significant value.
JPMorgan analysts stated that Microsoft's investment in OpenAI a few years ago may prove to be one of the smartest investments ever made.
As of May 2024, Microsoft operates over 300 data centers in 34 countries and regions worldwide, while Amazon has 215.
Amazon, of course, is not lagging behind. Amazon CEO Andy Jassy revealed at the re:Invent conference that the company plans to invest $75 billion in AWS and generative AI in 2025.
In November 2024, Morgan Stanley raised its forecast for Amazon's 2025 capital expenditure by 22% to $96.4 billion, with the increase primarily allocated to purchasing AI GPUs and servers for data centers. The forecast for 2026 was further raised to $105 billion. If Morgan Stanley's predictions are accurate, Amazon's 10-year $100 billion data center investment plan announced six months ago may need significant upward revision.

According to a research report by technology intelligence firm ABI Research, there will be 5,697 public data centers globally by the end of 2024, primarily located in the Asia-Pacific region, Europe, and North America. By 2030, 8,410 data centers will be operational. Owners of hyperscale data centers include well-known tech giants such as Microsoft (Azure), Amazon (AWS), Google (GCP), IBM, Oracle, and Alibaba (Alibaba Cloud). Microsoft and Amazon account for over half of the existing hyperscale data centers.
Google operates fewer data centers than Microsoft and Amazon, with a dozen in the U.S., but it was one of the first major companies to build hyperscale data centers. The Google data center in Council Bluffs, Iowa, which began construction in 2007 with a total investment of $5 billion, was once considered "unbelievably large." To ensure power supply and promote green energy, Google plans to build future data centers near solar and wind farms. The company recently announced a partnership with energy firms to invest $20 billion over the next decade in building multiple "industrial parks" equipped with renewable energy facilities to support data center operations.
Meanwhile, AI startups are rapidly catching up. OpenAI plans to collaborate with Microsoft on a $110 billion hyperscale data center project, including a super AI computer cluster called "Stargate," equipped with millions of servers and expected to launch in 2028. However, no updates have been released since Q3 2024.
But when it comes to audacity, no one surpasses Elon Musk. On September 3, 2024, Musk announced that his xAI team built the Colossus supercomputing cluster with 100,000 H100 GPUs in just 122 days. Moreover, the cluster will double in size in the coming months, expanding to 150,000 H100 and 50,000 H200 GPUs. Located in Memphis, Tennessee, Colossus is the world's largest super data center, serving as the foundation for Musk's vision of a general-purpose AI model, humanoid robots, and full self-driving (FSD) technology.
Three months later, Musk "revised" his statement again. xAI plans to expand the Colossus supercomputing cluster tenfold, eventually integrating over 1 million GPUs to surpass competitors like Google, OpenAI, and Anthropic. Musk even exaggerated, saying the number of GPUs could reach "a billion." Founded in July 2023, xAI recently completed a $6 billion Series C funding round, valuing the company at $40 billion. On January 3, Musk announced that xAI's next-generation AI model, Grok 3, is set to launch, with pretraining already completed. Grok 3's computational power is 10 times that of Grok 2, and Musk previously claimed it would be "the most powerful AI in the world."
The frenzy of data center investments has created persistent and massive demand for chips, servers, and electricity, spawning trillion-dollar supply chains in each sector.
For example, Supermicro became an AI sensation by pioneering liquid-cooled servers, while the need for high-speed connections between AI servers and switches in data centers recently gave rise to the niche AEC (Active Electrical Cable) concept in capital markets. Over the past month, stocks like Credo surged, driving related A-share concepts.
Chinese tech companies are also ramping up AI investments. Reports indicate that ByteDance invests 150 billion yuan annually, while Alibaba and Tencent invest hundreds of billions. However, even combined, these investments pale in comparison to Microsoft's. Therefore, companies like DeepSeek, which smartly leverage limited computing resources with a pragmatic approach, represent a realistic choice.
Fortunately, the transformative potential of AI is far from fully realized.