
The ins and outs of Microsoft's downgrade

Microsoft's demand has significantly slowed down in the past few months, mainly due to market digestion, design changes, and financial discipline. Although Microsoft was a major player in data center leasing in 2023 and the first half of 2024, it exited multiple large-scale deals starting in September 2024, abandoning approximately 1GW of letters of intent. The strained relationship with OpenAI may also have impacted its data center demand, leading to the suspension of construction on a large data center in Wisconsin
TD Cowen has released several reports over the past three months, with a few original reports at the end. Let's first outline the timeline:
January 24, 2025 Report: "Microsoft's Data Center Leasing and Land Acquisition Slowdown in the Second Half of 2024, We Believe This is Related to OpenAI/'Stargate' Project's Business Shift to Oracle"
Microsoft led the data center leasing rankings with 2.4GW of leasing volume in the Americas in 2023 and 1.5GW in the first half of 2024. However, our investigation shows that Microsoft's demand for data centers has significantly slowed down since September 2024, withdrawing from the market.
Specifically, our investigation indicates that Microsoft has exited multiple early or mid-stage negotiations for deals exceeding 100MW in capacity, abandoning approximately 1GW of letters of intent (we believe this includes Applied Digital), and has given up at least five signed land contracts.
Although our investigation shows that Microsoft entered the market in early January with a demand for over 500MW of capacity for delivery in 2026, the absence of Microsoft's hyperscale leasing team at the PTC conference (the first time this has occurred recently) has raised questions about its capacity needs and roadmap.
Additionally, we have heard several potential reasons for the slowdown in demand, primarily including: 1) Entering a digestion period after a large amount of leasing in 2023 and the first half of 2024; 2) Design changes; 3) Financial discipline.
Considering various factors, we believe the slowdown in Microsoft's demand is related to OpenAI seeking data center capacity elsewhere. In the first half of 2023 and 2024, Microsoft was the most active lessee of data center capacity, procuring capacity based on forecasts that included OpenAI's additional workloads. However, by September, we found that 33% of the capacity Microsoft had pre-leased was lagging behind schedule, which may have originally been planned to support OpenAI.
Around the same time, media reports began to emerge about tensions between Microsoft and OpenAI, with OpenAI leaders suggesting that Microsoft was too slow in meeting its data center needs to keep pace with its development. Furthermore, recently, Microsoft decided to suspend the construction of the second phase of a large data center campus in Wisconsin, which we previously emphasized was originally built to support OpenAI's workloads.
Finally, we believe the announcement of the "Stargate" project between Microsoft and Oracle further demonstrates a shift in the allocation of OpenAI's additional business between Microsoft and Oracle.
February 2025 Report: "Survey Shows Microsoft is Terminating Leases in Some Areas Due to Delays in Facility Construction or Power Supply"
Microsoft has also slowed down the conversion of signed Statements of Qualifications (SOQ, a prerequisite document for data center leasing) into formal leasing agreements. Currently, it is unclear whether this is merely a delay in SOQ to lease conversion or a complete termination of SOQ, ceasing any lease conversionsThe success rate of converting SOQ into formal lease agreements is close to 100%, and data center operators typically use this as a signal to initiate data center construction.
Additionally, our channel research indicates that Microsoft has reallocated a significant portion of its anticipated international spending back to the United States, suggesting a notable slowdown in its international leasing business.
As we highlighted in the key points from the 2023 Pacific Telecommunications Council (PTC) conference, Meta has also adopted a similar strategy. At that time, our research revealed that Meta canceled its $4 billion capital expenditure plan related to the metaverse (subsequently, Meta lowered its capital expenditure guidance by $4 billion two weeks later) and used this as a reason to cancel multiple data center lease agreements in the United States.
March 26, 2025 Report: "Microsoft's Contraction Creates Opportunities for Google and Meta"
Our channel research conducted at the NVIDIA GTC conference and DCD Connect event shows that although the number of lease cancellations or delays by Microsoft is still expanding, the overall demand for data centers has still increased year-on-year. This creates opportunities for Google and Meta to fill the capacity gaps.
The redesign of hyperscale data centers for higher density has led to a slowdown in data center equipment procurement, which we believe will negatively impact Vertiv's order volume in the first half of 2025.
The clear message conveyed at last week's NVIDIA GTC conference in San Jose and this week's DCD Connect event in New York is that, driven by the same factors we emphasized in late February, the overall demand for hyperscale data centers has increased year-on-year. However, specifically regarding Microsoft, further channel research indicates that the number of lease cancellations continues to rise, with leases being terminated in both the United States and Europe.
We believe that Microsoft's reduction in new capacity leasing is primarily due to its decision to no longer support OpenAI's incremental training workloads. However, we still believe that the cancellations and delays in leasing indicate an oversupply of data centers relative to its current demand forecasts. The delayed leases suggest that Microsoft is reserving for mid-term cloud computing and inference workloads, while the portions exceeding the updated mid-term capacity needs are being directly canceled.
On the positive side for third-party data center operators, our research shows that Google is stepping in to fill the capacity that Microsoft has abandoned in the international market, while Meta is filling the related capacity gaps in the United States. Both of these hyperscale data center operators have seen significant year-on-year growth in data center demand.
The reason for Google's demand growth is that it began to contract in late August 2024 to improve the utilization of existing data centers, but is now starting to replenish. As for Meta, its demand growth is to support the next generation of Llama training. Additionally, our research indicates that OpenAI is increasingly inclined to procure data center capacity directly from third parties, with its recent deal with CoreWeave being a prime exampleWe have observed that OpenAI has significant long-term data center capacity planning, intending to create multiple "stargate" projects, each with a capacity between 800 megawatts and 1.5 GW, with a cumulative long-term capacity demand potentially exceeding 6 GW. To meet this capacity demand, our research shows that OpenAI is recruiting personnel with design, construction, and capacity planning experience from other hyperscale data center operators, which we believe indicates that OpenAI may begin to build data centers on its own in the medium to long term.
Summary
Microsoft must have taken a step back; OAI's subleasing is not the reason. It is precisely because Microsoft has stepped back that OAI has turned to third parties like Oracle.
Meta, Google, AWS, and the new consortium led by OAI, along with third parties, are still continuing to invest. Are they smarter than Microsoft? Or are they lagging behind?
Information Equality, Original Title: "The Background of Microsoft's Downgrade"
Risk Warning and Disclaimer
The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investing based on this is at one's own risk