The staggering size and circular nature of recent deals raises both hopes and concerns, so keeping track will inform industry and investment decisions about where to go from here.
In sum – what to know:
Highly concentrated market – The last 30 days has seen a small number of major tech firms making enormous AI infrastructure deals, unprecedented in size and scope.
Bubble concerns – the frenetic pace of AI infrastructure investment is reminiscent of previous tech bubbles, so some are calling for a selective and measured approach to investment.
Quick rundown of recent deals – a closer look at the significant announcements and partnerships of the past month.
Understanding who is doing what in the world of AI infrastructure can give context about industry direction, economic impact, geopolitical influence, and resource allocation – now and into the future. During the last 30 days, there’s been a somewhat helter-skelter feel to the surge in capital expenditures and data center financing and acquisitions by investment and venture capital firms, hyperscalers, cloud computing, and hardware specialists. Just this week, Ares Management Director, Partner and Co-President Kipp deVeer stated that “the data center boom brings a risk of overbuilding,” which is why he emphasized a need to be “selective and measured in what we build.”
The recent flurry of AI infrastructure partnerships and deals has felt more like a race than an even jog, so for context, let’s take a closer look at recent deals and partnerships (with the understanding that this is by no means an exhaustive list):
One of the biggest deals – the one that seemed to rock the AI infrastructure world – was Nvidia’s $100 billion investment into OpenAI’s massive data center build (with an initial $10 billion investment that will progressively grow for each gigawatt OpenAI deploys). NVIDIA CEO Jensen Huang, OpenAI CEO Sam Altman, and OpenAI President Greg Brockman described the partnership as an enabler for “AI at scale for virtually every industry and user.”
That deal was the foreshock to the next seismic shift, when Oracle, SoftBank, and OpenAI projected a spend of $500 billion for 10 gigawatts of power by end of 2025, with $400 billion already committed to further expanding OpenAI’s Stargate Project. In this partnership, Nvidia agrees to provide the chips for new AI data centers purportedly slated for Texas, New Mexico, Ohio, and one undisclosed Midwestern location. Whether the recent OpenAI-AMD deal reshapes some of this partnership remains to be seen.
Other big deals included that between Microsoft and Nebius Group, with $17.4 billion in GPU slated to go toward capacity in the next five years; and that between CoreWeave and both OpenAI and Meta (the former deal worth $22.4 billion, and the latter worth $14.2 billion). Significant was the fact CoreWeave’s revenues would be less reliant on Microsoft now that Meta would be another major customer.
Not to be left in the dust of the U.S. infrastructure stampede, the U.K. began lassoing some big hyperscalers and neocloud providers in what are now evolving U.K.-U.S. partnerships – starting last month with a meeting among Nvidia’s CEO Huang, OpenAI CEO Altman, and U.K. Prime Minister Keir Starmer — the latter pledging £1 billion ($1.34B USD) for “AI Growth Zones” that would deliver 20x the computing power of current capacity). In the dealmaking, Nvidia and its partners CoreWeave (via CoreWeave Ventures), Microsoft and Nscale committed to building and operating AI factories across the United Kingdom, Norway and United States, by end of 2026. Starmer called Nvidia’s £2 billion ($2.67B USD) investment in the U.K.’s AI sector the “biggest-ever tech agreement” between the two countries.
In the same time period, Nscale announced it would scale up with 300,000 NVIDIA Grace Blackwell GPUs worldwide, dedicating up to 60,000 GPUs in the U.K., which dovetails with Nvidia’s promise to scale up with 120,000 NVIDIA Blackwell Ultra GPUs (investing up to £11 billion for U.K. data centers).
Not to be left out, Alphabet’s Google also pledged £5 billion over two years of AI investment in the U.K. (part of Alphabet’s $85 billion CAPEX forecast). Worth noting is that earlier in the year, Google committed $25 billion to AI infrastructure, as well as billions more in investment for hydropower upgrades across the United States. Alphabet’s CAPEX pledge rivals Microsoft’s projection of $80 billion, but is dwarfed somewhat by the AWS pledge of $100 billion.
The size and scope of these deals is raising hopes as well as concerns. The hope is for more jobs, GDP growth, and stable economies; the concern is risk of default and interest rate exposure to businesses, investors, and retirement accounts (which can now include “ alternative assets,” such as AI infrastructure investments). That’s especially true in cases where there is financing through debt rather than cash flows. To some experts, the incestuous nature of AI infrastructure investment creates a possible “AI bubble,” which may burst if returns don’t materialize as quickly as investments. Time will tell if AI’s rate of adoption will be fast enough to drive profitability, keeping the bubble intact.