Microsoft and OpenAI’s once‑rock‑solid alliance is showing signs of strain. In early April 2025, Microsoft quietly pulled out of two major data‑center deals—each capable of supplying hundreds of megawatts of power—that had been earmarked to support additional ChatGPT training workloads. This move, first reported by Windows Central, suggests Microsoft is deliberately dialing back its external commitments even as it insists on investing $80 billion in AI infrastructure this year [Windows Central].
When Microsoft first invested billions to become the exclusive cloud provider for OpenAI, it cemented one of the most talked‑about “tech bromances” in history. Azure powered every stage of GPT’s development, and Microsoft integrated ChatGPT deeply into products like Copilot and Bing. But when OpenAI unveiled its $500 billion Stargate project—with partners Oracle and SoftBank—to build a nationwide AI compute backbone, that exclusivity was bound to loosen.
Instead of an abrupt breakup, the new arrangement grants Microsoft a “right of first refusal” on any of OpenAI’s cloud capacity needs. In other words, OpenAI can shop around—but must give Microsoft the chance to match any rival offer first. It’s a departure from a single‑vendor lock‑in model, yet it still keeps Azure at the center of OpenAI’s roadmap through at least 2030 [Windows Central].
Why pull back now?
Source: Windows Central Microsoft doesn't want to support OpenAI's ChatGPT training — Is the multi-billion dollar tech bromance fraying further apart?
From Exclusive Ally to “Right of First Refusal”
When Microsoft first invested billions to become the exclusive cloud provider for OpenAI, it cemented one of the most talked‑about “tech bromances” in history. Azure powered every stage of GPT’s development, and Microsoft integrated ChatGPT deeply into products like Copilot and Bing. But when OpenAI unveiled its $500 billion Stargate project—with partners Oracle and SoftBank—to build a nationwide AI compute backbone, that exclusivity was bound to loosen.Instead of an abrupt breakup, the new arrangement grants Microsoft a “right of first refusal” on any of OpenAI’s cloud capacity needs. In other words, OpenAI can shop around—but must give Microsoft the chance to match any rival offer first. It’s a departure from a single‑vendor lock‑in model, yet it still keeps Azure at the center of OpenAI’s roadmap through at least 2030 [Windows Central].
The Two Mega Data‑Center Deals That Didn’t Happen
According to PC Gamer, Microsoft stepped away from two data‑center agreements that would have delivered roughly 2 gigawatts of combined capacity—enough to run two sizeable AI training farms [PC Gamer]. Reuters corroborated this, noting that Microsoft “no longer wants to offer additional support to OpenAI for ChatGPT’s training” and has allowed several hundred megawatts of capacity leases to lapse across the U.S. and Europe [Reuters].Why pull back now?
- Supply–Demand Calibration: Internal forecasts may have shown that existing on‑premises capacity, combined with Azure’s current footprint, is sufficient for Microsoft’s AI ambitions.
- Cost Discipline: With cloud revenue growth under investor scrutiny, retrenching on costly external leases can help protect margins.
- Strategic Flexibility: By redirecting spending toward in‑house data centers or different regions, Microsoft retains the agility to pivot as AI demand evolves.
What It Means for the AI Race
Is Microsoft signaling that the AI hype has outpaced real‑world infrastructure needs? Or is this a savvy financial pivot to balance its books while still committing massive capital to AI overall? Consider:- OpenAI’s Compute Independence: Stargate’s debt‑heavy financing puts pressure on SoftBank and Oracle to deliver capacity on time and on budget—failure could force OpenAI back toward Azure.
- Competitive Pressures: New entrants like DeepSeek—whose open‑source models reportedly rival GPT‑4 at a fraction of the cost—add urgency for both Microsoft and OpenAI to optimize spend and performance.
- Market Sentiment: Shares of energy and infrastructure firms dipped when news of Microsoft’s lease cancellations broke, underscoring investor concern about overcapacity and the true pace of AI growth.
Lessons from Tech History
Tech partnerships often begin with exclusivity and end in diversification. IBM once dominated enterprise servers before embracing open architectures; Apple and Google moved from uneasy bedfellows to fierce rivals in mobile OS. Microsoft and OpenAI’s journey mirrors these arcs: start narrow, build trust—and then expand the partnership’s terms to manage risk and fuel innovation.Looking Ahead
For Windows users and enterprises, Azure remains the primary gateway to OpenAI’s APIs and Copilot‑powered features. But the door is ajar for multi‑cloud strategies, potentially driving better pricing, performance, and specialized AI services from other providers. Will Microsoft’s measured retrenchment prove prudent, or will it cede ground to hyperscalers with more aggressive expansion plans? Only time—and the next wave of AI breakthroughs—will tell.Source: Windows Central Microsoft doesn't want to support OpenAI's ChatGPT training — Is the multi-billion dollar tech bromance fraying further apart?
Last edited:
