OpenAI Shifts Strategy from Building to Buying Compute Capacity Amid IPO Scrutiny

OpenAI Shifts Strategy from Building to Buying Compute Capacity Amid IPO Scrutiny March 22, 2026 In a significant strategic pivot, OpenAI is retreating from ambitious plans to build its own massive data centers and is instead focusing on securing computational capacity through partnerships with major cloud providers. This shift underscores the intense pressure the artificial intelligence leader faces to demonstrate fiscal discipline and a viable path to profitability as it prepares for a potential initial public offering later this year. The change in approach was highlighted by CEO Sam Altman during a recent industry summit, where he candidly discussed the operational challenges of large-scale infrastructure. Altman cited a severe weather incident at a key data center campus in Abilene, Texas—part of the flagship $500 billion "Stargate" project involving OpenAI, Oracle, and SoftBank—that temporarily disrupted operations. He also pointed to persistent supply chain issues and tight deadlines as realities forcing a reassessment. This strategic recalibration comes as OpenAI, valued at a staggering $730 billion in a recent funding round, transitions from a private market phenomenon to a company under the microscope of public market investors. Analysts note the move is a direct response to Wall Street's concerns over the company's previously "reckless" spending on growth. "The market wants to see OpenAI's revenues rolling at a pace in which the spending can be justified," said Daniel Newman, CEO of Futurum Group. OpenAI's operations are extraordinarily compute-intensive, requiring vast amounts of chips, power, and memory. Altman has repeatedly stated that compute is a major bottleneck, a constraint severe enough to force the company to rate-limit its products. To address this, OpenAI had embarked on a series of monumental infrastructure deals last year, including a highly publicized agreement with Nvidia that envisioned up to $100 billion in investment tied to deploying at least 10 gigawatts of AI data centers—a commitment that sparked fears of an AI bubble. However, the company is now tempering these ambitions. In recent investor communications, OpenAI has revised its total compute spend target for 2030 down to roughly $600 billion, a figure intended to align more closely with projected revenue growth. Concurrently, the company is emphasizing operational focus, declaring a "code red" to improve its core ChatGPT product and aggressively targeting high-value enterprise use cases. The practical difficulties of building gigawatt-scale data centers, a process experts say can take three to ten years, have proven formidable. Faced with construction hurdles and financing challenges, OpenAI has pivoted to a capital-light model, relying on partners like Oracle, Microsoft, and Amazon to piece together capacity. As part of a recent $110 billion financing round—which included $50 billion from Amazon—OpenAI committed to using 2 gigawatts of capacity on Amazon's Trainium chips and 5 gigawatts on future Nvidia systems. Nvidia's separate $30 billion investment in OpenAI, finalized recently, is notably decoupled from the specific deployment milestones of the earlier $100 billion deal, which now appears to be on hold. Nvidia CEO Jensen Huang indicated that such a massive direct investment is likely "not in the cards" and suggested this might be the last pre-IPO investment in OpenAI. "OpenAI is doing what it must do, which is gain access to compute at scale," Newman observed, framing the scramble for computational resources as the defining race in AI. "And because their cost structure is so high, their route to profitability will be scrutinized every step of the way." Source: CNBC