Build 250GW of data centers by 2033! Is OpenAI’s “ambition” realistic?

OpenAI plans to build 250GW of data centers by 2033. CEO Altman sees this as a path of "brutal industrialization" towards artificial general intelligence, but still faces huge challenges such as power supply, the need for trillions of dollars, and supply chain bottlenecks. Last week Wallstreetcn mentioned that the flagship data center site in Abilene, Texas was officially put into operation. As part of the $500 billion Stargate project being promoted by OpenAI and Oracle, OpenAI CEO Altman showed the media the initial results of this ambitious project. (OpenAI CEO Sam Altman at the Stargate data center in Abilene, Texas) On this 800-acre construction site, 6,400 workers are busy. The laid fiber optic cable alone could circle the earth 16 times. Altman said: This huge construction site is just a small part of the future scale and is not even enough to meet the needs of ChatGPT. According to internal disclosures at OpenAI, **the company expects to build data center capacity exceeding 2GW by the end of 2025, and plans to reach an astonishing 250GW by 2033, a scale accounting for about one-fourth of the current total U.S. installed power generation capacity (about 1200GW).** The core of Altman’s strategy is “scale computing.” This doesn’t mean algorithmic breakthroughs, but **driving AI towards general artificial intelligence (AGI) and artificial superintelligence (ASI) via millions of chips, large-scale data center campuses, gigawatt-level power, and massive amounts of cooling water in a “brutal industrialization” approach.** Under this logic, the criteria for measuring AI capability have undergone a fundamental shift. Altman explained: **At this scale, the number of GPUs is meaningless, replaced by the electricity consumption of the entire chip cluster—gigawatts (GW). GW has become the only standard for measuring how much effective computing power a company can maintain.** But the realization of this plan faces huge challenges, including power supply, funding needs, and supply chain bottlenecks. Industry insiders question whether such massive infrastructure investments are realistic, and whether it’s worth paying such a high price for AI development. ## Electricity demand equivalent to 250 nuclear power stations OpenAI’s 250GW target means enormous demand on the power system. **A typical nuclear power plant generates about 1GW. That means supporting OpenAI alone would require new generation capacity equivalent to 250 nuclear power plants.** As a comparison, Microsoft’s Azure cloud business, which ranks second, had a total operational power consumption of only around 5GW for all customers by the end of 2023. Reportedly, a large data center in the past usually consumed 10 to 50 megawatts, but now developers plan for a single campus to reach thousands of megawatts—a level comparable to the energy use of a city. However, the computing power Altman refers to is more than just electricity. That figure represents the entire industrial system, including: **data centers, chips, cooling and water systems, network fiber, and high-speed interconnects linking millions of processors into a supercomputer.** Sources cited in reports say OpenAI’s rapidly growing server demand has even surprised key supplier Nvidia’s executives. To address the power challenge, **OpenAI and its partners are exploring unconventional solutions, including building their own power plants instead of waiting for utilities to provide grid electricity, or locating facilities in remote regions with easier access to energy.** **The reason is that utility companies are inherently conservative about adding generation capacity—they will not risk building power plants that could cause overcapacity just for one company’s needs.** Reportedly, OpenAI is planning a mix of natural gas, wind, and solar power in Texas, but this still cannot easily fill the hundreds-of-GW gap. ## Trillion-dollar investment and supply chain bottlenecks Besides electricity, funding and supply chain are the other two constraints. **Altman admitted in an internal letter that OpenAI “has already invested hundreds of billions of dollars, but to do this well will require trillions more.” At the same time, it will need “activation of the entire global industrial base—energy, manufacturing, logistics, labor, and supply chain.”** **Currently, OpenAI has already poured significant funding into this.** According to reports, even before announcing the 250GW target, the company had signed contracts for about 8GW of computing power by 2028, which alone will require paying cloud providers like Microsoft tens of billions of dollars. **Additionally, at the current cost of building a 1GW nuclear power station—around $5 billion—just the investment in power infrastructure could amount to $1.25 trillion.** **Supply chain bottlenecks are equally severe.** To sustain such scale-up in computing power, chipmaker TSMC would need to provide more capacity to manufacture Nvidia’s GPUs, and lithography machine manufacturer ASML would need to supply more equipment. Expanding production in these sectors is not instantaneous; the whole upstream industry chain needs to invest and collaborate. Even if Nvidia pledges financial support for OpenAI’s data centers, adding new capacity will remain an arduous process. ## OpenAI’s "ambition" is basically an all-in bet Ultimately, OpenAI’s astonishing plan is a bet based on belief. **Altman and his competitors firmly believe that larger GPU clusters are the only path to stronger AI models and the key to unlocking AGI and ASI.** Just as the Hoover Dam and Apollo Program were historic mega-projects, behind them is strong faith in future technological transformation. Analysts believe that for investors and the market, how to view this bet depends on their judgment about AI’s future. **If one believes super AI can solve cancer and other human problems, a trillion-dollar investment is necessary. Otherwise, it could become a “huge project disaster” like California’s high-speed railway that is remembered by history.** Regardless of whether the 250GW target is ultimately achieved, this AI-driven, near-manic infrastructure boom has already begun. It is reshaping the energy, land, and capital markets in unprecedented ways, while society seems largely unaware of its immense costs and far-reaching impact. As Altman himself admits, when people use ChatGPT, few think about the sprawling, dusty construction sites and the industrial might they represent. Risk warning and disclaimer The market has risks and investments need caution. This article does not constitute personal investment advice, nor does it take into account the investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are appropriate for their particular situation. Investing accordingly is at your own risk.