OpenAI plans to build 250GW of data center capacity by 2033. CEO Altman views this as a "violent industrialization" pathway to artificial general intelligence, but the company still faces enormous challenges including power supply, trillion-dollar funding requirements, and supply chain bottlenecks.

Last week, the flagship data center site in Abilene, Texas officially became operational as part of OpenAI and Oracle's $500 billion Stargate project. OpenAI CEO Altman showcased the preliminary results of this ambitious project to the media.

(OpenAI CEO Sam Altman at the Stargate data center in Abilene, Texas)

At this 800-acre construction site, 6,400 workers are working intensively, with fiber optic cable installation alone extending enough length to circle the Earth 16 times. Altman stated:

"This massive construction site is only a small fraction of the future scale, not even enough to meet ChatGPT's demands."

According to information disclosed internally by OpenAI, the company expects to build data center capacity exceeding 2GW by the end of 2025, with plans to reach an astonishing 250GW scale by 2033. This target represents approximately one-quarter of the current total installed power generation capacity in the United States (about 1,200GW).

Altman's core strategy is "scaled computing." This doesn't refer to algorithmic breakthroughs, but rather to pushing artificial intelligence toward artificial general intelligence (AGI) and artificial superintelligence (ASI) through "violent industrialization" involving millions of chips, massive data center parks, gigawatt-level power, and substantial cooling water.

Under this logic, the standards for measuring AI capabilities have fundamentally shifted. Altman explained:

"At such scale, the number of GPUs becomes meaningless, replaced by the power consumed by entire chip clusters—gigawatts (GW). GW has become the only standard for measuring how much effective computing power a company can maintain."

However, realizing this plan faces enormous challenges, including power supply, funding requirements, and supply chain bottlenecks. Industry insiders question whether such massive infrastructure investment is realistic and whether it's worth paying such a high price for AI development.

**Power Demand Equivalent to 250 Nuclear Plants**

OpenAI's 250GW target represents enormous demand on the power system.

A typical nuclear power plant generates approximately 1GW, meaning that supporting just OpenAI's AI development alone would require building new generation capacity equivalent to 250 nuclear plants.

For comparison, Microsoft's second-ranked Azure cloud business had total operational power consumption of only around 5GW by the end of 2023 for serving all customers.

Large data centers previously typically consumed between 10 to 50 megawatts, but developers are now planning single campuses reaching thousands of gigawatts in scale, comparable to entire cities' energy consumption.

However, the computing power Altman refers to extends far beyond electricity. This figure represents an entire industrial system, including: data centers, chips, cooling and water systems, network fiber optics, and high-speed interconnect equipment connecting millions of processors into supercomputers.

Sources familiar with the matter revealed that OpenAI's rapidly growing server demands have surprised even executives at key supplier Nvidia.

To address power challenges, OpenAI and its partners are adopting unconventional approaches, including building their own power plants rather than waiting for utility companies to provide grid electricity, or locating facilities in remote areas with easier energy access.

This is because utility companies are inherently conservative in adding new generation capacity, unwilling to risk building power plants that could lead to overcapacity due to a single company's demands.

OpenAI has planned mixed energy solutions using natural gas, wind, and solar power in Texas, but this still cannot easily fill the massive gap of hundreds of GW.

**Trillion-Dollar Investment and Supply Chain Bottlenecks**

Beyond power, funding and supply chains are two other major constraining factors.

Altman candidly admitted in an internal letter that OpenAI "has already invested hundreds of billions of dollars, and completing this task will require trillions more." It also requires "activating the entire global industrial base—energy, manufacturing, logistics, labor, supply chains."

OpenAI has already invested heavily in this effort. Before announcing the 250GW target, the company had already contracted to obtain approximately 8GW of computing power by 2028, which alone requires paying hundreds of billions of dollars to cloud service providers like Microsoft.

Additionally, based on current costs of approximately $50 billion to build a 1GW nuclear plant, power facility investment alone could reach $12.5 trillion.

Supply chain bottlenecks are equally severe. Supporting such massive computing power expansion would require chip foundry giant TSMC to provide more capacity for producing Nvidia's GPUs, and lithography equipment manufacturer ASML to provide more equipment.

Capacity expansion in these segments cannot be achieved overnight, requiring risk investment and coordination across the entire upstream supply chain. Even with Nvidia's commitment to provide funding support for OpenAI's data centers, adding new capacity will remain an arduous process.

**OpenAI's "Ambition" is Actually a High-Stakes Gamble**

Ultimately, OpenAI's stunning plan is a high-stakes gamble based on belief.

Altman and his competitors firmly believe that larger-scale GPU clusters are the only path to more powerful AI models and key to unlocking AGI and ASI. Like historic grand projects such as the Hoover Dam and Apollo program, this reflects unwavering faith in future technological transformation.

Analysis suggests that for investors and markets, how to view this bet depends on judgments about AI's future.

If one believes super AI can solve human challenges like cancer, then trillion-dollar investment becomes necessary. Conversely, it could become a "massive engineering disaster" recorded in history like California's high-speed rail project.

Regardless of whether the 250GW target is ultimately achievable, this AI-driven, almost frenzied infrastructure construction boom has already begun. It is reshaping energy, land, and capital markets in unprecedented ways, while society seems not yet fully aware of the enormous costs and far-reaching impacts behind it.

As Altman himself acknowledged, when people use ChatGPT, few think about the dusty massive construction sites behind it and the industrial power they represent.