The biggest news in AI this year isn’t a new model or app, its infrastructure. Just yesterday, OpenAI announced that it will expand their “Project Stargate”, A plan to build colossal data centers across the U.S, with an additional 5 new data centres. Pushing the cost of its AI infrastructure project to about $400 billion dollars (Hammond, 2025). These facilities, often described as “AI factories”, are designed to run future generations of artificial intelligence at a scale the world has never seen.
The economics are simple. Lots of upfront costs, but very low marginal costs when the system is in place (Kenton, 2025). Building a data center stuffed with GPUs and advanced cooling systems costs tens of billions. But once the investment is made, the cost of running one more AI query is almost negligible. The more queries, models, and customers using Stargate, the cheaper it becomes per unit. That’s economies of scale in its purest form, and it explains why OpenAI is betting that size equals survival in the AI race.
Still, scale has its risks. If demand for AI plateaus or grows slower than expected, those massive data centres could sit underused, turning scale from an advantage into a liability. And even if the demand is there, this type of scale brings side effects: massive energy consumption, huge amounts of water used for cooling, and heavy reliance on global chip supply chains. If things don’t pan out the way OpenAI anticipates, the efficiencies of scale for Project Stargate could translate into higher costs for society.
My take: Stargate is both visionary and dangerous. On one hand, it could make advanced AI affordable, accelerating innovation across industries. On the other, it concentrates technological and economic power into very few hands, while massively contributing to environmental costs. Economies of scale may make AI cheaper, but the real question is: cheaper for whom, and at what price for everyone else?
References:
Hammond, G. (2025, 23 september). OpenAI expands Stargate AI project with five US sites. Financial Times. https://www.ft.com/content/9b6c7db8-9a14-4261-9c18-38ec84d869a0?
Kenton, W. (2025, 18 juni). Economies of Scale: What Are They and How Are They Used? Investopedia. https://www.investopedia.com/terms/e/economiesofscale.asp
OpenAI. (2025, januari). Announcing the Stargate project. https://openai.com/index/announcing-the-stargate-project/
The post raises some important questions about feasibility of ubiquitous use of AI and whether it is even needed in the first place. Will making it more affordable and accessible truly improve our everyday lives or lead to more challenges in terms of higher expected productivity and work stress?