OpenAI president: compute shortage will persist indefinitely

During a roundtable discussion in New York City on Thursday, OpenAI president Greg Brockman made a stark prediction about the future of the AI industry: "My prediction is very much that we will remain in this compute shortage no matter what." He elaborated that the shortage is forcing hard choices. "Right now we have to make hard decisions on what models we actually train, what products we actually scale," he said.

The shortage afflicts the entire industry. AI labs like OpenAI and Anthropic, alongside Big Tech companies including Microsoft, Amazon, Meta, and Google, are all struggling to meet demand for their AI services. They must juggle training new models while serving software to customers—a task that requires massive amounts of computing power from specialized graphics chips like Nvidia's. Despite collectively spending hundreds of billions of dollars on data centers, the industry cannot keep pace.

Google's situation illustrates the desperation. During the company's Q2 earnings call on Wednesday, CFO Anat Ashkenazi said that despite Google's heavy spending on its own data centers, the company will contract with third parties to access computing capacity in Q3. Most strikingly, Google signed a deal to rent out AI capacity from SpaceX for $920 million（約1500億円） a month in June. Brockman described a dynamic where even as OpenAI ramps up capacity, he expects surplus compute—only to see demand surge as the company develops new models and new applications. "I think that that is going to be how it will feel because we are shifting to this compute-powered economy," he explained.

Brockman also addressed how companies and their customers are using AI. He said customers are beginning to see a return on their investments in AI, citing improvements in productivity when the technology is applied thoughtfully. He pushed back against "tokenmaxxing"—using as much AI computing as possible simply to demonstrate capability—calling it wasteful. "I think people are starting to see that [return on investment] is there if you use [AI] right. If you just token max, then yeah, of course, you're just going to get max tokens. But if you value max, if you really try to apply these technologies to problems that matter, then you get the return," Brockman said, suggesting that the market may gradually shift toward more disciplined spending patterns.

The compute shortage in AI has become a structural constraint for the industry. Brockman's prediction that it will persist "no matter what" reflects a fundamental mismatch: even as companies like OpenAI, Anthropic, Microsoft, Amazon, Meta, and Google collectively spend hundreds of billions on infrastructure, the economic demand for AI services—and the ability to train ever-larger models—grows at least as fast. This is not a temporary supply-chain problem but a feature of a "compute-powered economy," in Brockman's phrase, where the limiting factor is the availability of specialized hardware (graphics chips like Nvidia's) and the data centers to house them.

The specifics underscore the severity. Google, despite being one of the world's largest technology companies with massive capital reserves, has resorted to renting capacity from third parties—paying SpaceX $920 million（約1500億円） per month as of June—in addition to building its own infrastructure. This signals that even the largest players cannot self-sufficiently meet their own demand. For OpenAI and other AI labs, the constraint forces strategic triage: they must choose which models to build and which products to prioritize, accepting that they cannot pursue all opportunities simultaneously.

Brockman's comments also hint at a secondary challenge: distinguishing signal from noise in how customers use AI. He dismissed "tokenmaxxing"—using as much AI compute as possible simply to demonstrate capability—as wasteful, and suggested that real return on investment comes from strategic application. This implies that even if the hardware shortage were solved, there remains a question of whether the market is being thoughtful about how to allocate scarce resources.

No comments yet. Be the first to share your thoughts!