Altman Dismisses AI Bubble Fears, Citing Unquenchable Compute Demand as Key Driver

The generative AI boom has ignited a gold rush of investment, raising alarms about a potential market bubble. Yet at OpenAI’s DevDay 2025, CEO Sam Altman pushed back, framing the sector’s exuberance as a natural phase in a technological revolution—one anchored by an insatiable demand for computational resources. 'People will overinvest in some places,' Altman conceded in a press Q&A. 'There will be numerous bubbles and corrections over that period, but what I don’t think [is that] this is totally divorced from reality—there’s a real thing happening here.'

Sabrina Ortiz/ZDNET

Altman’s comments arrive amid skepticism about AI’s return on investment (ROI). A recent MIT study found that 95% of enterprises see no measurable revenue or growth from AI deployments. But Gartner VP Analyst Gaurav Gupta contextualizes this: 'We have been talking about the AI bubble, where enterprises find it hard to achieve ROI beyond initial productivity gains. Hyperscalers and frontier labs continue spending to access more compute because a lot more work needs to be done on LLMs—it’s a race toward AGI.'

The compute deficit is stark. OpenAI’s new partnership with AMD—announced just before DevDay—commits to deploying six gigawatts of AMD Instinct GPUs by late 2026, with AMD granting OpenAI warrants for 10% of its stock. This follows a separate deal with Nvidia for 10 gigawatts of systems (representing millions of GPUs) and a $22.4 billion expansion with GPU cloud provider CoreWeave. Altman emphasized the immediate payoff: 'We can monetize every GPU we get our hands on super well. With 10x more compute, we could build so many more products.'

Product innovation underscores this demand. OpenAI VP Greg Brockman revealed that the Sora video generator’s development hinged on additional compute, while the new Pulse news-digest feature remains limited to paying subscribers due to processing constraints. For developers, this signals a critical bottleneck: scaling AI capabilities requires infrastructure investments far beyond current capacities, prioritizing hyperscalers and well-funded labs over smaller players.

Ultimately, Altman’s dismissal of bubble fears rests on a simple equation—compute equals opportunity. As AI models grow more complex and agents become autonomous, the GPU shortage may delay breakthroughs but also validates the sector’s foundational bets. The real risk isn’t overinvestment; it’s falling behind in securing the silicon to turn today’s prototypes into tomorrow’s tools.

Source: Based on reporting by Sabrina Ortiz for ZDNET