Key Takeaways

  • Roman Chernin, co-founder of Nebius, firmly believes the AI infrastructure boom is not a bubble, but instead the earliest phase of a massive, sustained expansion.
  • He argues that even technologically advanced companies are only beginning to apply AI, using it in a "small fraction" of their operations, indicating vast untapped potential.
  • The industry is “just at the beginning of this amazing moment when Jensen calls it like useful AI,” with truly scalable, effective AI use cases only now starting to emerge.
  • Compute demand is set to grow exponentially as more "unsolved problems" are tackled and enterprises integrate AI beyond initial experiments.

The Bubble Myth vs. Reality

For founders immersed in the AI whirlwind, the question often lingers: is this a bubble about to pop? Roman Chernin, co-founder of AI infrastructure firm Nebius, gives a direct answer: “No, I don't believe it's a bubble.” His conviction stems not from blind optimism, but a clear-eyed assessment of AI's current penetration—or lack thereof—in the real world.

Chernin argues that the perceived surge in AI investment and hype masks a deeper truth: we are barely scratching the surface of practical AI adoption. “If you take any large company, even pretty advanced technologically, you will see that they just starting,” Chernin observed. He's talking about the core operations of major enterprises, where AI is currently only applied to "a small fraction" of what's possible. This market isn't mature, nor is it one over-leveraged on future promises. Rather, it's a foundational technology just beginning to find its footing outside of labs and into real-world applications.

Think about it: how many mission-critical systems in Fortune 500 companies run on generative AI agents today? How many factory floors are optimized by predictive AI? Very few. Chernin's point is that the current investment wave is building the runways for an industry that hasn't even fully taken flight yet.

The Inevitable Explosion of Useful AI

If it's not a bubble, what is it? Chernin calls it “this amazing moment when Jensen calls it like useful AI.” He sees the industry poised for an explosion of genuine utility, a moment when AI moves from interesting demos to indispensable tools at scale. We are, he suggests, “just few months from the moment when we've got maybe first use case that's works in like in the scale.”

This isn't about incremental gains. This is about unlocking entirely new capabilities across every sector. As these "first use cases" prove their worth, demand for the underlying compute will surge. "We will see obviously like many many many more use cases and we will see much more adoption," Chernin stated. His logic is simple: the more problems AI can solve, the more compute power it will require. The total market "pie" is not fixed; it is expanding exponentially as we solve the "so many unsolved problems yet."

Founders often fear market corrections, but Chernin's perspective offers a different framing: the current compute investment is simply matching the potential demand, not exceeding the realized demand. The infrastructure built today will be stretched thin tomorrow, not sit idle.

What to Do With This

Don't build your AI company for a quick flip based on hype cycles. Instead, commit to the long game. Chernin’s take means the foundational opportunities in AI infrastructure and applications are immense and sustained, not fleeting. This week, pick one specific, currently "unsolved problem" within a large, traditional industry that could genuinely benefit from AI at scale. Don't just prototype; map out the full journey from bare metal to managed inference, anticipating the exponential compute needs when that solution becomes indispensable. Build for the world that will exist when "useful AI" is everywhere, not just the world that is.