The “AI Bubble” isn’t bursting, it’s expanding

We keep hearing it: “Is the AI bubble about to burst?” But the data from the field, and the numbers from Nvidia, say something quite different. This bubble is far from a pop. From our corner of the AI world, we see demand isn’t slowing, it’s accelerating.

Yesterday, 19 November 2025, Nvidia reported its Q3 (fiscal 2026) results:

• Revenue of $57.0 billion, up 22% quarter-on-quarter and 62% year-on-year
• Guided for Q4 revenue of $65.0 billion** (±2%), well ahead of many expectations

CEO Jensen Huang made it clear: “There’s been a lot of talk about an AI bubble. From our vantage point, we see something very different.”

Let’s unpack why the bubble narrative is misleading in this context:

Infrastructure build-out is still early

The transition underway isn’t just software; it’s hardware + systems + data centre scale + new architectures (training, inference, agentic AI). As Huang put it: “Compute demand keeps accelerating and compounding across training and inference, each growing exponentially. We’ve entered the virtuous cycle of AI.” That suggests we are not at the end of the story, but somewhere closer to the beginning or middle.

Real ROI signals are emerging

Earlier this month there was a piece in the Financial Times about research showing clear productivity gains from GenAI deployments. We wrote about it here.

• AI assistant before point of purchase boosted sales by 16.3 % and conversion by 21.7 %.
• Hybrid AI-+ human model delivered an 11.5 % sales lift compared to human-only.

These are real business impact numbers, not just hype.

It means investments aren’t just speculative; they’re starting to pay.

Bubble fears often focus on valuations, not fundamentals

Yes, valuations are high. Yes, there is risk. But valuation is not the same as demand drying up. High valuations may mark the price, but what matters to the infrastructure build-out is commitment: from cloud providers, enterprises, governments. And Nvidia’s backlog, supply constraints, and growth point to demand far exceeding simplistic “bubble” narratives. For example: demand for Blackwell chips reportedly “sold out”.

The “bubble” narrative mis-frames the shift

A bubble implies a speculative froth detached from utility and demand. But what we’re seeing looks like a deeper structural shift: AI embedding into workflows, expanding into new industries, creating new infrastructure. When that happens, what looks like a bubble often turns into a foundation for decades of growth. (Think: mobile internet, cloud infrastructure.)

From our perspective we’re not seeing any slowdown in demand. We’re seeing:

• More companies adopting AI tools at scale, not just pilots.
• More teams requesting AI-assisted workflows. Not just “let’s experiment” but “let’s deploy”.
• Bigger customers, broader use-cases, and higher expectations: enterprises want ROI, and they’re moving fast.
• Teams talking about outcomes (productivity, revenue, generative workflows) rather than “we should try AI someday”.

Rather than a bubble ready to burst, we appear to be in the growth phase of a structural transformation.

For businesses: If you’re waiting for the “AI hype to die down” before making a move, you may be putting yourself behind. Making early, thoughtful deployments gives you advantage, not just in technology, but in people, culture, workflow.

For investors: The Nvidia numbers suggest that the underlying demand drivers (chips, data centres, AI systems) remain extremely strong. If companies like NVIDIA are delivering 60+ % growth, guiding even higher, and selling out infrastructure, then the argument that the game is almost done looks weak.

For the ecosystem (partners, tools, service providers): This is an opportunity moment. As more firms adopt AI aggressively, the supporting ecosystem (services, integration, workflow tooling, inference operations) expands. At Vinton we see this every day. The demand is there. The momentum is building.

A word of balance (but not caution)

Of course I’d be remiss not to note: high growth comes with risk. Infrastructure build-outs are expensive. Talent bottlenecks persist. Energy and data-centre constraints loom. Valuations are stretched. But none of these equal a bursting bubble. They equal scaling challenges, which is exactly what you see when an industry enters its growth inflection.