Nvidia Forecasts 70% Growth and Jensen Huang Has Receipts

When Jensen Huang says Nvidia is going to grow 70% next year, you probably expect some vague hand-waving about AI demand. What you get instead is a fairly detailed case for why the chip giant believes it is still in the early innings of something enormous, and it is harder to dismiss than you might expect.

Huang made the remarks this week, outlining why Nvidia sees another year of extraordinary growth ahead despite already sitting at a market capitalisation that would have seemed absurd even five years ago. The company has its chips in data centres, in autonomous vehicles, in robotics, in defence applications, and increasingly in sovereign AI infrastructure programmes around the world. The pitch is simple: every single one of those sectors is still ramping, not plateauing.

The Circular Dealing Question

One of the more pointed questions Huang faced was whether Nvidia's growth is partly circular. The concern goes like this: Nvidia invests in AI startups, those startups raise money partly on the strength of being Nvidia-backed, they then use that money to buy Nvidia chips, and the cycle inflates revenue in a way that flatters everyone. Huang pushed back directly, arguing that Nvidia's venture positions are small relative to the chip revenue they generate, and that the underlying demand for compute is real and independent of any equity relationships.

It is a fair point to raise, and his answer is reasonable but not completely satisfying. The honest reality is that when a single company supplies the infrastructure for an entire industry it has also helped fund, the lines between organic demand and manufactured demand get genuinely blurry. That does not mean the numbers are fake. It means they deserve scrutiny.

What Is Actually Driving This

Strip out the noise and there are a few concrete things underpinning Nvidia's projections. First, the transition from training to inference is accelerating. For years, the big GPU spend was on training massive foundation models. That spend has not gone away, but inference, actually running these models at scale for millions of users, is now a parallel and growing workload. Inference favours Nvidia's architecture in ways that competitors have struggled to match at the high end.

Second, sovereign AI is a real and underappreciated driver. Governments from the Gulf states to Southeast Asia to Europe are building out national AI infrastructure, and they are largely buying Nvidia. These are not small purchases. Several countries have announced multi-billion-dollar GPU procurement programmes in 2026 alone.

Third, the robotics wave is arriving faster than most analysts predicted two years ago. Nvidia's simulation and training stack for physical AI, the software that lets robots learn to navigate the real world, is increasingly seen as essential infrastructure. That creates a new hardware demand curve that is just beginning.

The Competition Problem That Isn't

A common counterargument to Nvidia's dominance is that AMD, Intel, and a wave of custom silicon from Google, Amazon, Meta, and Apple will erode its margins. This is true in the long run. Custom silicon will capture a meaningful chunk of inference workloads, particularly for companies running their own models at scale. But that shift is slower than critics assume, because Nvidia's software ecosystem, CUDA chief among them, creates switching costs that are genuinely high. Rewriting optimised pipelines is expensive and time-consuming, and most companies have better things to do with their engineering talent.

The 70% growth figure is ambitious. If Nvidia hits it, the company will have achieved something almost without precedent for a business of its size. If it comes in at 50%, people will still call it a remarkable year. The structural tailwinds are real. The questions worth watching are margin pressure from custom silicon, any supply chain wobble in advanced packaging, and whether the inference transition plays out as quickly as Huang expects.

For now, the bear case on Nvidia keeps getting pushed further into the future, and that tells you something.