What Nvidia Tells Us About the AI Opportunity

Now that this wave of AI-revolution mania has pushed Nvidia's market cap past $1 trillion, Nvidia is no longer just the company that gaming nerds like me know from buying graphics cards. Its CEO, Jensen Huang, has become a bona fide celebrity — showing up on the evening news, and soon enough there'll be a flood of "Jensen Huang quotes" circulating, entrepreneurship wisdom and life lessons he may or may not have actually said. Even people with zero tech background — my retired-civil-servant mother, for instance — have started bringing up Jensen Huang to me.

But Jensen Huang has been showing up at Computex for years without fail, and Nvidia has been building toward this for a long time — it was already a hugely successful company. This wave of hype is better understood as the AI industry's influence finally taking a form ordinary people can grasp. Just as Steve Jobs embodied the smartphone revolution and Bill Gates embodied the personal-computer revolution, Jensen Huang and Nvidia have become the human face of this AI revolution.

From hardware to software: the electronics industry finds a new direction

At his keynote, Jensen Huang's main goal was to showcase a bigger push into selling more software and services — a bit like Amazon's playbook. He's already got the dragon-slaying sword that is the A100 GPU in hand; his underlying software services have long been a proven success, and now he's digging further into the application-layer software business. If that succeeds too, it's one more flywheel spinning; if it doesn't, it costs him nothing — the chip business and existing software services are unaffected either way. A lot of the applications he showcased already have corresponding software vendors behind them, and the application-software layer is inherently vast and fragmented by nature, so his applications business is really more of a complementary venture or a proof-of-concept than a core bet — but it's hard to see it failing badly either way, given he's still holding the dragon-slaying sword no matter what.

What has Taiwan's electronics industry so excited about Nvidia is that Nvidia used hardware to become the primary core supplier to software companies — which offers the electronics industry a template for imagining its own hardware-to-software pivot, because the two sides now understand each other in similar terms. In the past, the "software industry" that Taiwanese electronics firms understood mostly meant Google or Meta placing orders to manufacture servers or wearables. Pure software felt genuinely far removed. Some Taiwanese electronics firms have successfully invested in software ventures, but the scale simply doesn't compare to the world's major software and internet industries — and Taiwan's electronics success has never really needed a thriving software arm to succeed, the same way Silicon Valley companies in the Bay Area don't need serious manufacturing muscle to become unicorns.

So while it won't be easy, I still think it's a good thing that Nvidia is pushing Taiwan's electronics industry toward software. At the very least, ask any electronics-company boss today and none of them will tell you AI software doesn't matter. Future demand for AI is going to be huge and messy, and some companies are bound to come out of it with real results. Software talent is also, frankly, easier to grow than hardware talent — hardware is manufacturing, and iteration and experience take real time to accumulate. Jensen Huang himself has told the story of the early, brutal days when Nvidia had no money for production testing and had to bet everything on hardware simulators instead. Software, by comparison, can be iterated and rebuilt far more easily.

But if all the money's already been swept up by Nvidia selling A100s, and now it wants to sell you software services on top of that — what's left in the next wave of AI opportunity?

The answer is: everything Jensen Huang mentioned, and then some (which is to say, basically everything). Beyond chips, the main mainstream opportunity sits with software companies — look at the fact that the current number-one player in generative image AI, Midjourney, has just eleven employees. On top of that, the supply chains for automotive AI and marketing AI are anything but short, each packed with a range of different AI software vendors — and virtually every industry moving toward AI is generating a whole new wave of suppliers chasing demand.

Even once you've built your AI software, deployment needs might well be handled by yet another set of vendors entirely. That's why, watching interviews lately with the CEOs of publicly listed U.S. software companies, you can see them light up the moment analysts bring up AI — it's been a long time since there was a software narrative this big. Google, Meta, Salesforce and a handful of other major software companies have all risen anywhere from 40% to 100% between last year's low and this May — and that's just the big players.

Right now, the mainstream new early-stage AI-generative companies mostly cluster around a few directions: new training models; image and video generation and indexing; language-related generation and comprehension; and vector database models. What's notable is that these are all fairly foundational, tool-layer plays — Silicon Valley investors clearly believe the war has only just begun, and there's still plenty of room to invest in tooling-layer startups.

The application layer holds real opportunity — a decade-long transformation windfall

Beyond the foundational tools, the AI application layer right now is essentially boundless — anywhere humans want to be lazy, AI can be put to use. Nearly every major software company is transforming itself into an AI company, and that doesn't even count the small players. Pull up any international directory of AI tool startups and you'll find well over 3,500 companies — and that's just counting the ones with an actual visible product; it doesn't include whatever's brewing right now in someone's garage or lab. It adds up to a staggering explosion of software companies.

Going by past experience, the eventual number of winners among foundational software-tooling and generative-AI companies tends to be quite small — and hardware giants like Nvidia can still muscle into this space and offer tools at close to free, which makes competition brutal. I think the bigger opportunity going forward lies in application-layer companies, because that layer can absorb a far larger number of players — and once the early tooling settles, the application layer has at least a decade of transformation runway ahead of it. That's a full ten years of growth opportunity.

Right now, the compute layer — the frontmost piece of this — has already shown up in Nvidia's stock price. What follows, in order, should be the foundational-tools layer, then the application-tools layer, then digital transformation across every industry — and the whole ripple effect should keep playing out all the way to 2030. Investors can time their entry by tracking which industry's transformation timeline is coming up next.