Nvidia (NASDAQ: NVDA) has been the best-performing stock over the past two decades. In late 2022, it got a major shot in the arm when ChatGPT was released, triggering a tsunami of capital expenditure on artificial intelligence (AI) infrastructure by cloud giants like Amazon Web Services, Microsoft‘s Azure, and Google Cloud.
Put simply, they feared being left behind in the age of AI unless they whipped out their fat chequebooks and bought as many top-shelf Nvidia GPUs as they could get their hands on. This has seen GPU king Nvidia’s revenue surge from $27bn in FY23 to an expected $196bn in FY26 (starting in February)!
The widespread assumption has been that this AI expenditure would reach trillions of dollars by the 2030s. Until recently, that is, when little-known Chinese firm DeepSeek came along with a competitive open-source AI model that was reportedly trained on a shoestring $6m budget.
This has raised serious doubts about AI models requiring ever-increasing amounts of computing power to train and run. Yet this is a big part of Nvidia’s bull story.
So, is it game over for the stock? Here’s my take.
Uncertainty
What I’ve found fascinating about the recent DeepSeek bombshell is how it has further entrenched investor’s already-formed views on Nvidia and the AI revolution. I believe psychologists call this ‘confirmation bias’.
In other words, Nvidia optimists simply see DeepSeek as yet another bullish signal. Cheaper models will lead to quicker widespread AI adoption, driving even more demand for data centres packed with Nvidia’s GPUs. They have a name for this: Jevons Paradox. This is the idea that as the cost of using a resource falls, demand will rise, not drop.
For Nvidia bears though, DeepSeek is the red-hot pin that is ready to pop the AI bubble, bringing Nvidia’s share price crashing back down to earth.
Where do I stand? Well, all this has just further increased my uncertainty. This is why I reluctantly sold my long-held Nvidia position last year.
I have four basic worries. First, it took less than two years for Nvidia to become a $3trn+ company. To sustain such a valuation, it will one day need to regularly generate hundreds of billions in annual profit. But it’s not at all clear to me that AI investments will always be anywhere near this high.
Second, there’s China, where Nvidia makes around 15% of revenue. I think there will be increasing restrictions on Nvidia’s GPU workarounds for Chinese customers. Indeed, after DeepSeek, I wouldn’t rule out an outright ban.
Third, AI basically falls into two categories: training and inference (e.g., ChatGPT answering a question). As models evolve, demand may shift more toward inference.
Will that reduce demand for GPUs? Does Nvidia still have the same competitive edge there? I don’t know. Perhaps Meta Platforms, Microsoft, and the rest can compete more easily in this area.
Game over?
That said, I don’t think it’s game over for Nvidia stock. Investors can essentially gauge near-term GPU demand by listening to the major cloud providers. And Microsoft has just earmarked $80bn for annual AI spend, while Meta is planning as much as $65bn in 2025. These are figures representing around 30% of annual revenue!
More near-term growth seems certain for Nvidia then. Longer term though, I’m still uncertain, so not tempted to reinvest.
