Boy, have I been wrong about Nvidia (NASDAQ: NVDA) stock. When the chipmaker crashed nearly 70% in 2022, I didn’t invest, fearing that the so-called ‘everything bubble’ was bursting. I repeated my mistake when the launch of a slim-downed generative AI model by Chinese startup DeepSeek wiped $600bn off Nvidia’s market cap in just one day. Now valued at $4.5trn, is it time for me to throw the towel in and join the party?
Spending splurge
The fortunes of the chip manufacturer are inextricably linked to that of the hyperscalers. One might say they are joined at the hip.
Capital spending among the top five hyperscalers — Amazon Web Services, Alphabet, Microsoft, Meta, and Oracle — show no signs of abating. In 2024, spending by these firms building out their AI infrastructure reached over $200bn.
Back in the dotcom bubble, it was a similar story. In the late 1990s a mad rush of corporate spending to fend off the Y2K software bug, resulted in a surge in corporate profits for the big tech companies of the day.
More recently, in the wake of the work-from-home mandates, companies were forced to pull forward technology spending and accelerate a move to the cloud, in order to support remote working.
The only difference between these two previous peak spend cycles, was the length of the subsequent downturn. In 2000, tech stocks remained in the doldrums for a decade. In 2022, barely a year.
Return on investment
The big unknown is when will spending peak in this cycle. It’s certainly gone on a lot longer than I envisaged, that’s for sure. But the issues are mounting.
Despite ongoing investment, the hyperscalers are still to see a return on their huge investments. For me, the industry is big on promises but not so much on delivering ground-breaking, tangible innovations.
As reasoning models become increasingly more sophisticated, so too are their power demands. Huge upfront investments in data centre expansion are adding new layers of risk for the hyperscalers.
Society is growing increasingly concerned about the impact the proliferation of data centres is having on the natural environment. As huge swathes of land get swallowed up, a major public backlash against the industry could be looming.
Bigger than the internet
The jury remains out on whether the opportunity presented by generative AI will be bigger than the internet. What is becoming clear, though, is that the path the industry is on bears little resemblance to back then.
The internet was a great leveller. No longer did you need to be a global multi-national company to market your innovation. Instead, you could upload it to a few key platforms and reach a worldwide audience. That’s exactly how YouTube and Android started.
Today, however, the entire industry is becoming more and more centralised as the hyperscalers swallow up early-stage startups and entice top talent with pay awards as large as $1bn. Over the long-term I believe such a strategy will backfire as creativity and innovation dry up.
I’ve been wrong about Nvidia for so long, that I’m beginning to sound like a broken record. But with a price-to-earnings of 58 and nosebleed valuations across the rest of the Magnificent 7, nothing will entice me to buy in now. Never forget, bubbles only ever become obvious once they’ve burst.
