Why the Hyperscaler Spending Spree Might Put a Big Cushion Under Nvidia and AMD

There’s been much concern about what could happen to the semiconductor trade if one of the hyperscalers were to suddenly hit the brakes on AI-related CapEx. With Alphabet (NASDAQ:GOOG | GOOG Price Prediction) reporting earnings and more heated quarterly CapEx numbers, it certainly seems like the firm’s pace of spending could run towards a $200 billion per year cadence. Of course, time will tell how long these hyperscalers will continue raising the bar on spending if it means getting a muted reception to strong results and some earlier signs that spending is actually translating to earnings growth.

Any way you look at it, it feels like the major tech titans have already passed the event horizon that was getting into the AI race to begin with.

Now, it’s spending furiously to maintain that competitive advantage and to get all that early infrastructure built before a rival, domestic or foreign, has a chance to gain a leg up. It feels like any given quarter that the hyperscalers post will see strong cloud growth alongside commentary about how constraints held back what could have been. Indeed, that seemed to be the case for Google as well when it clocked in results after the close on Wednesday.

While there are serious risks of overspending on AI, it certainly feels like the hyperscalers have large enough cash cushions to absorb the blow far better than most other firms spending heavily on the effort that are leaning heavily on cap raises or excessive amounts of debt. In a way, the hyperscalers are stepping into the ring with some robust headgear while most others might be going without.

Nvidia and Advanced Micro Devices are holding their ground well and for good reason

Any way you look at it, it feels like we’re still a long way off from getting AI infrastructure to where it needs to be as firms scale aggressively. In due time, though, the chip wars and massive year-over-year efficiency gains could be the needle mover that helps get compute where it needs to be without breaking the bank. For Nvidia (NASDAQ:NVDA) and Advanced Micro Devices (NASDAQ:AMD), the slate of next-generation hardware is delivering on those enormous efficiency gains, and the big spenders are buying.

With AI demand continuing to overwhelm, the blame for bottlenecks is shifting to the fabs, which themselves are constrained. In any case, the order backlog provides clarity into the future of earnings, but beyond that, it feels like investors expect substantial sales and margin decay, as some look for demand to wind down gradually.

Given the constraints standing in the way of the great AI data center buildout, though, it feels like that cyclical downturn might still be far off.

It’s hard to imagine that the biggest cyclical upswing isn’t yet in the cards, but when you look at the data center buildout and the compute needs to power next-generation AI applications at scale, only then does it become apparent that the latest sell-off in semiconductors might have more to do with investor nerves than anything that’s changed regarding the state of the buildout.

In my view, bottlenecks across the board could stretch out the buildout over some number of years. Add Jevons Paradox into the equation, which cites more usage when efficiencies rise, and it’s hard to bet against semiconductors as their share prices take a big dive.

Betting on a hyperscaler CapEx drop might prove unwise

For Advanced Micro Devices and Nvidia, the hyperscaler CapEx is in the books and there are high hopes for more business as the buildout accelerates.

The hundreds of billions of CapEx could evolve into over $1 trillion as bottlenecks alleviate across the board; perhaps there’s potential for a more vicious cyclical upswing. Until then, though, Advanced Micro Devices and Nvidia are not wasting time as they look to meet sky-high demand while raising the bar ever higher and expanding the slice of the pie.

As both GPU titans start selling massive numbers of racks to hyperscalers while doubling down on software innovations, it feels like there’s a fat cushion being put underneath the GPU titans. Of course, that’s not to say that a sound financial cushion will cushion the stock, especially when fear takes control of the market.

A cheap stock, like Nvidia at just over 23.0 times forward price-to-earnings (P/E), can always get cheaper. If everyone doubts the firm’s future earnings potential, perhaps a single-digit P/E is justifiable to some. In any case, until hyperscalers start posting quarterly spend that suggests a slowdown rather than a speedup, Advanced Micro Devices and Nvidia might have more support than most other players in the sea of semis.

In my view, the custom silicon threat might be what thins the cushion that supports the GPU titans. Not lower CapEx overall, but less that goes into the hands of the third-party chipmakers.

Contact [email protected] for any questions or corrections.