Should Nvidia investors be concerned about this statement from Broadcom’s CEO?

There is no doubt Nvidia (NVDA 3.54%) was the big winner of the first stage of the AI ​​revolution. Following OpenAI ChatGPT’s November 2022 debut, virtually every enterprise and cloud giant has clamored for its market-leading GPUs to power generative AI.

Of course, as the absolute leader, Nvidia has managed to make a pretty penny for its chips, as evidenced by its skyrocketing revenues and margins over the past 18 months.

But in reaction, most cloud computing giants have started investing in their own custom accelerators, which are partly co-designed by ASIC companies. Broadcom (AVGO 2.79%) and Marvell Technologies.

Could the rise of custom ASICs kill demand for Nvidia? In his recent earnings release, Broadcom CEO Hock Tan offered words of caution.

“I flipped my vision in the last quarter”

Last week, Broadcom released its fiscal third quarter earnings. Broadcom beat analysts’ expectations, although its guidance came in a bit light.

As the name suggests, Broadcom has broad exposure to the chip industry, and many of its non-AI businesses are still mired in a two-year recession. However, Broadcom’s AI businesses through custom ASICs, networking and optical interconnect are all booming. Regarding custom ASICs, CEO Hock Tan noted on the conference call with analysts that Broadcom AI ASIC revenue has grown 3.5 times over the past year.

Despite having this ASIC business, Tan used to say that the “merchant” market, or rather neutral, third-party chipmakers like Nvidia will eventually win the day in the AI ​​accelerator market, in line with history semiconductors.

But when asked by an analyst last Thursday, Tan remarked: “I have changed my view. And we did that, by the way, in the last quarter, maybe even six months ago.”

Tan now says he believes the cloud giants will increasingly train large language models on their own custom silicon ASICs because of the cost advantages of making their own chips, as well as the added benefit of taking control from Nvidia .

Those GPUs are much more — or XPUs are much more important. And if that’s the case, what better thing to do than to take control of your own destiny by creating your own custom silicon accelerators? And that’s what I see everyone doing. It’s just that they do it at different rates and start at different times. But they’ve all started… they’re all going to go in that direction totally in ASIC or as we call them, XPU, custom silicon.

So what could this mean for Nvidia?

On the analyst call, Nvidia revealed that 45 percent of its revenue currently comes from hyper-scalar cloud, and more than 50 percent comes from large Internet companies — which means likely. Meta platforms — and other businesses.

When we look at the 45% of revenue from cloud operators, and then additional revenue from Meta, which also develops its own custom accelerators with Broadcom, that’s a good portion of Nvidia’s revenue that goes to customers looking for currently internal customers. alternatives.

Image source: Getty Images.

Now, Nvidia investors shouldn’t panic that all of its cloud demand will go to zero. It seems that we are still in the early days of AI development, and many cloud customers still like to use the latest and greatest Nvidia chips, their developers familiar with Nvidia’s CUDA software. Some cloud vendors have been making custom ASICs for years, but they still clearly buy a lot of Nvidia GPUs as well. Nvidia’s new Blackwell chip is slated to launch around the end of the year, which will provide a step-up in performance over the current Hopper line.

So there will still be demand for Nvidia chips, even from cloud providers.

But will Nvidia’s margins and multiples suffer?

While there will continue to be demand for Nvidia’s offerings, Nvidia likely won’t retain the 90%+ AI chip market it has amassed to date. It’s also possible that cloud companies will demand lower prices in the future, with perhaps more bargaining power, as they ramp up their custom ASIC offerings and offer them to customers at much lower prices.

One of the reasons Nvidia may have fallen after its recent earnings report is that its gross margins ticked off record highs, perhaps suggesting limits to the company’s pricing power, which seemed limitless throughout 2023 and early this year. year.

After this pullback, Nvidia stock is trading at 48 times trailing and 36 times forward estimates. So it’s not as expensive as it used to be, and it doesn’t have a weird rating for its quality.

But it’s still not “cheap” stock, which makes Nvidia vulnerable to any failure or crash. So any slowdown, whether caused by ASIC competition or another factor affecting the AI ​​market more generally, could limit the stock’s gains.