OpenAI Is Building Its Own AI Chip With Broadcom. Should Nvidia Investors Be Worried?

Shares of chipmaker Nvidia (NASDAQ: NVDA) have been the purest way to bet on the artificial intelligence (AI) boom. So when one of its largest customers shows up with a chip of its own, investors pay attention.

On June 24, OpenAI and Broadcom (NASDAQ: AVGO) pulled the wrapper off Jalapeño, OpenAI's first custom AI chip. Designed by the maker of ChatGPT and built with Broadcom, it's the first piece of a sweeping plan the two unveiled last fall to deploy 10 gigawatts of OpenAI-designed accelerators between the second half of 2026 and the end of 2029. Headlines billed it as a direct strike at Nvidia.

So should Nvidia investors be worried? A measured yes -- but probably less than the headlines suggest.

Image source: Getty Images.

The custom-chip threat is real

Broadcom's custom-silicon business is booming, and the numbers are hard to ignore.

In its fiscal second quarter of 2026 (the period ended May 3, 2026), Broadcom's AI semiconductor revenue jumped 143% year over year to $10.8 billion. The pipeline behind that looks even bigger: management said the company booked more than $30 billion of AI orders during the quarter alone.

And the company isn't shy about where this is heading. Broadcom reaffirmed that its AI chip revenue should exceed $100 billion in fiscal 2027 -- about double what it expects this year.

"Demand for XPUs and networking is simply insatiable," said Broadcom CEO Hock Tan in the company's fiscal second-quarter earnings call (XPU is Broadcom's label for the custom accelerators it co-designs with customers).

And OpenAI isn't the only major customer on that list.

Broadcom also designs custom chips for a short roster of large buyers that includes Alphabet's Google and Meta Platforms. The build-versus-buy calculation is playing out across nearly every deep-pocketed buyer of AI computing -- each has reason to design silicon tuned to its own workloads and to lean less on Nvidia's pricey graphics processing units (GPUs). Over time, this could pressure a critical element of the bull case: Nvidia's pricing power, and the gross margin of about 75% it protects.

Why Nvidia is still in a different league

Yet for all that momentum, Nvidia plays on another scale entirely.

In its fiscal first quarter of 2027 (the period ended April 26, 2026), Nvidia's revenue rose 85% year over year to $81.6 billion, with data center revenue up 92% to $75.2 billion. In other words, Nvidia sells more AI hardware in a single quarter than Broadcom expects to generate from its AI chip business over an entire year.

And scale isn't the only gap.

Jalapeño is an application-specific chip built for inference -- the step where a trained model answers a query -- and was taped out in about nine months. Chips like it can be cheaper and more power-efficient for a narrow set of jobs, but they're far less flexible than Nvidia's general-purpose processors and don't support CUDA, the software layer that keeps developers anchored to Nvidia. A custom inference chip is also additive as much as it is competitive: it absorbs a slice of OpenAI's inference workload while training, and much of the rest still runs on Nvidia.

Then there's the size of the pie. Nvidia guided for fiscal second-quarter revenue of $91 billion, up from $81.6 billion -- a sign demand is still expanding faster than any one competitor can absorb. When the market is growing this fast, custom silicon and Nvidia's chips can both win. The question is how the spoils get divided over time, not whether Nvidia keeps growing.

That backdrop also makes the stock look less stretched. At about $194 as of this writing, Nvidia trades at about 22 times forward earnings -- well off the nosebleed multiples it carried a year ago, suggesting the market has already baked in some erosion of its dominance.

Broadcom, by contrast, fetches a forward price-to-earnings ratio in the low 30s. So the enthusiasm for custom silicon is hardly a secret.

So, should Nvidia investors be worried?

It's worth watching. Custom chips from Broadcom and its customers are notable: not only are they performant, but they're also scaling quickly and will likely chip away at Nvidia's pricing power as the AI build-out matures. But "chip away" is the important phrase to note here. Nvidia's lead in performance, software, and sheer scale is still enormous, and the spending wave is growing at a breakneck pace. The larger near-term risk to the stock is arguably simply a shift in AI sentiment, not a homegrown chip from a customer that still buys Nvidia hardware.

For now, Nvidia's moat is narrowing at the edges -- not closing.

Daniel Sparks and his clients have no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Alphabet, Broadcom, Meta Platforms, and Nvidia. The Motley Fool has a disclosure policy.