Broadcom Stock Investors Just Got Good News From OpenAI -- Is Nvidia Losing Its Edge in AI Chips?

Semiconductor company Broadcom (NASDAQ: AVGO) has emerged as the second-largest supplier of artificial intelligence accelerators, albeit a distant second to Nvidia (NASDAQ: NVDA). But Broadcom recently announced a partnership with OpenAI that could bolster its market presence. Should Nvidia shareholders be worried?

Image source: Getty Images.

Where to invest $1,000 right now? Our analyst team just revealed what they believe are the 10 best stocks to buy right now, when you join Stock Advisor. See the stocks »

Broadcom recently announced a partnership with OpenAI

Broadcom develops a broad range of infrastructure software and semiconductor solutions. The company is particularly well known for its Ethernet switching and routing chips, where it has an 80% market share. Broadcom also has a 75% market share in high-end application-specific integrated circuits (ASICs), custom silicon built for specific workloads like artificial intelligence (AI).

In March, CEO Hock Tan said Broadcom was building custom AI chips for three hyperscale companies, as detailed below. He estimated annual revenue from those deals would range from $60 billion to $90 billion in fiscal 2027.

• Alphabet's Google works with Broadcom to build custom AI chips called TPUs (Tensor Processing Units).
• Meta Platforms works with Broadcom to build custom AI chips called MTIAs (Meta Training and Inference Accelerators).
• TikTok parent ByteDance works with Broadcom to build custom AI chips for video and networking tasks.

Hock Tan, on the most recent earnings call, said Broadcom was building custom AI accelerators for a fourth significant customer that had already placed orders totaling $10 billion. He did not name the company, but the Financial Times speculated it was OpenAI.

The companies confirmed that speculation with a press release on Oct. 13. Broadcom and OpenAI will co-develop 10 gigawatts of custom AI accelerators. Those chips will be assembled into racks connected with Broadcom networking solutions. Deployment will begin in the second half of 2026 and run through 2029.

Beyond those four customers, Broadcom has other ASIC prospects in the pipeline. Apple is reportedly working with the company to build custom AI accelerators code-named Baltra. The Information reports that those chips will be ready for mass production by 2026.

Broadcom stock looks expensive despite the recent deal with OpenAI

Broadcom, currently the leader in custom AI accelerators, is well positioned to benefit as more companies diversify beyond Nvidia GPUs. Indeed, Christopher Rolland at Susquehanna estimates Broadcom's share of the overall AI accelerator market will hit 14% by 2030, up from 6% today. Meanwhile, he expects Nvidia's share to fall to 67%, down from more than 80% today.

Broadcom is currently on pace to earn $13 billion in custom AI accelerator revenue in fiscal 2025, which ends in October. If it reaches the $60 billion to $90 billion target management quoted for fiscal 2027, the implied annual growth is 115% to 163% over the next two years. And those numbers only account for the first three hyperscale customers. They do not consider contributions from OpenAI or any other potential clients.

However, Broadcom trades at 91 times earnings, an expensive valuation for a company whose earnings are forecast to increase at 30% annually over the next three years. Those numbers give a price-to-earnings-to-growth (PEG) ratio above 3, and readings above 2 are generally considered overvalued.

Why Nvidia shareholders need not worry

Nvidia currently dominates the market for AI accelerators, but it faces competition from all angles. Notably, while Broadcom works with Alphabet's Google and Meta Platforms, its competitor Marvell Technologies helps Amazon and Microsoft build their own custom accelerators. Nevertheless, most industry observers believe Nvidia will maintain its dominance.

Morgan Stanley analysts in a recent note provided two reasons Nvidia is likely to retain its leadership in AI accelerators, both of which amount to the same thing: Nvidia systems have a lower total cost of ownership.

• While ASICs themselves may be cheaper, the adjacent technologies required to turn those chips into functional systems are generally more expensive. I am referring to components like memory, packaging, and interconnects.
• Nvidia has spent two decades creating an ecosystem of software development tools called CUDA, which helps programmers write AI applications for GPUs. No such ecosystem exists for ASICs, meaning the software tools must be built from scratch.

Morgan Stanley analyst Joseph Moore writes, "We have seen many threats to Nvidia come and go since 2018 -- something like a dozen start-ups, several efforts from merchant competitors such as Intel and AMD, and several custom designs. Most have come up short. Competing with Nvidia, a company that spends over $10 billion per year in R&D, is a difficult feat."

Nvidia currently trades at 54 times earnings, a reasonable valuation for a company whose earnings are forecast to increase at 36% annually over the next three years. Those numbers give a PEG ratio of 1.5, which means Broadcom stock is twice as expensive.

Trevor Jennewine has positions in Amazon and Nvidia. The Motley Fool has positions in and recommends Advanced Micro Devices, Alphabet, Amazon, Apple, Intel, Meta Platforms, Microsoft, and Nvidia. The Motley Fool recommends Broadcom and Marvell Technology and recommends the following options: long January 2026 $395 calls on Microsoft, short January 2026 $405 calls on Microsoft, and short November 2025 $21 puts on Intel. The Motley Fool has a disclosure policy.