One of Nvidia's Largest Clients Has Just Made a Major Agreement With Its Strongest Competitor

Nvidia ( NVDA 0.43%) has arguably been the leading winner from the surge in artificial intelligence advancements over the past three years. The company now boasts a market capitalization exceeding $4.3 trillion, making it the most highly valued business globally. This surge is primarily fueled by ongoing robust demand for its top-tier graphics processing units (GPUs), which are crucial for training and running large language models.

Nvidia’s financial records highlight a giant that continues to post rapid growth in both revenue and profits. However, some recent events have cast uncertainty over whether Nvidia can keep up this fast-paced growth. Notably, one of its largest clients recently struck a deal with a major Nvidia competitor.

The major risk facing Nvidia

While Nvidia’s sales have soared as major technology firms spend heavily to remain leaders in AI innovation, this has resulted in a significant portion of Nvidia’s income coming from just a few clients. According to Nvidia, two direct customers accounted for 39% of its revenue in the second quarter, while six customers made up 85% of sales.

Of course, only a limited number of companies have the resources to invest in building large-scale data centers. Many such businesses, including Microsoft, serve numerous enterprise clients who access Nvidia chips via their cloud services. Therefore, Nvidia’s customer concentration risk is somewhat mitigated, but it remains a notable factor for investors to keep in mind.

This issue has become more prominent recently because OpenAI, a leading user of Nvidia hardware, is said to be developing its own AI accelerator chip in partnership with Broadcom ( AVGO 0.19%). The timing coincides with Broadcom’s announcement that it has secured a new qualified custom AI chip customer and already received $10 billion in orders.

If OpenAI does turn out to be Broadcom’s new client, it signals a significant change for the company. As the leading force behind The Stargate Project, OpenAI is planning to invest $500 billion from 2025 to 2028 in AI infrastructure. The project’s announcement mentioned Nvidia among its technology partners, but notably did not include Broadcom.

The shift is already underway

OpenAI may not be alone in its move toward more specialized silicon and less dependence on Nvidia. According to Broadcom CEO Hock Tan, the company continues to expand its presence with three other custom AI chip customers, specifically Meta Platforms, Alphabet, and ByteDance. Microsoft, in addition, is reportedly preparing to greatly increase its use of its next-generation custom accelerator, which could result in $10 billion to $12 billion in orders in 2027, based on Fubon Research’s findings.

This follows a recent agreement in which OpenAI will use Alphabet’s Google Cloud and its Tensor Processing Units (TPUs), which Broadcom also helped design. Google published research indicating its average energy use for a text AI prompt is just 0.24 watt-hours—described as “considerably lower than many public estimates.” The company also pointed to a 33-fold efficiency improvement over the last year. Google credited much of this efficiency to its TPU chip. Given OpenAI’s massive scale, shifting away from Nvidia’s infrastructure could lead to substantial cost savings.

For investors, the key question is whether the market has already reflected the potential market share Nvidia may lose and the possible gains for competing chipmakers. Nvidia’s stock trades at about 38 times projected earnings, still above its average since early 2023. This is despite the expectation that Nvidia’s earnings growth won’t repeat the extraordinary momentum of the previous two years.

However, Broadcom’s stock isn’t a bargain either. The share price has surged to over 50 times earnings following news of a major contract win. While custom AI chips are becoming increasingly important for Broadcom’s revenue and profits, it’s worth noting that currently only about half of its semiconductor sales relate to AI, and it also operates a mature software division. Over time, AI-related spending will weigh more heavily on Broadcom’s overall results, but for now, the market prices it as if all of its revenue is AI-driven.

At this point, both chipmakers appear pricey. If forced to choose, Broadcom seems to offer more potential upside than Nvidia despite the current valuations. Nevertheless, there are several other AI-related stocks with more attractive pricing for investors to explore.