When it comes to which chip stock will outperform in 2026, there is no better existential battle than the one between Nvidia (NVDA +1.50%) and Broadcom (AVGO +2.62%). This is about whether Nvidia can maintain its dominance in the artificial intelligence (AI) infrastructure market with its graphics processing units (GPUs), or whether the market will begin to shift more toward custom AI ASICs (application-specific integrated circuits), like what happened in cryptocurrency mining.
Let's examine the case for each stock to see which will outperform in 2026.
The case for Nvidia
NASDAQ: NVDA
Key Data Points
Nvidia is currently the dominant player in the AI infrastructure market, with its GPUs providing the muscle needed to run AI workloads. This dominance can be seen in its extraordinary growth over the past few years. Last quarter, it grew its revenue an impressive 62% to $57 billion, while on a two-year basis, its revenue more than tripled, and on a three-year basis, it's up nearly tenfold.
The company has garnered an over 90% market share in the data center GPU space, in large part due to the ecosystem it's built around its chips. Its GPUs are the primary chips used to train large language models (LLMs), as most foundational AI code was written on its CUDA software platform. Meanwhile, the company's NVLink interconnect systems allow its chips to essentially act as one powerful unit, maximizing performance and discouraging customers from mixing and matching chips from other vendors within an AI cluster.
While ASICs have been gaining some ground, Nvidia's GPUs still have some big advantages. ASICs are pre-programmed chips, so they lack the flexibility of Nvidia's GPUs, which can be reprogrammed and have nearly two decades of AI libraries and tools built on top of CUDA to optimize their performance. In a fast-moving tech landscape, this is important. The company's chips are also readily available and can be used by any customer within any AI framework.
The case for Broadcom
NASDAQ: AVGO
Key Data Points
While Nvidia's GPUs dominate the AI infrastructure landscape, companies that own large data centers, called hyperscalers, have been increasingly working to lower their reliance on the chipmaker to lower costs. One solution they have been turning to is custom AI ASICs.
While ASICs lack the flexibility of GPUs, they generally consume less power and are more cost-efficient for the specific tasks for which they have been designed. This becomes even more important as the market turns toward inference, which is an ongoing cost. Meanwhile, the company that hyperscalers are increasingly tapping to help them design custom AI ASICs is Broadcom.
Broadcom helped Alphabet design its well-regarded Tensor Processing Units (TPUs), which are now viewed as one of the best alternative chips to Nvidia's GPUs. Given the success of TPUs, other hyperscalers have started to flock to Broadcom for help designing their own custom AI chips. Earlier this year, Broadcom said that its three furthest along AI ASIC customers were an over $60 billion opportunity in its fiscal 2027. Meanwhile, a fourth unnamed customer surprisingly placed a $10 billion order that will start being delivered in the second half of next year.
Perhaps most importantly, though, when OpenAI was making deals with chipmakers, it committed to deploying 10 gigawatts of custom AI chips from Broadcom to be deployed before the end of 2029. Based on Nvidia GPU pricing, that would be a $350 billion deal. Given that Broadcom is set to only generate total revenue of just over $63 billion this fiscal year, the AI chip opportunity in front of it is just enormous.
Image source: Getty Images.
The verdict
While I think Nvidia will remain the AI chip leader, I think Broadcom is the better stock to own in 2026. Its revenue base is much smaller, and the company looks poised to see explosive growth in the coming years from its ASICs business.
Both stocks should perform well next year, as AI data center spending continues to ramp up, but Broadcom just has the better growth prospects given its smaller size and the growing number of customers it is working with on custom AI chips.
