AMD has emerged as a stock market winner in 2026 despite being ranked last in a year-start projection, while Nvidia has lagged expectations and Broadcom has performed as anticipated. The three compete in AI infrastructure differently: AMD and Nvidia offer general-purpose GPUs, while Broadcom partners directly with hyperscalers like Alphabet, Meta, Anthropic, and OpenAI to design custom AI chips with better cost-performance. Broadcom's custom chips are scheduled to ramp production late 2026 and into 2027, potentially becoming popular as companies optimize AI spending, though Nvidia's entrenched market position and GPU flexibility give it lasting structural advantages.
Summaries like this, in your inbox every morning.
Sign up free →What happened
AMD has outperformed significantly in 2026 so far, reversing an early-year prediction that ranked it last among AMD, Broadcom, and Nvidia. Broadcom has performed as expected in the middle, while Nvidia has underperformed. The three companies compete in AI infrastructure but take different approaches: AMD and Nvidia both offer GPUs (graphics processing units for complex computing tasks), while Broadcom designs custom AI chips in partnership with hyperscalers like Alphabet, Meta, Anthropic, and OpenAI, with production ramping up late this year and into next.
Why it matters
Broadcom's custom chip strategy may create a cost-performance advantage over GPU-based computing, potentially making it increasingly popular with major AI companies seeking to optimize spending. However, Nvidia maintains a commanding lead in the data center space and many firms have already built ecosystems around its products, giving it structural advantages. AMD faces an uphill battle to gain traction against entrenched Nvidia ecosystems. For investors, this suggests the AI infrastructure market is diversifying beyond pure GPU dominance.
What to watch
Production timelines for Broadcom's custom chips are scheduled to ramp up late 2026 and into 2027, which will be a key indicator of whether the custom chip strategy gains meaningful traction against GPU-based approaches. Nvidia and Broadcom are viewed as tied in terms of business approach potential, while AMD remains structurally disadvantaged in competing for data center share.
The article presents a reshuffled competitive landscape in AI infrastructure stocks within a single year. The author's early-2026 ranking—Nvidia first, Broadcom second, AMD third—has been upended by actual performance, with AMD becoming a major winner and Nvidia a loser, though the timeframe for this reversal is compressed to just the start of 2026. The fundamental driver is the divergence in business strategies: Nvidia and AMD compete directly with GPU-based solutions, while Broadcom has carved a different niche by working directly with hyperscalers to create bespoke silicon.
Broadcom's custom chip partnerships with Alphabet (via TPUs), Meta, Anthropic, and OpenAI represent a deliberate shift away from general-purpose computing toward optimization for specific workloads. The article notes that production ramp schedules for Meta, Anthropic, and OpenAI chips extend into 2027, suggesting that much of Broadcom's growth narrative remains forward-looking. The author's logic is that custom chips will likely gain popularity as AI hyperscalers prioritize spending efficiency, but acknowledges that GPU flexibility keeps many companies invested in that approach. For Nvidia, the commanding lead and entrenched ecosystem provide structural moats; for AMD, the same ecosystem advantage works in the opposite direction, making it harder to displace Nvidia in data centers. The framing positions the market not as a zero-sum GPU battle but as a segmenting market where custom chips and GPUs can coexist, with Broadcom and Nvidia both positioned to benefit.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
