The artificial intelligence sector is continuing its sharp surge despite fears of a potential ‘bubble’, and Nvidia is leading the space with over $4.5 trillion market cap.
But, heading into 2026, one Wall Street analyst says AMD could finally overtake Nvidia as the better AI play next year.
In his latest piece with The Motley Fool, analyst Geoffrey Seiler argues that while Nvidia remains the heavyweight champion, AMD’s smaller base gives it far more room to run.
He says AMD’s “distant No. 2” status creates a scenario where the stock could surge on developments that would barely move the needle for its much larger rival.
Why this call matters now
The timing is critical. With the “AI infrastructure boom” entering its second phase, moving from training to broader inference deployment, investors are hunting for the next leg of growth.
The market is paying attention: AMD shares have shown resilience, signaling that Wall Street is finally buying into the idea that the chipmaker can capture double-digit market share in a sector Nvidia once monopolised.
Seiler’s argument isn’t that AMD will beat Nvidia on raw performance, but that it will beat it on stock returns thanks to some specific catalysts that hit in 2026.
The centerpiece of the bull case is the definitive agreement for OpenAI to deploy 6 gigawatts of AMD’s Instinct MI450 GPUs.
Seiler notes this deal alone validates AMD’s hardware at the highest level, creating a guaranteed revenue stream that anchors 2026 estimates.
The report highlights a critical, under-the-radar development: Microsoft’s release of new toolkits that seamlessly convert Nvidia’s CUDA models to AMD’s ROCm code.
This effectively lowers the switching cost for enterprise customers, making AMD’s cheaper chips a viable alternative for massive inference workloads.
Risk check: Why Nvidia stock still matters
Despite the optimism, betting against Nvidia remains a dangerous game.
The company still commands over 90% of the AI GPU market, and its “AI factory” approach, integrating H100 and Blackwell chips with proprietary networking, creates a defensive moat that is incredibly difficult to cross.
For training the largest frontier models, Nvidia remains the default choice. Hyperscalers may use AMD for inference (running models), but they are still buying Nvidia to build them.
AMD has a history of promising big and delivering late. Investors will be watching the supply chain closely; any delay in shipping the MI450 could derail the entire 2026 thesis.
If AMD’s software stack (ROCm) creates even minor friction for developers, the cost savings won’t matter. Nvidia’s premium multiple is partly a “peace of mind” tax that CTOs are happy to pay.
For 2026, the trade comes down to growth style: own Nvidia for steady, dominant compounding, but look to AMD if you want the high-beta play on a market share breakout.
The post Can AMD outperform Nvidia stock in 2026? This analyst isn’t doubting it appeared first on Invezz