Amazon Is in Talks to Sell Nvidia Rival Chips to Other Firms

Amazon is in talks to sell its custom AI chips to third-party data center operators, potentially expanding a business designed to challenge Nvidia’s dominance in AI accelerators. The company has already begun discussions, though no customers were named. The move signals a broader commercialization opportunity for Amazon’s chip strategy and could modestly affect sentiment around both Amazon and the AI hardware competitive landscape.

Analysis

This is less about near-term chip revenue and more about Amazon turning AWS silicon into a strategic wedge against the most profitable layer of the AI stack. If Amazon can prove third-party deployment at scale, the real economic prize is not just chip ASPs but leverage over enterprise AI procurement, with downstream pressure on NVIDIA’s networking, software attach, and pricing power. The second-order winner is anyone exposed to alternative inference economics: large cloud buyers, model operators with heavy token volumes, and semiconductor supply chain vendors that can ride a multi-sourced AI capex cycle.

The market may be underestimating how long this takes to matter financially. Even if discussions turn into pilot deployments quickly, design wins in data-center silicon usually translate into meaningful revenue over 12–24 months, not weeks. The nearer-term effect is sentiment: every credible alternative to NVIDIA tightens the multiple dispersion in AI hardware, and that can compress NVDA on any sign of slower order growth or margin moderation even before Amazon’s chips become material.

The main contrarian risk is that Amazon is not necessarily trying to “beat” NVIDIA so much as improve its own negotiating position. If customers use Amazon chips as a bargaining chip rather than a full replacement, NVDA may lose pricing power at the margin without losing the volume franchise, which is harder to handicap and usually less dramatic in the stock. Also, if third-party adoption exposes software compatibility or performance gaps, the narrative can reverse quickly; the gatekeeper here is ecosystem confidence, not raw chip performance.

Key Decisions for Investors

• Maintain a tactical short NVDA vs long AMZN pair for 1-3 months: the asymmetric read-through is toward margin-pressure risk for NVDA and strategic optionality for AMZN; stop if NVDA re-accelerates on hyperscaler capex commentary.
• Sell out-of-the-money NVDA calls into strength for the next 4-8 weeks: implied vol should remain bid on AI supply-chain headlines, but upside may be capped if investors begin pricing in multi-source chip competition.
• Add to AMZN on pullbacks over a 3-6 month horizon: this is an optionality trade on silicon monetization and bargaining power, with limited fundamental downside if the external chip effort stalls.
• Consider a long semi-equipment basket vs NVDA if the market starts rotating from monopolistic AI hardware into diversified buildout: competition usually benefits tools and fab-enablers before it hurts the broader capex complex.