Andy Jassy Says This Could Be a $50 Billion Business for Amazon

Amazon CEO Andy Jassy said demand for its AI chips (Trainium) is “quite possible” to reach a ~$50B annual cash run rate from potential third-party rack sales. While Amazon’s revenue was $717B last year, the market narrative is that this AI-chip path could be a new growth catalyst as the stock has lagged (+~10% vs. the S&P 500’s ~20% over 12 months). The article also flags valuation support with AMZN trading at ~30x P/E versus ~38x for the Technology Select Sector index ETF.

Analysis

This is more important as a margin story than a new revenue stream. If AMZN can substitute cheaper in-house silicon for third-party accelerators inside AWS, the first-order payoff is not chip revenue — it is higher AWS gross margin, better price/perf for customers, and a longer reinvestment runway into AI infrastructure. The market is still valuing AMZN like a mature megacap platform; sustained evidence of silicon-led operating leverage is the path to multiple expansion.

NVDA is the obvious second-order pressure point, but the risk is not an immediate share loss event. The real threat is that hyperscalers become more disciplined buyers and reserve NVIDIA for frontier training while shifting more inference and cost-sensitive workloads to custom ASICs, which can slow pricing power at the margin. That dynamic would also encourage the other cloud names to accelerate internal chip programs, widening the moat for whichever platform can manufacture AI at the lowest effective cost.

The catalyst path is probably 1-3 quarters, not days: a disclosed external customer, improving AWS margin data, or commentary that Trainium utilization is tightening supply. The thesis breaks if AWS capex rises faster than monetization, or if NVIDIA’s next-gen platform keeps the performance gap wide enough that Trainium remains a niche optimization tool. Over 6-18 months, the question is whether AMZN’s AI stack becomes a flywheel for FCF growth rather than another capital-intensive promise.

Contrarian view: the consensus may be overestimating how quickly this becomes a material standalone business and underestimating the balance-sheet benefit of keeping AI economics inside AWS. The stock does not need a $50B external chip franchise to work; even a modest reduction in compute costs can lift AWS economics enough to justify a re-rate.

Key Decisions for Investors

• Long AMZN on pullbacks over the next 1-3 months; thesis is multiple expansion via AWS margin leverage, not chip revenue. Risk/reward improves if the market stays focused on headline revenue rather than operating leverage.
• Relative-value pair: long AMZN / short NVDA in small size as a hedge on hyperscaler custom-silicon adoption. Use only if NVDA valuation re-accelerates ahead of evidence that merchant GPUs retain full pricing power.