Amazon: You Aren't Bullish Enough

Amazon is seen as having significant upside as Trainium AI chips may expand from internal use to external sales, adding to the company's $225B revenue commitment story. Rapid Bedrock growth and the TaaS model are cited as drivers of AWS revenue acceleration and margin expansion, with EPS expected to rise more than 20% over the next three years. The commentary argues Wall Street price targets are too conservative and likely to see upward revisions.

Analysis

The market is still treating AWS as a mature utility, but the mix shift here matters more than the headline growth. If Trainium expands externally, Amazon isn’t just selling cloud consumption; it is moving up the stack into a more captive compute ecosystem, which should improve gross margin durability and reduce customer churn versus pure software-layer AI offerings. That also creates a second-order benefit for AWS networking, storage, and inference-related spend, so the revenue opportunity is likely larger than the chip line item implies.

The competitive pressure is asymmetric. Microsoft and Google can match model access, but Amazon’s combination of lower-cost custom silicon and broader enterprise distribution can force price discipline across hyperscalers, especially in AI training workloads where capex intensity is becoming the gating factor. The likely loser is any competitor leaning on third-party accelerators or fragmented AI tooling; their economics get squeezed if Amazon proves it can undercut per-token or per-training-run economics while still expanding margins.

The setup is still underappreciated because consensus tends to linearize AWS growth, while this is a classic operating leverage inflection. A 20%+ EPS compounding path over three years can justify multiple expansion if investors start underwriting a higher terminal AWS margin and a faster capex payback cycle. The contrarian risk is execution: if Trainium adoption stalls outside internal workloads or AI demand normalizes faster than expected, the market will re-rate this as another incremental cloud story rather than a platform step-change.

Near term, the catalyst path is months, not days: product availability, enterprise design wins, and next few AWS prints will matter more than the article itself. The key downside risk is not demand collapse but capital intensity overshoot—if Amazon has to spend aggressively to win share, the EPS story can lag even if revenue accelerates. That makes this a quality-growth trade with asymmetric upside, but one that should be monitored for margin slippage and any evidence that external chip sales cannibalize higher-margin services.

Key Decisions for Investors

• Go long AMZN on any post-print consolidation; use a 3-6 month horizon and target upside from multiple expansion if AWS margin reacceleration becomes visible. Risk/reward favors owning the name into evidence of external Trainium traction rather than chasing after a clean breakout.
• Buy AMZN Jan-2027 calls or call spreads to express the 2-3 year EPS compounding thesis with defined downside. Best structure is a moderate-strike vertical to avoid overpaying for implied vol while keeping exposure to upward estimate revisions.
• Pair trade: long AMZN / short a basket of higher-cost AI infrastructure beneficiaries that rely on third-party accelerators. The trade works if Amazon’s silicon economics pressure industry pricing and shifts share toward integrated cloud platforms over the next 2-4 quarters.
• Add to AMZN only if the next AWS commentary shows external Trainium design wins rather than just internal deployment. If adoption remains internal-only, trim exposure on any rally because the market may already be pricing in a larger monetization runway than justified.