Marvell stands to gain from boosted Amazon-Anthropic partnership

Amazon's expanded partnership with Anthropic includes more than $100 billion of AWS spending over the next decade and up to 5 gigawatts of new AI capacity, a setup that analysts say benefits Marvell and Astera Labs. JPMorgan highlighted Marvell as AWS's long-standing ASIC design partner on Trainium, while RBC raised Marvell's price target to $170 from $115 and Astera Labs' target to $250 from $225. The deal should support demand for AI chips, optical connectivity, ethernet switches, and related data-center hardware into FY28 and beyond.

Analysis

The market is correctly reading this as a multi-year content expansion story, but the bigger implication is not just incremental chip revenue for MRVL — it is a higher utilization regime for Amazon’s AI stack that should steepen the attach rate of networking and optical silicon across every layer of the buildout. As Trainium nodes scale, the bottleneck shifts from compute alone to scale-up/scale-out interconnect, which tends to benefit the vendors with the most embedded design wins and the highest switching costs. That makes MRVL the cleaner second-order winner than AMZN itself on a near- to medium-term basis, because the economics of the build increasingly favor custom silicon plus high-speed connectivity over merchant GPU dependency.

The underappreciated upside is that this partnership likely pulls demand forward for a broader cohort of AI infrastructure suppliers, especially where bandwidth density and latency matter more than raw FLOPS. ALAB’s positioning is attractive because the market still tends to underwrite it as a “switching” story, while the real option value is in its exposure to repeated architectural refreshes as AWS moves through successive Trainium generations. If Trainium 4 meaningfully increases optical scale-up content, the mix shift could expand dollar content per rack faster than unit growth alone would imply.

The key risk is timing: the stock reaction can overshoot before revenue inflects, while much of the content benefit appears back-half weighted and may not fully hit the P&L for several quarters. There is also execution risk on AWS’s custom-chip ramp; if deployment timelines slip, the bullish narrative remains intact but gets pushed out, which is dangerous at elevated multiples. A secondary risk is that investors extrapolate this into a generic AI infrastructure melt-up, leaving the market vulnerable to disappointment if procurement concentration stays limited to one hyperscaler.

Consensus seems to be missing the asymmetry between direct revenue and ecosystem leverage: AMZN captures the strategic AI moat, but MRVL and ALAB capture the marginal dollars per incremental workload expansion. That makes the move potentially underdone for the suppliers and potentially overdone for the platform owner in relative-return terms. The best trade here is not simply long AI hardware, but long the names with the highest content gain per generation shift and the least dependence on headline model demand.

Key Decisions for Investors

• Long MRVL vs AMZN on a 1-3 month horizon: buy MRVL into any post-news consolidation and hedge with a smaller AMZN short; the setup favors the supplier with direct content leverage over the platform name, with upside tied to successive Trainium ramps rather than sentiment alone.
• Add ALAB on pullbacks for a 6-12 month trade: the market is likely underestimating switch attach as AWS scale-up intensifies; risk/reward improves if the stock de-rates back toward pre-news multiples while the order book remains intact.
• Consider a pairs trade: long MRVL / short a basket of broader semi names with weaker hyperscaler exposure over 2-4 quarters; this isolates AWS-specific content gains and reduces beta to generic AI-capex enthusiasm.
• Use call spreads in MRVL rather than outright equity for the next 2-3 quarters: elevated pricing already reflects part of the news, so defined-risk upside participation is preferable unless you expect an additional catalyst on Trainium 4 disclosures.
• Set a tactical take-profit level on MRVL after a further 10-15% rally unless there is fresh evidence of revenue acceleration; the near-term risk is multiple expansion outrunning the actual earnings bridge.