Herman Jin
Formerly in Asia FICC at Goldman Sachs, Macro & Semiconductor Investing
Astera Labs may be a cleaner way to play the shortage in AI connectivity components than Marvell. Marvell's story is still a narrative whose earnings payoff is far off, leaving it vulnerable to sell-offs in between. The Google-Marvell arrangement that some see as a threat to Astera Labs is a 2028 issue, not an immediate one. x.com x.comÂ
The market is lagging badly on CXL demand for Astera Labs and Marvell. The next two quarters for both companies will show just how demanding the industry has already become. x.comÂ
Lumentum's 109% year-on-year growth and sold-out capacity through 2029 make selling its AI optics exposure questionable. Bloomberg reported Marvell is in advanced talks to buy it; the appeal depends on the price, especially when holders of an $80 billion Lumentum would become holders of a $320 billion-plus Marvell growing 37%. x.comÂ
Astera Labs looks relatively expensive because similarly fast-growing AI infrastructure names trade at much lower multiples. Broadcom and Nvidia are at 14x, Credo at 15x and Lumentum at 18x; comparable peers look 50-80% cheaper. Astera is a great company and not necessarily expensive in absolute terms. Its roughly 2x multiple over Credo may persist, but the setup looks harder; with no valuation support, it could fall 50% if weakness continues. x.com x.com x.com x.comÂ
The Lumentum merger makes Marvell's path toward a $1 trillion valuation look clearer if it executes well. Jensen Huang had called Marvell the next $1 trillion company. Lumentum's demand is sold for the next two to three years, though whether that gives Marvell an edge remains an open question. Marvell looks like a long-term opportunity, but the route to building a position is still unclear. x.comÂ
River Eletec's speculative AI/HPC-adjacent technology looks worth more than its $38 million market cap. Its low-jitter 1.6T product may be sampled by Marvell, though that is a guess; past earnings have been disastrous. x.com x.comÂ
Astera Labs carries a huge premium for flawless execution against a CXL market that is small near term. Switch-based CXL memory pooling capturing $15-16B in 2026-2028 is low probability: Marvell guides its XConn switch business to only about $100M by fiscal 2028, PCIe included. Pooling software likely stays immature into 2028, per a Samsung white paper, and Nvidia is contesting the warm KV-cache tier. Astera does CXL best, but SK hynix and Micron offer better risk-adjusted returns. x.com x.comÂ
Marvell's end-to-end networking approach is well suited to heterogeneous GPU, CPU and ASIC compute and mixed optical/copper technologies. Its shift from AI loser to winner deserves a lot of credit: its old custom-ASIC focus had lost out to Broadcom and Alchip, but it is now back in ASICs as Google's TPU v10 partner. The stock's excellent run reflects how it is now perceived. x.comÂ
The market may still be underestimating Marvell as it becomes a key supplier of custom silicon and connectivity for the AI build-out. At its Investor Day, Marvell put its addressable market at about $400 billion by 2030, targeted about $20B of FY28 revenue and said revenue could reach $70-90B by FY31. Last quarter revenue rose 37% to $2.74B on AI and data-center demand, and Nvidia's Jensen Huang reportedly called Marvell the next trillion-dollar company. x.com x.comÂ
Astera Labs' valuation has gotten nuts given its medium- and long-term competitive and technology risks, especially next to NVIDIA and Broadcom. The aixPhotonics acquisition was smart, but even if those optical offerings ramp well, they are unlikely to offset the long-term cannibalization of retimers, AECs and similar products by NPO and CPO. x.comÂ
Astera Labs' chip connects every GPU in AI servers, positioning it to ride a data-center buildout that is just getting started. Quarterly revenue reached $392 million in Q2 2026, versus $65 million in Q1 2024. x.comÂ
