Marvell’s enormous targets, Lisa Su’s trip to South Korea… Is Big Tech’s move away from Nvidia getting underway in earnest?

The New York editor presents Marvell’s New York Investor Day and AMD CEO Lisa Su’s trips to Taiwan and South Korea together as signals that Big Tech is moving away from Nvidia. The editor reports that Marvell raised its fiscal 2028 revenue target from 180 hundred-million to 200 hundred-million dollars, presented an exceptionally long-term target of 700~900 hundred-million dollars for 2031, and saw its stock rise 6% in a single day. The editor says Marvell cited supplies to the world’s 4 major hyperscalers, the possibility of a long-term contract with Google worth up to 1200 hundred-million dollars, and a target of more than 120 hundred-million dollars in custom chip revenue in 2029 as grounds for its confidence. The editor reports that, emphasizing silicon photonics that overcomes the copper wall and optical interconnects in the 3.2 tera class, Marvell conveyed that whichever chips customers use, they inevitably have to rely on Marvell’s connectivity technology. Lisa Su reportedly announced in Taipei that supply expansion would continue in 2026 and increase substantially in 2027, met Foxconn and TSMC, and then headed straight to South Korea to attend to the Samsung and SK hynix HBM supply chain. Using an analogy involving clothing and luxury goods, the editor explains Big Tech’s two-track strategy for reducing dependence on Nvidia: custom ASIC chips for repetitive computations from Marvell, and merchant GPUs from AMD as general-purpose alternatives. The editor concludes by identifying execution risk for a company with annual revenue of 82 hundred-million dollars, margin pressure as the custom business grows, AMD’s lack of specific guidance and the CUDA barrier, and power-grid and cost constraints as risks, while naming the share of in-house technology within capital expenditure, HBM contracts, and silicon photonics delivery timing as points to watch. All directional outlooks are the speaker’s conditional judgments, not definitive short-term price predictions.