Huang’s Computex slide sparks hunt for next Marvell

At Computex, NVIDIA CEO Jensen Huang displayed a slide mapping an AI-factory ecosystem and called Marvell “the next trillion-dollar company.” Marvell shares climbed from below $100 in April to roughly $300, a gain of about 241% year to date with an all-time high near $316.

Huang divided NVIDIA’s offering into three layers: RTX as the GPU “worker,” DGX as the system-level “team,” and DSX as the full “factory” that combines racks, power, cooling, networking and software. He described DSX as “a blueprint, a reference design for building and operating AI factories at maximum efficiency and profitability,” and named components including DSXSim for digital-twin planning, DSX OS to run operations, and DSX Max LPS and DSX FLEX for packing more GPUs into constrained power budgets and flexing with the grid.

NVIDIA said DSX aims to speed deployment and lower operating costs for large AI installations. The company projected up to 100 gigawatts of AI factory capacity online by 2030 and estimated a one-gigawatt AI factory can cost between $30 billion and $100 billion. The slide included a roster of partners NVIDIA expects to work with to design, build and run these facilities.

The map lists companies across design and simulation, data-center equipment, power and cooling, construction, utilities and operational software. NVIDIA published the reference designs and assembled an AI Factory Ecosystem so partners can coordinate civil works, high-voltage lines, chillers and other components required to deploy multi-gigawatt sites.

Market response was immediate. Marvell’s stock, which traded in a $50–$100 range in April, moved into the high $200s and low $300s year to date. Investors have been reviewing the slide to identify other public companies that appear on the DSX map.

Analysts say appearing on the slide does not equal material revenue exposure. Some firms on the map, such as CoreWeave and Vertiv, derive a substantial portion of revenue from AI infrastructure. For other listed companies, including heavy-equipment and utility firms, revenue from AI data-center builds represents a smaller share of total business. The slide does not quantify how much demand or pricing power individual suppliers will capture.

NVIDIA presented the slide as a reference for how large-scale AI facilities are designed and delivered. Industry participants and investors are using it to assess which suppliers may participate in the projected build-out of gigawatt-scale AI capacity.