Not Just Marvell! Jensen Huang Names 30+ "AI Factory" Stocks — Top Gainer Already Up 480% This Year
CoreWeave CRWV | 0.00 | |
NEBIUS NBIS | 0.00 | |
IBM Corp IBM | 0.00 | |
Cadence Design Systems, Inc. CDNS | 0.00 | |
Procore PCOR | 0.00 |
Since the beginning of the year, the AI wave has swept global markets. At the GTC conference, Jensen Huang's new narrative catalyzed a surge in related concept stocks—Applied Optoelectronics up 480%, Marvell up 243%, Nebius up 211%. Now, he is introducing the next chapter.
At the latest GTC in Taipei, Huang declared that the era of Agentic AI has officially arrived. Unlike generative AI, agentic AI understands human intent and autonomously completes tasks through hundreds or thousands of reasoning steps—each step consuming tokens.
"Today, we are building AI factories, not traditional data centers. Give it energy, and it produces valuable products called tokens," Huang said. Tokens, as the core profit unit for AI companies, are driving global computing demand to skyrocket.
From Selling Shovels to Building AI Factories
A 1GW AI factory costs $20-30 billion to start, potentially approaching $100 billion. Beyond cost, these projects face complex engineering challenges: chips, racks, networking, power, cooling, and grids must be designed end-to-end. Currently, AI factories are over-allocated for power by up to 40%, stranding massive computing resources.
NVIDIA's answer is the DSX platform—a full "factory-building suite" covering hardware, software, and infrastructure. Using Omniverse digital twins, partners validate designs before the first rack is built. DSX OS handles operations and maintenance. DSX MaxLPS enables liquid cooling and dynamic power distribution, deploying more GPUs within the same power budget.
NVIDIA is no longer just selling shovels; it is selling the blueprints and project management for entire AI factories—transforming from a chip supplier into the infrastructure hegemon of the AI era.
The DSX Ecosystem: Nearly 100 Public Companies
Huang's AI factory empire relies on a broad ecosystem:
- AI Cloud Services: GPU cloud infrastructure (first stop after tokens are generated)
- AI Factory Software: Operating systems and containerization (IBM holds influence)
- Design & Construction: Digital twins and simulation (Cadence, Dassault, Siemens)
- Energy & Cooling: The most critical bottleneck (Delta Electronics, Hitachi, Mitsubishi Electric)
- Infrastructure & Data Centers: Land and facilities (Equinix, Digital Realty)
- Computing Systems & ODM: AI server manufacturing (Dell, HP, Lenovo)
As global AI factory construction accelerates, order visibility for all these partners will expand.
The Next Trillion-Dollar Company?
Huang publicly endorsed Marvell, calling it "the next trillion-dollar company." The two companies announced an expanded partnership to build critical networking infrastructure for next-gen AI data centers. Marvell surged over 40% that week and is up over 240% year-to-date.
Huang explained that agentic AI requires distributing computing tasks across entire data centers, creating immense demand for high-speed interconnectivity. Marvell's optical interconnect, ultra-high-speed DSP, and silicon photonics are at the heart of this infrastructure.
Meanwhile, NVIDIA announced its Spectrum-X Ethernet silicon photonics technology is now in full production, delivering 5x better energy efficiency and AI uptime.
The main AI investment theme is shifting from a single-node GPU computing race to a full-stack AI factory system.
The supply chain bottleneck has expanded from "buying massive quantities of GPUs/ASICs" to solving the entire AI data center delivery chain: power equipment, liquid cooling, data center CPUs, memory (DRAM/NAND/HBM), optical interconnects, high-speed Ethernet/DCI, transformers, and gas turbines.
Among these, storage and optical interconnects remain the most critical bottlenecks in current AI computing systems.