NVIDIA's GTC conference is scheduled for March 16-19, 2026. As a bellwether event in the global AI computing sector, the conference is expected to unveil core specifications for next-generation GPUs, including the Rubin and "Feynman" architectures. It will also showcase technological breakthroughs and commercialization progress in computing infrastructure areas such as CPO switches, power architectures, and liquid cooling systems.
According to the official schedule, CEO Jensen Huang will deliver a keynote speech from 11:00 AM to 1:00 PM Pacific Time on March 16 (2:00 AM to 4:00 AM Beijing Time on March 17). Huang will take the stage at the SAP Center in San Jose to outline NVIDIA's strategic plans for the coming year to developers, analysts, and media representatives.
Several securities firms have released preview reports analyzing potential industry shifts and investment opportunities arising from the event.
Guolian Minsheng Securities noted that NVIDIA has established a global CPO supply chain covering core chips, optoelectronic devices, materials, and packaging. The firm highlighted that key focus areas for this GTC include the roadmap for next-generation AI chips and network architecture solutions, particularly technical details of scale-out network architectures such as Quantum-X and Spectrum-X CPO switches.
Additionally, Guolian Minsheng pointed out that the GTC conference may reinforce market trends centered on "speed + power efficiency," suggesting investors pay close attention to related sectors like PCBs, CPO technology, memory, power supplies, and thermal management.
Guojin Securities recommended focusing on CPO, LPU, AI liquid cooling, AI power systems, orthogonal backplanes, and the Feynman architecture. The firm anticipates that NVIDIA may introduce a new inference chip integrating Groq LPU technology to address growing demand for high-efficiency, cost-effective AI inference solutions amid rising competition. They emphasized that ongoing upgrades in large models and the emergence of AI applications like OpenClaw continue to drive strong demand for computing power, reinforcing the value of NVIDIA's AI hardware.
Guotai Haitong Securities suggested that the most significant aspect of GTC 2026 may not be individual chip specifications, but rather whether NVIDIA will use the event to transition the industry from simply "purchasing GPUs" to "deploying AI factories" through systematic implementation of the Rubin platform, roadmap disclosures for Feynman, and integrated upgrades in optical interconnect, power supply, and liquid cooling technologies. The firm advises monitoring AI chips, computing power, and memory sectors.
Dongfang Securities indicated that the upcoming GTC is likely to feature updates on the new Vera Rubin platform and the next-generation Feynman architecture. The VR platform may adopt microchannel cold plate technology, which could require higher liquid cooling performance compared to the GB300 platform, potentially increasing the value of liquid cooling systems. As NVIDIA and Google ramp up shipments and chip cooling demands grow, domestic suppliers integrated into international supply chains are expected to see further development opportunities.
Citic Securities highlighted that with the GTC 2026 conference approaching, NVIDIA is likely to expand its chip product portfolio. In addition to the six core chips of the Vera Rubin AI platform, the company may disclose more details about the Rubin Ultra chip and cabinet designs, introducing innovations in data interconnect and power delivery architectures. Visibility for new products like orthogonal backplanes and CPO is also expected to improve. NVIDIA may also launch an LPU inference chip to complement its CPX series and expand its inference capabilities. Further insights into the future Feynman architecture and NVIDIA's vision for AI infrastructure and industry trends are anticipated. The conference is expected to strengthen market confidence in the sustained growth and incremental potential of the AI sector.
Comments