Full Coverage: Jensen Huang's "Full-Stack AI" Speech - New Blueprint for Computing Power Industry About to Be Unveiled

At Tuesday dawn Beijing time, the highly anticipated keynote speech at NVIDIA’s GTC conference is about to begin. NVIDIA founder and CEO Jensen Huang will deliver approximately a two-hour keynote, with an official preview stating that the speech will cover “the entire technology stack,” including chips, software, models, and applications.

According to previous reports from Cailian Press, the main focus points of GTC 2026 keynote include:

1. Updates on product roadmaps, including the upcoming Vera Rubin platform, Rubin Ultra chips, and the next-generation Feynman architecture with potential revenue expectations;

2. Collaboration with Groq and LPU related news, with Rubin CPX possibly being shipped separately;

3. Data transmission and power supply systems, involving popular concepts such as optical communication, CPO, HVDC, PCB, etc.;

4. Enterprise-oriented AI open-source intelligent platform NemoClaw, and discussions on which publicly traded companies will be mentioned in AI application layers.

Jensen Huang’s speech coincides with U.S. stock trading hours, so the capital markets will respond promptly to his remarks. As the world’s most valuable publicly traded company, NVIDIA’s stock has been in a sideways trend since last summer. Analysts generally expect that NVIDIA needs to deliver some real surprises to break the current stalemate.

At 02:00, before Jensen Huang takes the stage, NVIDIA’s stock price rose over 2%, still fluctuating around the mid-range of last year’s prices. Huang Huang seems to have arrived a bit late, as the speech has not started on time.

(Source: TradingView)

Pre-Analysis: 【Wedbush Securities: NVIDIA Likely to Provide a Very Bullish New Outlook】

Wedbush Securities, an investment bank, stated that NVIDIA is very likely to present a “very bullish” latest outlook at GTC.

The analysts wrote in their report: “We expect the company to provide a very optimistic update centered around enterprise AI demand, and anticipate Jensen Huang will make an ‘unreserved’ positive outlook on the entire AI industry and order pipeline opportunities. We believe this scale will easily surpass the previous target of $500 billion in shipments for the Blackwell and Rubin platforms by 2026.”

The analysts further pointed out: “This also aligns with NVIDIA reaching a $3 trillion to $4 trillion market opportunity in AI infrastructure by 2030. We expect the company to update on this at GTC. Meanwhile, the product roadmap for the next few years may also be further disclosed at GTC. The company is continuously expanding its AI product portfolio and further targeting the ‘Physical AI’ market opportunities related to automation.”

Pre-Analysis: 【Guo Ming: Nvidia’s Investment in Groq Could Boost LPU Shipments by Over 10 Times】

Earlier Monday, well-known analyst Guo Ming stated on social media that after NVIDIA’s investment in Groq, the LPU ecosystem for ultra-low latency inference is entering an explosive phase. Supply chain estimates suggest that LPU shipments will reach 4 to 5 million units in 2026–2027, a tenfold increase year-over-year, mainly benefiting from deep integration with NVIDIA’s software ecosystem such as CUDA, as well as the expanding demand for low-latency inference driven by AI agents, real-time consumer applications, and Physical AI.

Guo Ming further indicated that NVIDIA plans to increase the single-rack LPU density from 64 to 256 units to meet the KV-cache needs of long-context inference while maintaining low latency during decoding. The new generation rack is expected to enter mass production by late 2026 to early 2027. This architecture upgrade will also drive deep integration of network interconnects, compilers, and development frameworks, triggering a new industry cycle for high-end PCB materials.

(Source: Cailian Press)