What will Jensen Huang say during NVIDIA GTC Washington, D.C. Keynote?

Jensen Huang's keynote topics consistently differ based on the nature of the event. At technical conferences, including GTC, his presentations predominantly focus on advancements in hardware architecture roadmaps, such as the Blackwell platform. These keynotes also cover the evolving AI software ecosystems, the concept of AI factories, and the development of agentic AI ^{[1][2][3][4][5]}.

Conversely, Huang addresses broader societal implications during policy-focused engagements. During these appearances, his discourse emphasizes themes such as national competitiveness, the strategic importance of “sovereign AI,” the necessity of robust domestic energy infrastructure, and the alignment of AI leadership with national security interests. This distinct thematic approach highlights a deliberate communication strategy, tailored to the specific audience and objectives of each conference or event ^{[1][2][3][4][5]}.

Prediction markets suggest Cosmos will feature more prominently at GTC D.C. Prediction market odds indicate Cosmos is more likely to be mentioned by Jensen Huang at the December 1, 2026 GTC D.C. Keynote, with an 83-85% likelihood, compared to Omniverse at 68-79% ^[1][2][3][4]. Strategically, NVIDIA positions Omniverse as the core platform for real-time simulation and digital twins crucial for public sector operations, with Cosmos functioning as a component within the broader Omniverse ecosystem ^[5][6].

Omniverse serves as NVIDIA's foundational platform for public sector digital twins. NVIDIA explicitly states Omniverse is the platform for real-time simulation and digital twins, essential for improving planning and operational readiness across the public sector ^[5]. It enables agencies to develop industrial digital twins and physical AI simulation applications in secure training environments, providing prebuilt OpenUSD, rendering, and simulation capabilities for developers to build virtual worlds for physical AI training and testing ^[5]. During the October 2025 GTC D.C. keynote, Jensen Huang described Omniverse DSX as a blueprint for building and operating gigascale AI factories and the Omniverse computer as the second computer in the physical AI stack ^[7][8].

Cosmos provides generative world models integrated within the Omniverse ecosystem. The 2026 NVIDIA Annual Review clarifies that "Omniverse, with Cosmos World Foundation Models, simulates worlds, robots, and workflows with physical fidelity," indicating Cosmos is integrated within Omniverse rather than being a standalone public-sector platform ^[6]. In the October 2025 GTC D.C. keynote, Cosmos was referenced as the generative model that creates infinite virtual environments for training ^[7][8]. For public sector applications, Cosmos world models are available on GitHub and Foundry for autonomous vehicle development and robotics ^[9].

Jensen Huang will likely address U.S. AI policy and export controls. During his December 1, 2026, keynote at NVIDIA GTC Washington, D.C., Jensen Huang is expected to address the volatile U.S. AI policy and export controls concerning AI chips to China ^{[4][5][1][2][3]}. This focus is primarily driven by ongoing regulatory shifts and heightened scrutiny following October 2026 reports of a suspected Chinese informant allegedly targeting Huang ^[7].

U.S. export controls on AI chips remain dynamic and unpredictable. The U.S. export controls on AI chips to China continue to be a dynamic and unpredictable policy area as of late 2026 ^[4][5]. A January 2026 rule formalized a case-by-case review for H200-class chips, marking a continuation of significant regulatory adjustments throughout 2025 and 2026 ^[4][5]. Jensen Huang has openly criticized these controls, deeming them a "failure" that inadvertently benefits competitors and fosters indigenous Chinese chip development ^[4][6].

Public scrutiny intensified due to suspected Chinese informant reports. Public and policy scrutiny intensified in October 2026 due to reports concerning a suspected Chinese informant allegedly targeting Jensen Huang ^[7]. These developments bring renewed focus on the enforcement of advanced AI hardware export controls, especially regarding the potential diversion of chips from third-country data centers to Chinese end-users ^[7].

Limited official NVIDIA 2026 transcripts permit some term tracking for "Cosmos" versus "Gaming." Specifically, directly searchable official transcripts are available for NVIDIA's FY26 Q4, dated February 25, 2026 ^[1], and FY27 Q2, dated August 26, 2026 ^[3]. The FY26 Q4 transcript provides specific data points, noting Gaming revenue at $3.7 billion and explicitly mentioning "NVIDIA Cosmos" in the context of physical-AI and robotics models ^[1][2].

Comprehensive word counts for 2026 keynotes remain largely unavailable for term frequency analysis. While NVIDIA’s GTC 2026 keynote is accessible on demand, a complete transcript suitable for word counting is not exposed ^[4]. Similarly, a summary from Quartr for the GTC Berlin 2026 keynote, dated October 21, 2026, does not offer auditable "Cosmos" versus "Gaming" word counts ^[5]. Although official investor-relations pages list various 2026 calls and keynotes, including links to transcript PDFs for FY27 Q1 and FY27 Q2 ^[6][7], these materials do not specify whether they contain mentions of "Cosmos" or "Gaming," nor do they provide full auditable word counts for these terms from all available 2026 documents, beyond the specific mentions found in the FY26 Q4 transcript ^{[1][2][3][4][5]}.

NVIDIA consistently emphasizes Physical AI through strategic releases and declarations in 2026. The company has highlighted "Physical AI" across multiple events throughout the year, underscoring its strategic importance and ongoing development. At CES 2026 on January 5, NVIDIA released Isaac GR00T N1.6 open models designed for physical AI applications ^[4]. This strategic focus continued at GTC 2026 in March, where NVIDIA declared that 'the big bang of physical AI has started' and announced the Physical AI Data Factory Blueprint in collaboration with partners ^[1]. These recurring mentions and significant releases strongly indicate the probable inclusion of "Physical AI" in the upcoming GTC Washington, D.C. Keynote.

The Cosmos platform is an expanding, frequently discussed theme throughout 2026. The "Cosmos" platform has been a recurring and expanding theme in NVIDIA's activities. On January 5, 2026, CES saw the release of Cosmos Transfer 2.5, Cosmos Predict 2.5, and Cosmos Reason 2 open models ^[4]. Further advancements were showcased at SIGGRAPH 2026 on July 20, with the launch of Cosmos 3 Edge, a 4-billion-parameter open world model for local physical AI, and Cosmos-Dreams closed-loop simulators. Cosmos 3 is available in Edge (4B), Nano (16B), and Super (64B) configurations ^[2]. The platform's capabilities were further demonstrated at Actuate 2026 on August 18–19, where 'Cosmos 3 for Robotics' was presented as an omnimodal world foundation model ^[5]. By October 22, 2026, the Cosmos platform saw expansion with new training tools, simulation checkpoints, and Cosmos Data Factory blueprints, achieving top-five rankings on six different open models leaderboards ^[3]. A Kalshi market pricing 'Cosmos' at 70% for NVIDIA's November 18, 2026 earnings call further supports the expectation of these terms in near-term communications ^[6].

NVIDIA's fiscal Q3 2027 earnings are scheduled for November 17, 2026 ^[1][2][3]. The company guided revenue to $108 billion ±2% (89% YoY growth) and gross margin to 74% ±50 bps, explicitly excluding China data center compute revenue ^[1][2][3]. Polymarket indicates a 97% chance of beating earnings, with consensus EPS near $2.09 on ~$92B revenue ^[4]. Kalshi prediction market traders expect Jensen Huang to say 'TSMC' (41%) more than 'Taiwan' (18%) during earnings, with 'Tariff' at 18%, 'Trump' at 8%, 'H20' at 11%, and 'GM' at 4% ^[4]. Following earnings, NVIDIA GTC Washington, D.C. 2026 runs November 30–December 3, 2026, with Jensen Huang's keynote on Tuesday, December 1 at 2 p.m. ET ^[5][6][7]. The conference theme is 'super intelligence (SI) accelerating American economic growth and scientific discovery' ^[5][6][7]. Analyst William Stein expects NVIDIA to reiterate its ~$500 billion backlog and $3 trillion to $4 trillion TAM by 2030 at GTC, noting comments on market sizing and product introductions should be a 'modest positive' for the stock ^[8].

The Vera Rubin platform began production shipments in August 2026 and is expected to make up about 20% of data center revenue in fiscal Q3 2027 ^[9][2]. Commercial shipments are expected to scale through H2 2026 into 2027, with supply remaining a bottleneck at least through end of fiscal 2028 ^[9][2]. At GTC San Jose 2026 (March 16, 2026), Huang declared 'the inference inflection has arrived' and doubled demand visibility to $1 trillion in Blackwell and Vera Rubin orders through 2027 ^{[10][11][12][13]}. This follows a prior GTC DC keynote in October 2025 where $500 billion in cumulative revenue visibility for Blackwell and Rubin products through calendar 2026 was disclosed ^[14][15]. A 20-year, 12-gigawatt compute agreement tied to OpenAI and SB Energy points to roughly $600 billion in potential cumulative revenue before 2030 ^[9]. Hyperscaler capital spending is on track to top $1 trillion in 2027, with Meta raising its 2026 capex range, Alphabet projecting a notable rise in 2027 outlays, Microsoft expecting accelerated spending in H1 2027, and Amazon reaffirming its full-year capex plan ^[1].

Bearish risks include gross margin compression to 71-72% in fiscal Q4 2027 due to rising memory prices, down from 75% in fiscal Q2 2027 ^[9][2]. Other headwinds include in-house custom ASICs at large cloud customers, China export controls, and a DOJ probe of NVIDIA's $12.9 billion Hugging Face purchase ^[9][2]. Wall Street consensus remains a 'Strong Buy' with an average 12-month price target in the low-to-mid $300 range, showing wide dispersion from ~$180 to $515 ^{[9][11][16][17][18]}. BofA, Citi, and JPMorgan all set $300 targets after GTC 2026, while one analyst raised to $325, and BNP Paribas raised to $345 in October 2026 ^{[9][11][16][17]}.