At CES 2026, NVIDIA opened the show with a keynote by founder and CEO Jensen Huang. The annual tech event took place Jan. 6–9 in Las Vegas, with Huang’s keynote delivered on the evening of Jan. 5 and focused on “AI infrastructure, physical AI and the industrial revolution.” According to a series of reports from Investing.com, NVIDIA’s headline at this CES was the formal unveiling of its next‑generation AI platform Rubin, promising a 10× reduction in inference token costs and announcing six new chips. Huang also emphasized automotive AI progress, saying NVIDIA will begin testing a self‑operated Robotaxi service in 2027 to address competition from Waymo, Tesla and others. NVIDIA showcased multiple AI models and robotics hardware, further consolidating its leadership in the AI ecosystem. The NVIDIA booth drew heavy interest during the show, which attracted more than 100,000 attendees. This article synthesizes Investing.com’s live coverage to summarize NVIDIA’s major announcements and analyze their technical and industry implications.
Keynote core: Rubin platform launch, sharp inference cost reduction
In his address Huang formally introduced the Rubin platform, the next‑generation AI architecture following Blackwell, named in honor of astronomer Vera Rubin. Investing.com reported that Rubin’s headline claims include a 10× reduction in inference token cost, along with 3.5× training performance and 5× inference performance improvements versus Blackwell. Huang said the Vera Rubin GPU has entered full production and will ship at scale during 2026.
The Rubin platform includes six new chips covering scenarios from data centers to edge computing:
- Rubin Ultra: the flagship for hyperscale AI training and inference.
- Rubin GPU: optimized for high‑performance inference, with single‑chip inference throughput up to 5× that of Blackwell.
- Four additional variants targeting specific domains such as automotive, robotics and edge AI.
The announcement addresses the industry pain point of high AI compute costs. Huang said Rubin achieves energy‑efficiency leaps via architectural optimizations, including more efficient NVLink interconnects, advanced HBM memory and new transistor designs, bringing token costs down to 1/10 of Blackwell. That reduction could materially lower cloud operating costs for trillion‑parameter models and accelerate enterprise adoption of large‑scale AI.
Automotive AI and Robotaxi plan: testing a self‑operated service in 2027
Automotive AI was another focal point of Huang’s talk. He emphasized progress on the NVIDIA DRIVE platform and argued that “physical AI” will reshape transportation. Investing.com reported Huang’s announcement that NVIDIA plans to launch and begin testing a self‑operated Robotaxi service in 2027, signaling the company’s move into the robotaxi market. NVIDIA already partners with multiple automakers, including Mercedes‑Benz and Rivian, but an owned Robotaxi service would directly challenge Tesla Robotaxi and Waymo.
Technically, the Robotaxi effort will use Rubin‑based automotive chips and aim to support L4/L5 autonomy. Huang demonstrated a live example of end‑to‑end vehicle decision‑making in complex urban environments, using the NVIDIA Cosmos world model to predict pedestrian behavior and traffic flow. The announcement comes amid competitive pressure—Tesla has signaled Robotaxi plans for 2026 and Waymo operates in multiple U.S. cities. Huang argued NVIDIA’s open ecosystem and compute advantages will enable it to catch up and compete effectively.
AI models and robotics hardware: multi‑domain demonstrations
NVIDIA also unveiled several new AI models and robotics technologies at CES, with physical AI as a highlight, including:
- An upgraded NVIDIA Cosmos open model for robot training and digital twins, supporting more realistic physical simulation.
- New robotics hardware: in partnership with Agility Robotics, Figure and others, NVIDIA demonstrated humanoid robots in real time, using Rubin GPUs to accelerate perception‑to‑decision loops.
- Omniverse platform extensions: new industrial AI tools and a joint keynote with Siemens showcasing smart‑factory applications.
Huang stressed these technologies will drive an “industrial AI revolution,” citing deployments such as Hitachi’s HMAX system and Schaeffler’s manufacturing optimizations. Agentic AI building blocks were highlighted as enabling multi‑step task automation for enterprises.
Ecosystem partnerships and show‑floor interactions
NVIDIA’s booth featured more than 20 live demonstrations with partners including Siemens, Lenovo, Mercedes‑Benz, Snowflake and Abridge. A joint keynote with Siemens on Jan. 7 focused on industrial AI, and a session with Lenovo on Jan. 8 covered AI PCs. A pre‑show panel hosted by Bank of America included guests from the automotive, healthcare and robotics sectors to spotlight ecosystem collaboration.
Impact analysis: a cost revolution and intensified competition
The Rubin platform launch has been described as an AI compute “cost revolution.” A 10× reduction in inference cost could substantially lower cloud AI service prices and spur adoption of trillion‑parameter models by small and midsize companies. Investing.com analysis suggests this will compress margins for competitors that rely on high‑cost inference, such as OpenAI and Anthropic, while stimulating demand—NVIDIA’s stock rose during CES.
The Robotaxi plan marks a strategic shift for NVIDIA from chip supplier toward service operator, directly confronting Tesla. The planned 2027 tests will serve as a test of NVIDIA’s end‑to‑end stack, though regulatory and safety hurdles remain significant.
Overall, NVIDIA’s CES 2026 announcements set the tone for AI development in 2026: a shift from pure scale to efficiency and application. Rubin’s 10× cost advantage, the 2027 Robotaxi objective and new robotics hardware underscore NVIDIA’s full‑stack ambitions. In an increasingly competitive landscape, NVIDIA is leveraging an open ecosystem to consolidate leadership and drive AI from labs into real‑world deployments. The event highlighted both technical advances and an ongoing investment wave in AI infrastructure.
▶️ Enjoy this story? Creating content is hard!
Follow our YouTube channel:
AI Rosetta
Your likes, subscriptions and shares are our greatest support! 🙏
More AI news at: https://www.airosetta.com