At the GTC 2026 conference, NVIDIA CEO Jensen Huang predicted that the Blackwell and Vera Rubin platforms would generate at least $1 trillion in revenue by the end of 2027, doubling last year’s forecast, far exceeding Wall Street expectations and driving the stock price up.

NVIDIA unveiled the Vera Rubin platform, a complete supercomputer consisting of seven chips—including the Vera CPU, Rubin GPU, and Groq 3 LPU—and five rack systems, built on TSMC’s 3nm process and designed for agentic AI.

The new Vera Rubin platform delivers 16 PetaFLOPS of inference compute at FP8 precision, an 8x improvement over the H100. Huang introduced “token factory economics,” stating that inference token cost has dropped to less than one‑tenth of what it was in the GPT‑4 era.

NVIDIA launched the Groq 3 LPU, optimized for low‑latency large language model inference. It integrates large‑capacity SRAM and achieves token throughput of up to 1,500 tokens per second to support complex multi‑agent interactions.

NVIDIA introduced the Quantum3400 CPO switch, which reduces transmission loss by 60% and improves energy efficiency by 5x. Rack power density for next‑generation high‑performance computing clusters has risen to 240‑260 kW per rack, making liquid cooling the standard thermal solution.

Cloud service provider Akamai purchased thousands of NVIDIA Blackwell GPUs to build a globally distributed AI inference platform, deploying compute at the edge to reduce latency and egress costs.

Chinese GPU companies such as Moore Threads and Moore Threads (MoX) recently went public, accelerating capital market activity in the sector. However, some AI computing centers report utilization rates below 50% for domestic chips, shifting industry focus from raw specifications to ease of use and full‑stack solutions.

The US and China are diverging in their computing development paths: the US emphasizes cutting‑edge technologies such as 3nm and 1.6nm processes and space‑based computing, while China focuses on integrating computing with the real economy, implementing a “computing‑electricity synergy” strategy that uses green power from the west to address east‑west supply‑demand mismatches.

NVIDIA is collaborating with automakers including BYD and Geely to develop L4 autonomous vehicles based on the DRIVE Hyperion platform. At the same time, the company is working on data‑center computers designed for deployment in space to enable real‑time sensing and autonomous decision‑making.

Following the popularity of the open‑source project OpenClaw, NVIDIA launched the NemoClaw platform to help enterprises securely deploy agents. The AI ecosystem is shifting toward “agent‑as‑a‑service,” with tokens increasingly becoming a new unit of value in the industry.