NVIDIA CEO Jensen Huang announced that the company expects to generate at least US $1 trillion in cumulative revenue from its Blackwell and Vera Rubin AI platforms by the end of 2027, at the GTC 2026 conference.
This updated forecast doubles the US $500 billion opportunity the company projected for the period ending in 2026. Huang attributed this growth to the arrival of the inference inflection, where demand for running AI models in real-time now matches the demand for training them.
The Vera Rubin platform and data centre expansion
A centre of this strategy is the new Vera Rubin architecture, which follows the Blackwell generation. The company introduced the Vera Rubin NVL72, a liquid-cooled rack-scale system that integrates 36 Vera CPUs and 72 Rubin GPUs. This system is designed for models with more than one trillion parameters and provides a 10X reduction in costs for inference tasks compared to previous hardware.
The platform also includes the NVIDIA Vera CPU, which uses 88 custom cores to deliver twice the efficiency of traditional data center processors. To manage the heat generated by these high-density systems, NVIDIA moved to 100% fanless direct liquid cooling. Additionally, Huang discussed Vera Rubin Space-1, a project aimed at deploying AI data centers in outer space through partnerships with satellite providers.
Advancements in graphics and robotics
NVIDIA introduced DLSS 5 (Deep Learning Super Sampling), which the company describes as a neural rendering model. Unlike previous versions that focused on upscaling and frame generation, DLSS 5 uses AI to inject photorealistic lighting and materials into game frames. The technology analyses scene semantics like fabric, hair, and translucent skin to bridge the gap between computer-generated imagery and reality. It is scheduled for release in Fall 2026, with support from publishers like Bethesda, Ubisoft, and Capcom.
Market shift
NVIDIA is diversifying its hardware stack to include central processing units (CPUs) and language processing units (LPUs). The company showcased the NVIDIA Groq 3 LPU, which utilises technology licensed from the startup Groq in a US $17 billion deal. This expansion allows NVIDIA to offer a full-stack “AI Factory” that combines various chip types to handle agentic AI, autonomous systems that can reason and execute tasks.
While the financial outlook remains high, investors showed a measured response to the news. NVIDIA shares rose 1.2% following the keynote, as the market weighs the US $1 trillion forecast against rising competition from custom silicon developed by cloud providers like Google and Amazon.
Timeline for New Technology
- Blackwell Systems: Currently in full production and shipping to customers.
- DLSS 5: Expected to launch in Fall 2026.
- Vera Rubin Systems: Scheduled for general availability in the second half of 2026.
- NVIDIA Groq 3 LPU: Expected to ship in the third quarter of 2026.