Nvidia set a massive tone for the future of enterprise technology and autonomous transportation at its annual event. During the Nvidia GTC 2026 keynote in San Jose, Chief Executive Officer Jensen Huang outlined a sweeping vision that spans next-generation hardware, enterprise software, and global vehicle networks. The announcements highlighted how rapidly the artificial intelligence landscape is evolving and scaling.
The core of the presentation focused on aggressive financial projections and hardware rollouts. Nvidia now expects its business to reach unprecedented financial milestones driven by extreme demand for artificial intelligence infrastructure. Alongside these hardware gains, the company detailed major software collaborations and safety platforms designed to push automated systems into everyday commercial use.
Accelerating Hardware and Trillion-Dollar Projections
Demand for advanced computing power continues to supercharge Nvidia’s business operations. Huang announced that the company now expects its Blackwell and Rubin chips to generate at least $1 trillion in revenue by the end of 2027. This marks a significant upward revision from the previous forecast, which projected $500 billion in sales from these chips by the close of 2026.
To further expand its computing architecture, Nvidia unveiled the Groq 3 Language Processing Unit. This marks the first major hardware release following Nvidia’s $20 billion asset purchase of Groq last December. The new processor is designed to accelerate language model performance and is expected to begin shipping in the third quarter of this year.
Looking further ahead, the company provided a glimpse into its long-term hardware roadmap. The upcoming 2028 Feynman family will introduce a new generation of graphics and language processing units, accompanied by a new central processing unit designated as Rosa.
Mistral Forge Enables Custom Enterprise Models
Moving beyond core hardware, the event showcased new ways for organizations to build specialized artificial intelligence. French startup Mistral AI utilized the conference to launch Mistral Forge, a platform that allows enterprises and governments to train custom models from scratch using their own proprietary data.
Many enterprise projects struggle because standard models lack institutional knowledge. Rather than relying on retrieval augmented generation or fine-tuning existing systems, Mistral Forge fundamentally retrains models using internal documents and workflows. Customers can build upon Mistral’s library of open-weight systems, including the newly introduced Mistral Small 4.
To ensure organizations can properly manage synthetic data pipelines and evaluation frameworks, Mistral is deploying specialized engineers to embed directly with customers. The platform is already being utilized by early adopters such as Ericsson, the European Space Agency, and Dutch chipmaker ASML.
NeMoClaw Secures Autonomous AI Agents
Security and privacy remain critical hurdles for widespread automated system adoption. To address this, Nvidia introduced NeMoClaw, a dedicated software stack built for the OpenClaw open-source agent platform. This tool allows developers to install Nemotron models and the new OpenShell runtime using a single command.
NeMoClaw creates a sandboxed privacy and security layer directly beneath autonomous agents. A built-in local privacy router enables these agents to dynamically choose between tapping into cloud-based frontier models or keeping sensitive data processing strictly on-device. The software is highly versatile, running on standard GeForce RTX personal computers, RTX PRO workstations, and enterprise-grade DGX systems.
Additionally, the newly formed Nemotron Coalition announced its first major project. Nvidia and Mistral AI co-developed a new base model trained entirely on DGX Cloud infrastructure. This foundational model will be open-sourced upon release and is slated to underpin the upcoming Nemotron 4 family.
Expanding the Global Robotaxi Footprint
Nvidia is also aggressively stepping up its automotive partnerships to establish a stronger foothold in the global autonomous vehicle sector. Leading automakers BYD, Geely, Isuzu, and Nissan are officially adopting the Drive Hyperion platform. This system integrates the necessary computing power, sensors, and software to build Level 4 autonomous vehicles.
The ridesharing industry is simultaneously adopting these technologies to scale automated fleets. Lyft will integrate the Hyperion platform to develop its own customized robotaxis. Meanwhile, Nvidia provided an update on its ongoing collaboration with Uber, revealing plans to establish a global network of 100,000 autonomous vehicles by 2027. This rollout will eventually span 28 markets across four continents, with Los Angeles and San Francisco scheduled as the first launch cities early next year.
To manage the complex edge cases associated with driverless navigation, Nvidia introduced Halos OS. Operating as a strict safety guardrail, the system is engineered to physically intervene if automated models attempt to make unsafe driving decisions. This fail-safe architecture ensures that if any primary computer or sensor fails, the vehicle can still independently navigate to a secure location.
