Nvidia Wants Robots to Behave
Nvidia led the day's news, planting flags across robotics, science, and the data center. Its biggest swing is Halos for Robotics, which the company bills as the industry's first full-stack safety system for "physical AI" — the catch-all term for AI that controls machines moving through the real world. The pitch is simple to grasp: as humanoid robots start working shoulder-to-shoulder with people, they need a comprehensive safety layer so they stop bumping into the humans around them. For a chipmaker that has ridden the AI boom to dominance, it's a deliberate step deeper into the messy physical world beyond the server rack.
Nvidia didn't stop there. At the ISC conference in Hamburg, it rolled out a new suite of AI software aimed squarely at scientific research, part of a broader push into the laboratory. That effort dovetails with the National Science Foundation's National Artificial Intelligence Research Resource — a U.S. government program, powered by Nvidia hardware, that gives researchers access to heavy-duty AI tools and is already reshaping how science gets done. On the applied side, Eco Wave Power is tapping Nvidia's AI infrastructure and digital-twin technology to turn ocean waves into electricity.
The company also addressed the AI industry's quieter crisis: heat and water. In a post cheekily titled "Hotter Than a Hot Tub," Nvidia detailed a cooling approach that runs the liquid coolant at a toasty 45°C, keeping the most demanding AI machines cool while using far less water.
The Compute Land Grab
The infrastructure story of the day came from an unexpected landlord: SpaceX. Elon Musk's rocket company signed a computing-power agreement worth up to $6.3 billion with Reflection AI, an open-source AI startup, turning its Colossus data center into a major AI play. The terms are eye-popping — Reflection will pay roughly $150 million a month to rent access to Nvidia's coveted GB300 chips. It's a striking sign of how scarce top-tier compute has become when an open-source lab leases its hardware from a space company.
The buildout continues elsewhere, too. Microsoft is planting one of the largest data center projects in its history in Pecos, West Texas, a campus that expands its global AI footprint into the oil country of the Permian Basin.
OpenAI Goes Defensive — and Visual
OpenAI had a two-front day. It introduced Daybreak, a set of tools the company says are meant to help "every organization in the world" find, validate, and patch software vulnerabilities at scale, with a stated plan to fix bugs in open-source code. Turning a model better known for writing software into one that secures it is a notable pivot.
OpenAI also struck a surprise licensing deal with Getty Images, bringing Getty's vast library of licensed photos and visual content into ChatGPT. Investors cheered: Getty's stock soared on the news, a rare bright spot for a company that has spent years fighting over how AI uses creative work.
Hollywood Meets the Algorithm
Finally, Google is investing $75 million in A24, the beloved independent film studio, as part of a new AI partnership, according to the Wall Street Journal's Ben Fritz. The deal hints at how seriously Big Tech is courting Hollywood as generative tools edge into filmmaking — and how cautiously a prestige studio is testing those waters.
The throughline across today's stories: the AI race is no longer just about bigger models. It's about the chips to run them, the power and water to cool them, the safety to deploy them around people, and the rights to the data and images that feed them.