https://cdn.gophotonics.com/news/vv_639114203925166355.webp712370
Lambda, the Superintelligence Cloud, has been named a launch partner for NVIDIA’s Vera CPU platform and NVIDIA STX at NVIDIA GTC 2026, and also announced the deployment of NVIDIA Quantum-X800 InfiniBand Photonics Co-Packaged Optics (CPO) networking in an AI factory with more than 10,000 NVIDIA Blackwell Ultra GPUs, along with the launch of Lambda Bare Metal Instances as a core cloud offering for frontier AI workloads.
“Lambda’s mission is to expand humanity’s energy and computational capacity,” said Stephen Balaban, Co-founder and CEO of Lambda. "Today's announcements advance that mission, enabling the world's top AI teams with the infrastructure they need to do their best work."
With custom Olympus CPU cores and up to 1.2 TB/s of memory bandwidth, Vera keeps thousands of sandboxed AI environments running in parallel, ensuring NVIDIA GPUs stay fully utilized across reinforcement learning (RL) and agentic workloads. It gives frontier labs access to a compute architecture designed for the demands of next-generation agentic AI systems.
While the industry is shifting toward agentic AI, long-term memory and the processing of massive context windows are critical bottlenecks in inference. Powered by NVIDIA Vera Rubin, BlueField-4, Spectrum-X Networking, and NVIDIA AI software, NVIDIA STX brings a modular reference architecture for rack-scale AI storage platforms, accelerating inference, analytics, and training through next-generation hardware integration and optimized KV-cache management.
As AI factories scale into the tens of thousands of AI accelerators, network architecture becomes as important as the accelerators themselves. Co-packaged Optics (CPO) networking delivers higher efficiency, longer sustained application runtime, and greater resiliency than traditional pluggable transceivers. Lambda is leading one of the largest deployments of NVIDIA Quantum-X800 InfiniBand Photonics Co-Packaged Optics switches to date, in an AI factory with 10,000+ NVIDIA Blackwell Ultra GPUs. The deployment builds on Lambda's November announcement of early CPO adoption.
“The race to build AI factories isn’t won on GPU counts alone. Network architecture is what determines whether those systems can perform at scale,” said Dave Salvator, Director of Accelerated Computing at NVIDIA. “Getting this right is what allows AI infrastructure to power services used by hundreds of millions of people around the world.“
Lambda’s new Bare Metal Instances, paired with custom networking and system-level optimizations, provide infrastructure and research teams with direct hardware access and eliminate virtualization overhead for distributed training workloads. The new offering strengthens Lambda's full-stack AI infrastructure platform, expanding the tools available to frontier labs, enterprises, and hyperscalers. Infrastructure teams gain complete control over the hardware stack while benefiting from Lambda’s reliability, uptime, and operational expertise.
These announcements reflect Lambda’s decade-long collaboration with NVIDIA, as well as the company’s commitment to continuously develop its Superintelligence Cloud: a platform engineered for fast deployment, density, and cooling to meet modern AI demands and maximize the intelligence produced per watt.
