NVIDIA's Rubin platform enters full production — six new chips aimed at one AI supercomputer
NVIDIA's next-generation Rubin platform — six new chips designed to act as a single AI supercomputer — is in full production, with Vera Rubin-based instances set to arrive at AWS, Google Cloud, Microsoft, and Oracle Cloud in the second half of 2026. The cadence from Hopper to Blackwell to Rubin keeps the frontier's compute doubling on schedule.
The six-chips-one-machine framing is the design philosophy. Rubin is not a single accelerator but a rack-scale system in which CPU, GPU, networking, and memory are co-designed to behave as one supercomputer, because at frontier scale the bottleneck is moving data between chips, not the chips themselves. Selling the rack, not the die, is how NVIDIA keeps the performance curve bending.
The hyperscaler line-up is the demand signal. AWS, Google Cloud, Microsoft, and Oracle all committing to Vera Rubin instances in the same half-year means the largest buyers have pre-absorbed the next generation before it ships — the opposite of a market waiting to be convinced. The capex those instances represent is the same trillion-dollar buildout the grid is struggling to power.
The strategic read is that the annual-cadence roadmap is now the moat. By shipping a full new platform every year, NVIDIA forces competitors to chase a moving target and customers to standardise on its stack to stay current. Rubin entering full production on schedule is less a product launch than a demonstration that the cadence holds.
NVIDIA — NVIDIA kicks off the next generation of AI with Rubin — six new chips, one AI supercomputer → · iTiger — Nvidia announces Rubin AI chips for 2026, trillion-dollar data center boom →