Gimlet Labs and the Multisilicon AI Cloud
Could a software startup weaken Nvidia’s dominance without manufacturing a single chip? Gimlet Labs has raised $300 million at a $3 billion valuation to build a “multisilicon cloud” that routes each phase of an AI inference workload to the processor best suited for it—including GPUs, CPUs and specialized accelerators.
This Short explains why Andreessen Horowitz, Arm and Microsoft’s M12 are backing the company, what Gimlet claims its platform can deliver, and why Nvidia’s CUDA ecosystem remains the biggest obstacle. We also separate investor enthusiasm and company performance claims from results that still require independent validation.
Source: AI Weekly, September 4, 2026, linking Bloomberg News reporting by Dina Bass. Additional first-party context from Gimlet Labs’ Series B announcement by Zain Asgar, Michelle Nguyen, Omid Azizi, James Bartlett and Natalie Serrino.
What to watch
Routing workloads across hardware can offer flexibility, but benefits depend on software compatibility and real workloads. Funding and investor support do not independently validate performance claims. The relevant comparisons include end-to-end cost, reliability, and the effort required to move an existing application.
Related reading: our overview of AI tools in shared workplace documents.
Watch and listen
Watch the YouTube Short above or listen to the full episode on Spotify.
