Gimlet Labs Raises $300 Million to Build a Multi-Silicon Inference Cloud for Agentic AI

Gimlet Labs, a San Francisco-based AI infrastructure company, raised $300 million in a Series B led by Andreessen Horowitz, valuing the company at $3 billion. The round includes new investors Arm, M12 and Sapphire Ventures alongside previous backers Menlo Ventures and Factory, and arrived just six months after Gimlet's $80 million Series A, bringing its total funding to $392 million. Since that Series A, the company has tripled its customer base and added one of the top three frontier AI labs and one of the top three hyperscalers as customers, a pace of enterprise adoption that helps explain why a16z returned to lead a round this large so soon after the first institutional check.

What Gimlet Builds

Gimlet Labs is building a multi-silicon inference cloud for agentic AI, infrastructure designed to run AI workloads across multiple types of processors rather than locking a customer into a single hardware architecture. The company works across NVIDIA, AMD, Intel, Arm, Cerebras and d-Matrix chips, letting its infrastructure blend GPUs, CPUs and purpose-built AI accelerators depending on what a given workload actually needs. That flexibility is the core pitch: as agentic AI systems move from single model calls to long chains of reasoning and tool use, the cheapest and fastest chip for any one step in that chain can change constantly, and most inference infrastructure today is built assuming it won't.

Why the Chip Diversity Bet Matters

The AI infrastructure market has largely organized itself around a single vendor's hardware for the last several years, and that concentration has become its own risk, both in pricing power and in supply. A company that can route inference workloads across six different chip families is making a specific bet: that the winners in applied AI will be the companies that can adapt to whatever hardware is fastest or cheapest at a given moment, not the ones locked into a single supplier's roadmap. Arm's participation as a new investor in this round is a signal in the same direction, since Arm-based silicon is one of the architectures Gimlet explicitly supports alongside the incumbent GPU vendors.

What This Means for Founders

Gimlet's raise is a useful data point for any founder building on top of AI infrastructure rather than around it. The pace here, tripling customers and closing a $300 million round six months after an $80 million Series A, reflects how quickly enterprise AI infrastructure spend is moving right now, and it is a reminder that infrastructure-layer startups solving a real cost or reliability problem for frontier labs and hyperscalers can raise at a speed that outpaces almost every other category. For founders building applications on top of inference infrastructure, Gimlet's multi-silicon approach is also worth watching directly, since a chip-agnostic inference layer changes the calculus on which hardware dependencies are worth building into an application versus abstracting away entirely. Founders working in AI infrastructure, chip-adjacent tooling or agentic AI systems looking for investors with a thesis in this exact layer can browse AI and ML focused investors on AngelLinx.

Find investors who are actively deploying by creating your pitch listing on AngelLinx @ angellinx.ai/register.

AngelLinx Intelligence | angellinx.ai