Gimlet Labs raises USD $300m in Andreessen-led round

Gimlet Labs has raised USD $300 million in a Series B funding round led by Andreessen Horowitz, valuing the artificial intelligence infrastructure company at USD $3 billion.

The San Francisco-based group said the latest round brings total funding to USD $392 million, with participation from Sapphire Ventures, M12, Arm, Menlo Ventures and Factory.

Gimlet Labs develops software and cloud services for artificial intelligence inference, an area drawing growing investor interest as companies seek faster, cheaper ways to run AI models after training. It focuses on so-called multi-silicon inference, which distributes workloads across different types of chips rather than relying on a single hardware architecture.

The new funding will be used to expand the company's cloud platform operations and grow its workforce.

Investor backing

Andreessen Horowitz described the investment as a bet on a shift in how AI inference infrastructure is built, as data centre constraints and chip shortages continue to shape the sector's economics.

"AI demand is growing exponentially, while data centers and silicon can't keep pace. The answer isn't just more infrastructure - it's a better architecture. Gimlet has built a new kind of inference cloud, heterogeneous by design, that matches each workload to the right silicon. By making GPUs and purpose-built accelerators work as one system, Gimlet delivers dramatically more throughput, more interactivity and more intelligence from every watt. We believe this is where inference infrastructure is headed," said Raghu Raghuram, Managing Partner, Andreessen Horowitz and Gimlet Labs board member.

Gimlet Labs is positioning itself around a challenge emerging across the AI market: the growing burden of inference workloads as large models are deployed at scale. It argues that relying on a single category of chip creates bottlenecks in speed, power use and hardware utilisation, particularly for agentic AI systems that require repeated decision-making and fast responses.

Cloud build-out

According to the company, its software breaks model inference into stages and assigns each one to the chip type it considers best suited. It offers the service through its own managed cloud and as a managed service within customer environments.

Its hardware partners include NVIDIA, AMD, Intel, Arm, Cerebras and d-Matrix, reflecting a broader industry push to expand the range of silicon used in AI computing beyond Nvidia's dominant graphics processors.

Gimlet Labs emerged from stealth in late 2025 and said earlier this year that it had tripled its customer base. It also said it had signed one of the top three frontier labs and one of the top three hyperscalers as customers, though it did not identify them.

The company also said it has secured billions of dollars in contracted revenue for Gimlet Cloud and is scaling to hundreds of megawatts in managed heterogeneous infrastructure. Those figures suggest investors are backing the company not only for its technology pitch but also for early signs of commercial demand in a market where many infrastructure start-ups are still proving adoption.

Chief Executive Officer and Co-Founder Zain Asgar said the company is responding to surging demand for AI inference.

"We've reached a turning point where inference is the dominant AI workload and the demand for tokens is explosive. With Gimlet, our customers are able to serve massive volumes of tokens at very low latency, even as their AI workloads continue to grow across all dimensions. We're able to deliver unprecedented performance because Gimlet software intelligently slices and orchestrates their workloads across different types of hardware from both mainstream and emerging chipmakers," said Zain Asgar, Chief Executive Officer, Gimlet Labs.

The fundraising comes as investors continue to pour money into AI infrastructure businesses, particularly those promising to lower the cost of running models and reduce dependence on a narrow pool of chip suppliers. As more companies move from training models to serving them in production, inference has become an increasingly important battleground across the sector.

Gimlet Labs said its system combines GPUs, CPUs and purpose-built AI accelerators in a single environment and can improve throughput and interactivity by up to ten times for some workloads.