The archive · AI & Models · Product decision · 2025–2026
Gimlet Labs' multi-silicon bet lands a $300M Series B at a $3B valuation
Gimlet Labs bets each AI inference phase should run on its best chip — and that heterogeneous cloud closes the 3-10x efficiency gap in 18 months.
Gimlet Labs
What the business is
Gimlet Labs sells a multi-silicon inference cloud: software that disaggregates AI workloads and runs each phase on the best available chip, delivered as its own managed cloud or inside customer data centers.
Starting capital：Total funding reached $392M: an undisclosed seed led by Factory with angels including Sequoia's Bill Coughran and Intel CEO Lip-Bu Tan, an $80M Series A led by Menlo Ventures (March 2026), and a $300M Series B led by Andreessen Horowitz with Sapphire Ventures, M12 and Arm (September 2026).
How it started
Zain Asgar, a Stanford adjunct professor and exited founder, built Gimlet Labs with Pixie co-founders Michelle Nguyen, Omid Azizi and Natalie Serrino. His starting observation was that AI applications use already-deployed hardware only 15-30% of the time — "you're wasting hundreds of billions of dollars because you're just leaving idle resources" — so he set out to make AI workloads 10x more efficient on silicon that already exists.
What happened
Gimlet Labs emerged from stealth in October 2025 with what it calls the first multi-silicon inference cloud and claims it reliably speeds AI inference 3-10x for the same cost and power, slicing even a single model across different chip architectures. It works with NVIDIA, AMD, Intel, Arm, Cerebras and d-Matrix. In March 2026 the company raised an $80M Series A led by Menlo Ventures at a total of $92M raised, with 30 employees and customers including a top-three frontier lab and a top-three hyperscaler. On September 4, 2026 it announced a $300M Series B led by Andreessen Horowitz — whose managing partner Raghu Raghuram calls heterogeneous inference "where inference infrastructure is headed" — valuing the company at $3B.
How it ended up
Gimlet Labs is scaling a multi-silicon cloud toward hundreds of megawatts of managed heterogeneous infrastructure and says it holds billions of dollars in contracted revenue, backed by $392M in total funding at a $3B valuation as of September 2026.
Background
Gimlet Labs was built by the team behind Pixie, the open-source Kubernetes observability startup acquired by New Relic in 2020. CEO Zain Asgar's founding observation was brutal: AI applications use already-deployed hardware only 15-30% of the time, so the industry was wasting hundreds of billions of dollars on idle compute. His bet was that a software layer could make AI workloads 10x more efficient on silicon that already exists.
The product is a multi-silicon inference cloud: orchestration software that disaggregates agentic AI workloads and runs each phase on the chip best suited to it — inference is compute-bound, decode is memory-bound, tool calls are network-bound — across CPUs, GPUs and purpose-built accelerators, either as Gimlet's own managed cloud or inside customer data centers. The company claims 3-10x throughput gains for the same cost and power, and partners with NVIDIA, AMD, Intel, Arm, Cerebras and d-Matrix.
Gimlet Labs emerged from stealth in October 2025 claiming eight-figure revenue at launch. In March 2026 it raised an $80M Series A led by Menlo Ventures — total funding then $92M — and said its customer base had tripled to include a top-three frontier AI lab and a top-three hyperscaler. On September 4, 2026 it announced a $300M Series B led by Andreessen Horowitz with Sapphire Ventures, M12, Arm, Menlo Ventures and Factory, valuing the company at $3B.
The company says it has secured billions of dollars in contracted revenue for Gimlet Cloud and is scaling to hundreds of megawatts of managed heterogeneous infrastructure. The bet now is that inference — not training — is the dominant AI workload, and that the industry's $765 billion of AI capital expenditure this year needs software to keep every one of those chips busy.
What has to be true
- The pain is quantified: McKinsey-sized data center spending heads toward trillions while Asgar says deployed hardware sits idle 70-85% of the time — an efficiency problem, not a supply problem.
- The wedge is technical and defensible: splitting one AI workload across chips requires orchestration software no silicon vendor owns, so Gimlet sits above the hardware wars instead of inside them.
- The traction is dated and specific: October 2025 stealth exit at eight-figure revenue, a customer base tripled by March 2026, and billions in contracted revenue by September 2026.
- The capital path confirms the thesis: $80M from Menlo Ventures in March 2026, then a $300M Series B led by Andreessen Horowitz at a $3B valuation six months later.
What can be applied
Don't bet on which chip wins — bet on the layer that profits from every winner: Gimlet's software monetizes whichever silicon gets deployed, so the bet pays off whatever architecture prevails.
Aftermath
As of September 2026 Gimlet Labs is a $3B-valued, 392M-funded independent company scaling its multi-silicon cloud, with contracted revenue it describes as billions of dollars and plans to build out hundreds of megawatts of heterogeneous infrastructure. The open questions are whether its claims of 3-10x efficiency survive at hyperscale, and whether chipmakers' own software stacks close the gap before Gimlet's orchestration layer becomes the default.
Sources
- Startup Gimlet Labs is solving the AI inference bottleneck in a surprisingly elegant way
- Now Valued at $3 Billion, Gimlet Labs Raises $300 Million in Series B Led by Andreessen Horowitz for Industry's First Multi-Silicon Inference Cloud for Agentic AI
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card