Funding Brief｜Gimlet Labs Series B $300M

🌏 中文版

Funding Details

What This Company Does

Gimlet Labs builds a "multi-silicon inference cloud" — instead of betting on a single chip architecture, it splits an AI inference workload into stages and dynamically routes each stage to whichever hardware, GPU, CPU, or a specialized accelerator, fits it best.

Its platform, which it calls the industry's first multi-silicon inference cloud, supports chip architectures from NVIDIA, AMD, Intel, Arm, Cerebras, and d-Matrix, and is available either through Gimlet's own managed cloud or deployed inside a customer's own data center. The company claims this heterogeneous-hardware orchestration delivers 3-10x faster inference at the same cost and power envelope — squeezing more compute out of the same footprint, which lines up neatly with agentic AI's appetite for low latency and high interactivity.

Gimlet emerged from stealth in October 2025 with eight-figure revenue already in hand, serving both AI-native companies and Fortune 500 customers. By March it had tripled its customer base and landed one of the top-three frontier labs and one of the top-three hyperscalers as customers; by this September round, the company says it has secured "billions of dollars" in contracted revenue and is scaling its managed heterogeneous infrastructure toward hundreds of megawatts.

What This Funding Signals

Implications for the Agent Ecosystem

Inference has become software's biggest workload, but data centers were largely designed for training or for a single chip type — that mismatch is exactly what this round of funding targets. Gimlet says the new capital will scale its multi-silicon cloud operations and keep growing the team; co-founder and CEO Zain Asgar has also said the company is now helping design data centers themselves, since different chips demand different cooling, power, and rack configurations, turning a software problem into a physical infrastructure one. In other words, Gimlet isn't trying to win by picking the fastest chip — it's betting that the software layer coordinating many chips could become an infrastructure layer in its own right.

What Investors Are Betting On

a16z managing partner and Gimlet board member Raghu Raghuram put the thesis plainly: "The answer isn't just more infrastructure — it's a better architecture." The investment lands right after a16z expanded its Growth Fund to $8.5B in late August and launched a $1.1B Machine Age Fund targeting chips, memory, data centers, and robotics — putting Gimlet squarely inside a16z's broader bet on the physical infrastructure underneath AI, alongside existing portfolio companies like OpenAI, Databricks, xAI, Anduril, and SpaceX. Compared with fellow inference-infrastructure players Groq (a $350M round at a $3.5B valuation) and Cerebras (a $1B round at $23B, later a Nasdaq debut valued above $56B), Gimlet isn't trying to build a superior chip of its own — it's betting that the multi-chip orchestration software can grow into an independent infrastructure layer.

Numbers Worth Watching

• Valuation jumped from $400M at Series A to $3B at Series B — a 7.5x increase in six months, steeper than enterprise agent-platform peer Wonderful's 2.5x rise over the same window.
• $392M raised across seed ($12M), Series A ($80M), and Series B ($300M) in under a year since its October 2025 stealth launch.
• A customer list already including a top-three frontier lab and a top-three hyperscaler — a customer density rarely seen at this stage for an infrastructure startup.

Watchlist Status

Gimlet Labs is not yet tracked in the watchlist. Recommend adding it to section A3 (Inference Infrastructure), with tracking focus on: multi-silicon inference cloud for agentic AI, $300M Series B at a $3B valuation, led by Andreessen Horowitz.

Today's Takeaway

Most inference-infrastructure funding stories over the past six months have centered on "whose chip is fastest" (Groq, Cerebras). Gimlet Labs' valuation jump shows the market is also betting on a different route — not building a chip, but building the layer that lets every chip work together. If agentic AI workloads keep getting more heterogeneous, that "translation layer" could end up worth as much as any single chip.

References

• 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
• Andreessen-backed Gimlet Labs hits $3B valuation with $300M round as AI goes multi-chip
• Gimlet Labs Raises $300M in Series B Round; Valuation at $3B