Gimlet Labs
Category: AI Infrastructure
A multi-silicon inference cloud platform designed to optimize and orchestrate complex agentic AI workloads across heterogeneous hardware architectures. Gimlet Labs was founded in 2023. The company is led by Zain Asgar. Based in San Francisco, USA. Team size: 11-50. Total funding raised: $392.0M. Latest round: Series B. Key investors include Andreessen Horowitz, Sapphire Ventures, Menlo Ventures, Factory, Eclipse, Prosperity7, Triatomic, Arm, M12, Samsung Ventures, Tiger Global Management, 645 Ventures.
- Founded
- 2023
- Headquarters
- San Francisco, USA
- Team size
- 11-50
- Total funding
- $392.0M
Value proposition
Eliminates AI inference bottlenecks and reduces operational costs by dynamically routing workloads to the most efficient hardware (Nvidia, AMD, or specialized AI chips) without requiring manual code refactoring.
Products and solutions
Gimlet Cloud (serverless multi-silicon inference for agents), kforge (autonomous PyTorch kernel generation for CUDA/ROCm/Metal), Agentic Workload Orchestrator / heterogeneous disaggregation, Hardware-agnostic AI runtime & scheduling, Managed multi-silicon inference (cloud or customer environments)
Unique value
Unlike traditional cloud providers that lock users into specific GPU instances, Gimlet provides a 'liquid' infrastructure that abstracts the underlying silicon, allowing AI agents to hop between different chip architectures mid-workflow for optimal performance.
Target customer
Enterprise AI developers, LLM-based application builders, and companies deploying 'agentic' AI systems that require high-scale, multi-step inference.
Industries served
Artificial Intelligence & Machine Learning, Cloud Infrastructure, Enterprise Software, Financial Services, Healthcare Technology
Technology advantage
Leverages proprietary hardware-software co-design to solve the 'agentic bottleneck'—where multi-step AI tasks are slowed by latency and high costs—by using an intelligent scheduling layer that treats diverse hardware as a single unified pool of compute.
How they differentiate
Proprietary multi-silicon orchestration that abstracts hardware (Nvidia, AMD, NPUs), allowing agentic AI workloads to dynamically route to the most efficient chip mid-workflow without code changes.
Main competitors
Together AI, Anyscale, CoreWeave, Lambda Labs
Key partnerships
MLCommons (member; agentic inference benchmarks), Arm (strategic chip-software collaboration), NVIDIA, AMD, Intel, Cerebras, d-Matrix, Andreessen Horowitz (Series B lead), Menlo Ventures, Factory, Sapphire Ventures, M12
Notable customers
Top-3 frontier AI lab, Top-3 hyperscaler, AI-native and Fortune 500 deployments, Billions in contracted Gimlet Cloud revenue (as of Series B)
Major milestones
Emerged from stealth Oct 2025 with $12M Seed (Factory) and eight-figure revenues, $80M Series A Mar 2026 led by Menlo Ventures; customer base tripled incl. top-3 frontier lab and hyperscaler, Joined MLCommons Jun 2026 for agentic inference benchmarks, $300M Series B Sep 2026 led by Andreessen Horowitz at $3B valuation (total raised $392M)
Growth metrics
Launched Oct 2025 with eight-figure revenues; by Series B (Sep 2026) reported billions in contracted revenue for Gimlet Cloud and scaling to hundreds of MW managed capacity; $3B valuation.
Market positioning
Infrastructure-as-a-Service (IaaS) and orchestration layer for high-scale agentic AI deployments.
Geographic focus
Global, with primary operations in North America (San Francisco, CA).
Patents and IP
Proprietary orchestration logic and hardware abstraction layers (specific public filings not disclosed, but core IP centers on multi-silicon scheduling).
About Zain Asgar
Co-founder and former CEO of Pixie Labs (acquired by New Relic); GPU architect at NVIDIA; Engineering Lead at Google AI (Google Lens); Director Engineering at Trifacta (employee #6); Adjunct Professor of Computer Science at Stanford University; PhD in Electrical Engineering from Stanford.
Latest news about Gimlet Labs
More AI Infrastructure companies
Official website: https://gimletlabs.ai/