Gimlet Labs raises $300 million led by Andreessen Horowitz for software that splits AI workloads across chip architectures.
The AMW Read
A $300M round roughly quadrupling a known AI Infrastructure player's prior capital base meaningfully updates the compute-orchestration layer's funding baseline, but falls under the $500M mega-round threshold so no cross-capital tag applies.
Gimlet Labs raises $300 million led by Andreessen Horowitz for software that splits AI workloads across chip architectures.
Gimlet Labs, an AI Infrastructure company, announced a $300 million funding round led by a16z. The company builds software designed to distribute AI compute workloads across different chip architectures rather than a single accelerator family. Per the AI Market Watch index, Gimlet Labs was founded in 2023 and had raised roughly $92 million in total funding prior to this round, a base this new round expands several times over in a single close (coverage, not a census).
Software that allocates inference and training workloads across heterogeneous chips speaks directly to a live tension in AI infrastructure: buyers want the price, availability, and performance benefits of running on multiple accelerator vendors, but doing so has historically meant rebuilding tooling per chip family. A round this size, led by a top-tier venture firm, signals continued investor conviction that the orchestration layer sitting above raw silicon — not the chips themselves — is where durable software value can accrue as inference volume grows and compute costs stay under scrutiny.
For builders, a maturing cross-chip scheduling layer lowers the switching cost of diversifying accelerator vendors, a hedge enterprise buyers increasingly want against GPU scarcity and pricing volatility. For investors, the round is a marker that capital continues flowing into infrastructure middleware sitting between hardware and applications, not only into frontier model labs — reinforcing the compute-execution layer as a distinct, separately fundable category.