Gimlet Labs raises $300 million as it outlines hundreds of MW of AI inference capacity by end-2027

Gimlet Labs has announced a $300 million financing round led by Andreessen Horowitz (a16z), valuing the AI inference cloud startup at $3 billion. The company also says it has cloud-services contracts worth several billion dollars, although it has not disclosed a precise value, counterparties or contract duration.

The company’s infrastructure model uses software orchestration to allocate stages of AI inference across different processor types. Gimlet Labs says it limits the chip mix within each data center to two to four types in order to reduce operational complexity. Suppliers referenced by the company include Nvidia, AMD, Intel, Arm, d-Matrix and Cerebras.

For data-center infrastructure, the principal disclosed target is the planned deployment of hundreds of MW of energy capacity by the end of 2027. This is a forward plan, not commissioned capacity. Gimlet Labs has also said it intends to present GW-scale data-center power plans in the coming years, without specifying capacity, sites, geography, phased delivery schedules or grid-interconnection arrangements.

The strategy places heterogeneous-accelerator compatibility, cooling, network design and power efficiency at the center of inference-cloud economics. Rather than relying solely on the scale of a GPU fleet, the approach seeks to match individual inference stages to different chip architectures while containing operational complexity at the facility level.

If the company executes its stated build-out, it could become a source of demand for high-density data-center infrastructure, liquid cooling, network systems and power capacity. The disclosed plans, however, provide no detail on the facilities, power procurement or grid connections required to support the proposed expansion.