Lambda

Lambda (formerly Lambda Labs) is a leading AI cloud infrastructure company headquartered in San Francisco, California, providing GPU compute for AI training, fine-tuning, and inference workloads. The company raised over $1.5 billion in its Series E round in November 2025, led by TWG Global with participation from the U.S.

Innovative Technology Fund, and is targeting a public market debut in the second half of 2026 with pre-IPO financing led by Mubadala Capital. Lambda's largest customer is NVIDIA itself, which leases back 18,000 GPUs for $1.5 billion—a unique validation of Lambda's infrastructure quality.

Additional major customers include Microsoft (a multi-billion-dollar, multi-year agreement to deploy tens of thousands of NVIDIA GPUs), Writer, Sony, Samsung, Pika Labs, and Intuitive Surgical. Lambda publishes its on-demand rates rather than quoting them on request.

Checked on 30 July 2026, its list prices were $3.99–$4.29 per GPU-hour for H100 SXM, $6.69–$6.99 for B200 SXM6 (180GB), and $2.79 for A100 SXM 80GB, across more than a dozen GPU types. The platform offers on-demand GPU instances, reserved clusters, and a developer-friendly API with no lock-in.

Lambda serves AI researchers, startups, and enterprises that need high-throughput training compute without the complexity of hyperscaler contract negotiations. Its focus on transparent pricing, hardware availability, and ML-optimised networking makes it particularly popular with AI research teams building foundation models and large-scale fine-tuning pipelines.

What separates Lambda from CoreWeave and Nebius is that a single engineer can put a card on a credit card without talking to anyone: instances are self-serve and first-come, with multi-node clusters the only tier that routes through sales. That makes it the closest thing in this category to a hyperscaler substitute for research teams, and its transparent published list is a genuine differentiator against quote-on-request rivals — though on those 30 July 2026 figures it is a cheaper on-demand option rather than a dramatically cheaper one, and anyone repeating an older price should recheck the page.

That Series E set the valuation at around $5.9 billion, and the pre-IPO round is reported at roughly $350 million. Buyer caveats: self-serve and first-come means popular GPU types can be unavailable in a given region at the moment you need them, which is the trade for having no minimum commitment; and the NVIDIA relationship noted above cuts both ways, since NVIDIA is at once an investor, the supplier of the hardware and the largest customer — read it as concentration as much as validation.

Best fit: research teams and startups that value immediate access and posted prices over negotiated rates and SLAs.