Lambda sells access to Nvidia GPU computing for model training and inference, serving individual AI builders as well as enterprises and hyperscale customers. The company said it closed a $1.01 billion senior secured term loan to buy and develop the servers and related infrastructure for three committed deployments across multiple data centers.
A customer can launch a small GPU instance within minutes, reserve an interconnected cluster for a larger job or contract for a dedicated supercluster containing thousands of processors. For large installations, Lambda can manage the hardware and orchestration layer while the customer runs training and inference workloads through Kubernetes or Slurm.
The new loan is a delayed-draw facility, meaning Lambda can borrow as the financed clusters reach commissioning milestones instead of taking the entire amount before the equipment enters service. It carries a 6.78% fixed rate, amortizes through May 2033 and is secured by the funded infrastructure and associated customer cash flows.
According to Lambda, the three deployments are committed by two investment-grade hyperscale customers. The same contracts and GPU systems that create the revenue also support the debt, tying the financing to a small set of identifiable installations rather than Lambda’s general expansion plans.
Lambda closed a separate $926 million term loan on August 27 for one committed customer deployment. That earlier facility carried floating pricing of SOFR plus 3 percentage points and matured in 2030; the latest loan fixes the rate for longer and spans two hyperscale customers.
Lambda’s product range explains the financing need. Small users can rent processors as needed, but dedicated superclusters require GPUs, networking and data-center capacity to be installed before customers can use them; the delayed-draw loan funds that buildout against contracted demand.
Published versions remain available when an article is updated or corrected.
- Revision 1 · Initial publication
Initial publication