Lambda Closes $1B Fixed-Rate AI Infrastructure Loan

A GPU can move from frontier hardware to a prior generation long before a conventional infrastructure loan reaches maturity. Lambda has placed that mismatch inside a $1.008B financing structure designed to release capital only as new clusters enter service and begin supporting contracted customer deployments.

The AI infrastructure company announced the senior secured delayed-draw term loan on October 1, 2026. The facility carries a 6.78% fixed interest rate, matures on May 30, 2033, and will finance more than 30,000 NVIDIA GB300 and VR200 GPUs across data centers in Seattle, Kansas City, and Dallas. Those systems are tied to three committed deployments for two highly rated counterparties.

The transaction matters because the financing case rests on more than Lambda's corporate growth. Lenders are underwriting GPU assets, commissioning milestones, service obligations, and the contracted cash flows around them. That moves part of the AI buildout closer to project finance, where infrastructure is funded against a defined operating schedule and a specific source of repayment.

What Lambda Financed

The official headline rounds the transaction to $1B, while the facility's exact size is $1.008B. It is a senior secured, fixed-rate term loan with delayed-draw mechanics, meaning Lambda can take capital in stages as the associated GPU clusters are commissioned. The loan then begins amortizing after each GPU tranche stabilizes, rather than treating every dollar as if it entered service on closing day.

Lambda said the 6.78% coupon is paid semi-annually and that the facility was oversubscribed. Morningstar DBRS assigned the borrower and loan an A (low) rating with stable trends, while Lambda reported a Baa1 rating from Moody's. J.P. Morgan served as sole coordinating lead arranger, structuring agent, and bookrunner.

The funded assets and contracted cash flows secure the debt. According to Morningstar DBRS, more than 30,000 GPUs will be installed across facilities in Seattle, Kansas City, and Dallas. The rating agency describes take-or-pay contracts with two highly rated counterparties, a structure that limits volume risk because the customers have committed to capacity rather than paying only when they use it.

Why the Delayed Draw Matters

AI compute requires large payments for chips, data-center construction, power systems, cooling, networking, and installation before a customer workload begins producing revenue. A delayed-draw loan narrows that timing gap. Capital is released as deployment milestones are met, allowing Lambda to match borrowing more closely with the moment a cluster becomes an operating asset.

That discipline matters at a 6.78% fixed rate. Idle borrowed money creates cost without revenue, while a fully funded project still faces delivery and acceptance risk. By connecting each draw to commissioning and each amortization schedule to stabilization, the facility converts a broad growth story into a series of measurable infrastructure handoffs.

The two customers remain unnamed, and Lambda has not disclosed how much of the facility was drawn at closing. Those limits matter. The transaction is evidence that rated investors accepted the contract and security package, not proof that every dollar is already deployed or that every cluster has completed acceptance.

The Contract Around the Silicon

The GPUs are valuable collateral, but their economics depend on the agreements surrounding them. Morningstar DBRS cited take-or-pay cash flows, achievable service-level thresholds, limited cancellation rights, lender step-in rights, and a minimum 1.20x debt-service coverage ratio as important protections. The financing also benefits from draw conditions and a fixed-price GPU supply contract with Dell for outstanding tranches.

The risk analysis is unusually physical for a story filed under cloud computing. GPUs still have to arrive, data centers still have to be completed, and new hardware still has to pass customer acceptance. Morningstar DBRS specifically identified the short operating history of NVIDIA's GB300 GPUs and the lack of operating history for VR200 GPUs as risks, alongside installation delays, data-center completion, and Lambda parent-level default during the construction period.

That combination explains the transaction's real market signal. Fixed-income investors are becoming comfortable with AI compute when the underwriting can be attached to named assets, controlled draw schedules, and durable customer contracts. The rating applies to this structured project, not to every GPU cloud provider or every AI infrastructure promise.

Lambda's Expanding Capital Stack

The October financing is Lambda's second institutional credit facility and its third major debt-market step in 2026. It is separate from a $1B syndicated senior secured credit facility announced in May and the $926M term loan B that closed in August. The August facility backed another committed deployment and used a floating-rate structure priced at SOFR plus 3.00%.

Lambda also raised more than $1.5B in Series E equity in November 2025. Equity, bank lending, syndicated term debt, and insurance or fixed-income capital serve different jobs inside the same buildout. The mix gives Lambda more ways to finance capacity without asking one pool of capital to absorb every stage of risk.

The leadership structure has shifted with that scale. Michel Combes is CEO, while co-founders Stephen Balaban and Michael Balaban serve as CTO and CPO. The operating challenge now extends beyond buying hardware: Lambda has to coordinate capital, supply, construction, acceptance, and customer delivery across several large deployments at once.

What This Signals for AI Infrastructure

The AI infrastructure market is developing a financing layer that looks increasingly familiar to lenders accustomed to power, telecom, and other contracted assets. Customer demand becomes useful to capital markets when it is expressed through enforceable commitments, deployment schedules, service levels, and cash-flow protections. Lambda's facility shows one way that translation can happen.

It also shows where the model remains exposed. The useful life of the financing must coexist with rapid hardware cycles, and the stability of contracted revenue must survive installation and operating problems. Two counterparties may diversify one facility beyond a single customer, but the economics still depend on a concentrated set of large buyers keeping their commitments.

Across Seattle, Kansas City, and Dallas, the loan will move from a credit document into more than 30,000 machines, their power and cooling systems, and the customer workloads waiting for them. Each cluster has its own commissioning clock. Lambda's next financial result will be produced rack by rack as those clocks begin matching the repayment schedule.