Lambda는 Microsoft의 AI 워크로드 수요 가속화에 대응해 GPU 클라우드 용량을 확대하기 위해 10억 달러 규모의 추가 부채 조달을 완료했다.

San Jose-based Lambda has secured approximately $1 billion in short-term private debt to purchase additional Nvidia GPUs for a major computing agreement with Microsoft, according to individuals familiar with the transaction. JPMorgan Chase structured the offering for private placement investors, and the company expects lease revenue from corporate clients to service the debt through operational cash flow.

Closed this week, the financing marks the latest in a series of debt transactions the San Jose-based neocloud provider has leveraged to fuel its rapid expansion, as reported by BiGGO Finance. Lambda rents computing chips and AI infrastructure to enterprise clients and intends to direct the new capital exclusively toward Nvidia GPUs slated for leasing to Microsoft. While the company has maintained a vendor relationship with Microsoft since 2018, it substantially scaled the partnership in November with a multibillion-dollar, multi-year agreement to deploy tens of thousands of Nvidia GPUs, including liquid-cooled GB300 NVL72 supercomputing clusters, according to Lambda.

Lambda plans to deploy the newly acquired chips rapidly. The loan structures are explicitly tied to designated customer deployments, linking the debt directly to specific hardware allocated for particular client contracts. Per the BiGGO Finance report, the purchased GPUs themselves serve as collateral. Representatives for Microsoft, Nvidia, and Lambda did not immediately respond to requests for comment regarding the arrangement.

Founded in San Jose in 2012 by twin brothers Stephen and Michael Balaban, Lambda initially operated as a facial-recognition software developer. According to Forbes, the founders first constructed an internal GPU cloud to reduce their own Amazon Web Services costs before pivoting in 2017 to commercialize that infrastructure for external AI developers. This strategic shift established Lambda as one of North America’s leading neocloud providers, operating on a specialized model centered on procuring cutting-edge Nvidia hardware and leasing compute capacity to large enterprises.

Lambda has simultaneously expanded its physical footprint across the Bay Area. Last August, the company leased a 20,000-square-foot office building in downtown San Francisco, joining a broader wave of AI-driven commercial real estate activity documented by Hoodline. In September 2025, Lambda deployed zero-emission, hydrogen-powered Nvidia GB300 NVL72 supercomputing systems at ECL’s data center facility in Mountain View, leveraging off-grid hydrogen fuel cells to circumvent local power-grid constraints.

This week’s approximately $1 billion financing is neither Lambda’s first major debt transaction this year nor its largest. On August 27, the company closed a $926 million senior secured Term Loan B facility led by Morgan Stanley and MUFG, specifically earmarked for the acquisition and installation of Nvidia GB300 GPUs. The facility received a Baa2 investment-grade rating from Moody’s, was priced at SOFR plus 3.00%, and matures in December 2030. It is reportedly the first broadly syndicated investment-grade Term Loan B executed by a private neocloud provider.

Prior to that transaction, Lambda closed a $1 billion secured credit facility in May 2026. Going further back, the company pioneered the GPU asset-backed debt market in April 2024 by securing a first-of-its-kind $500 million special-purpose financing facility led by Macquarie Group, which enabled the purchase of Nvidia hardware without mandating long-term customer lock-in contracts. Executive leadership is currently helmed by CEO Michel Combes, a former Alcatel-Lucent chief executive and SoftBank veteran, who guides the executive team alongside co-founders Stephen and Michael Balaban as the firm transitions into institutional debt markets.

Lambda’s reliance on debt financing is underpinned by rapid top-line growth. The company reported $760 million in annual recurring revenue for 2025, representing a 79% year-over-year increase, and forecasts revenue exceeding $1.5 billion in 2026. According to PitchBook data, Lambda achieved a $5.43 billion post-money valuation during a November 2025 venture round following a $1.5 billion equity raise that same month. The company is reportedly negotiating to secure up to $3 billion in pre-IPO funding as it prepares for a potential public market debut next year.

Lambda’s strategy reflects a broader industry trend. By early 2026, specialized neocloud providers—including CoreWeave, Lambda, Crusoe, and Fluidstack—had accumulated over $20 billion in outstanding private loans primarily collateralized by Nvidia GPUs, according to a Substack analysis by Dave Friedman. Major Wall Street institutions such as BlackRock, JPMorgan, Carlyle, and PIMCO have emerged as active participants in this GPU-backed lending space. Furthermore, banks and technology firms have collectively raised more than $400 billion in AI-related debt globally in 2026, according to Bloomberg data.

This surge in borrowing introduces risks related to hardware depreciation. As noted in the same Substack analysis, spot rental prices for Nvidia H100 GPUs fell between 50% and 70% from early 2024 levels, dropping to $2–$4 per hour by late 2025 amid the market entry of newer architectures like the GB300. Rapid product release cycles can accelerate the economic obsolescence of older hardware generations serving as loan collateral—a vulnerability that directly impacts Lambda, given that projected lease revenues form the foundation of its debt repayment strategy. Additionally, the BiGGO Finance report indicates Lambda holds a separate contracted project with Nvidia, though specific terms remain undisclosed.

The structure also provides a strategic workaround for Microsoft. The technology giant secures access to constrained Nvidia supply without absorbing the full hardware cost on its balance sheet upfront, while Lambda assumes the operational burden of deployment and maintenance. As Lambda advances toward a potential 2027 initial public offering, its increasing dependence on private placements and term loan facilities mirrors a wider Silicon Valley evolution—one where physical compute capacity is being capitalized similarly to traditional utility and real estate infrastructure.