What is Lambda?
The AI-only GPU cloud and supercomputing platform for training, fine-tuning, and deploying frontier models.
- Category
- AI Infrastructure / GPU Cloud
- Headquarters
- San Jose, California
- Founded
- 2012
- Employees
- ~750
- Total Funding
- $3.1B+ (equity + debt)
- Valuation
- $5.9B (Nov 2025 Series E)
What is Lambda?
Lambda is an AI-only GPU cloud and supercomputing platform founded in 2012 and headquartered in San Jose, California. The company builds and operates hyperscale 'AI factories' — dense, liquid-cooled data centers packed with NVIDIA GPUs — and rents that compute to hyperscalers, frontier AI labs, enterprises, and researchers who need massive scale without the friction of building their own infrastructure.
Lambda's cloud GPU rental business generated over $520 million in revenue over the trailing twelve months ending September 2025, with its annualized revenue run rate reaching approximately $760 million by late 2025 — up 79% year-over-year. The company serves over 200,000 AI developers and teams, including enterprise customers such as Apple, Amazon, Microsoft, MIT, Stanford, Harvard, Caltech, and the U.S. Department of Defense. In a signal of just how supply-constrained high-end GPU compute remains, NVIDIA itself is a customer — leasing back approximately 18,000 chips through a multi-year $1.5 billion agreement signed in September 2025.
Unlike general-purpose hyperscalers, Lambda dedicates 100% of its engineering, operations, and support to AI workloads. Its flagship products — on-demand GPU instances, 1-Click Clusters (16 to 2,000+ GPUs deployable in minutes), and private cloud deployments — run exclusively on the latest NVIDIA architectures including H100, H200, B200, and GB300 NVL72 Blackwell Ultra systems. Lambda achieved NVIDIA Exemplar Cloud status in October 2025, guaranteeing performance within 5% of published benchmarks for large-scale training workloads — a distinction it holds as one of the first GPU cloud providers worldwide to earn the certification.
A landmark multi-billion-dollar, multi-year agreement with Microsoft announced in November 2025 to deploy tens of thousands of NVIDIA GPUs in Lambda's liquid-cooled U.S. data centers cemented the company's position as an infrastructure partner to the world's largest technology companies. With a $1.5 billion Series E closed in November 2025 at a $5.9 billion post-money valuation, a $350 million pre-IPO convertible notes round led by Mubadala Capital in early 2026, and investment banks Morgan Stanley, J.P. Morgan, and Citi retained, Lambda is actively preparing for a public offering targeted in the second half of 2026.
What does Lambda offer?
Lambda's product portfolio spans on-demand GPU instances, reserved and private clusters, bare-metal instances, and the Lambda Stack software environment — all purpose-built for AI workloads.
- On-Demand GPU Instances· Cloud Compute
- 1-Click Clusters (16–2,000+ GPUs)· Cloud Compute
- Bare Metal Instances· Cloud Compute
- Private Cloud Deployments· Cloud Compute
- Superclusters· Cloud Compute
- NVIDIA GB300 NVL72 Blackwell Ultra· Hardware Platform
- NVIDIA H100 / H200 / B200· Hardware Platform
- Lambda Stack (CUDA / PyTorch / TensorFlow)· Software
- Managed Slurm· Software
- Cloud Metrics Dashboard· Software
- Filesystem S3 Adapter· Software
- AI Factory Design & Build· Services
- Professional Services· Services
How does Lambda make money?
Lambda makes money primarily by renting NVIDIA GPU compute by the hour or under reserved contracts, operating a capital-intensive cloud infrastructure model with approximately 50% gross margins overall (and roughly 61% excluding legacy hardware lines).
The core revenue engine is GPU cloud rental across several pricing tiers. On-demand H100 SXM instances run $2.99–$3.29/GPU-hour in 1x-to-8x configurations; H100 PCIe instances are available at lower rates. Newer B200 instances are priced at approximately $4.99–$5.29/GPU-hour, and 1-Click Clusters — which require a minimum two-week commitment — are priced around $6.16/GPU-hour for H100 configurations supporting 16 to 2,000+ interconnected GPUs. Lambda's zero-egress policy is a meaningful differentiator for data-intensive teams: while AWS charges $0.09/GB and Google Cloud $0.12/GB, Lambda transfers data out at no cost, saving large teams $437–$550 per month or more. Reserved compute contracts (multi-month to multi-year) carry lower per-hour rates but lock in capacity and underpin Lambda's ability to invest in GPU procurement ahead of customer demand.
A second revenue layer comes from large-scale enterprise and hyperscaler agreements structured as multi-year, committed take-or-pay contracts. The Microsoft agreement (announced November 2025) and the NVIDIA leaseback deal ($1.5B over four years) exemplify this model: Lambda deploys the capital to build and operate the data centers, then earns a durable fee stream. These deals require enormous upfront investment — hence the debt facilities, equity rounds, and $275M credit facility added in August 2025 — but generate predictable long-term cash flow once the hardware is deployed.
Growth is driven by three structural forces: NVIDIA GPU supply constraints that make Lambda one of the few reliable access points for frontier silicon; the increasing appetite of frontier AI labs and enterprises for dedicated large-scale clusters; and Lambda's ability to rapidly redeploy existing infrastructure to new GPU generations. The company's gross margin of approximately 50% (61% excluding legacy hardware) is healthy for a capital-intensive infrastructure business, and Lambda ran a net loss of approximately $24 million in H1 2025 on revenues exceeding $250 million — deliberately reinvesting into data center capacity ahead of its planned IPO.
Who leads Lambda?
Lambda's co-founders Stephen and Michael Balaban built the company for over a decade, and in May 2026 brought in seasoned infrastructure operator Michel Combes as CEO, shifting Stephen to CTO and Michael to CPO as the company prepares for a public offering.
- Michel CombesChief Executive OfficerMay 2026–presentFormer CEO of SoftBank International, Sprint, and Alcatel-Lucent; also Chairman/CEO of Brightspeed. Brings two decades of capital-intensive infrastructure operating experience; hired to lead Lambda's gigawatt-scale expansion and IPO.
- Stephen BalabanCo-Founder & Chief Technology Officer2012–presentMachine learning researcher who co-founded Lambda in 2012; published at NeurIPS and SPIE; served as CEO from founding through April 2026 before transitioning to lead technology and product vision as CTO.
- Michael BalabanCo-Founder & Chief Product Officer2012–presentTwin brother of Stephen; scaled Lambda's product and operations from early GPU workstations to hyperscale clusters; now leads product strategy connecting infrastructure to developers.
- John DonovanChairman of the Board2026–presentFormer CEO of AT&T Communications and CTO of AT&T; brings deep expertise in large-scale network and infrastructure deployments critical to Lambda's gigawatt buildout.
- Jerry HunterVice Chairman of Compute Delivery2026–presentFormer COO of Snap and early AWS infrastructure leader; brings hyperscale data center operations experience directly relevant to Lambda's compute delivery function.
- Charles FisherChief Financial Officer2026–presentFormer CFO of Turo and senior finance executive at Charter Communications; leads Lambda's financial operations ahead of the planned IPO.
- David ConnollyChief Legal Officer2026–presentFormer General Counsel at Altice; oversees Lambda's legal and compliance functions as it scales to gigawatt-class infrastructure and public-company readiness.
How do you contact Lambda's leadership?
Lambda's primary email domain is lambda.ai; internal executive contacts follow the firstname@lambda.ai pattern. The legacy press email domain is lambdal.com. For official media outreach use pr@lambdal.com or the lambda.ai/contact form.
firstname@lambda.aiHow much funding has Lambda raised?
Lambda has raised over $3.1 billion in disclosed equity and debt across seven major equity rounds since 2012, with a $5.9 billion post-money valuation set by its November 2025 Series E — making it one of the most heavily capitalized private AI infrastructure companies globally as it approaches a planned H2 2026 IPO.
Lambda's earliest capital came from seed tranches totaling roughly $4 million between 2015 and 2018, backing the company's transition from GPU hardware sales to cloud. A $15 million Series A equity round closed in July 2021 (alongside a $9.5 million debt facility), with investors including Gradient Ventures, Bloomberg Beta, 1517 Fund, Razer, and Georges Harik, valuing the company at $87.5 million. A November 2022 venture round added $39.7 million, followed by the $44 million Series B in March 2023 led by Mercato Partners at a $205 million valuation, with angel participation from Greg Brockman, Garry Tan, Adam D'Angelo, Jeff Hammerbacher, and Lukas Biewald.
The pace of capital dramatically accelerated with the AI compute boom. In February 2024, USIT (Thomas Tull's US Innovative Technology Fund) led a $320 million Series C at a $1.5 billion valuation — Lambda's unicorn milestone — with participation from B Capital, SK Telecom, T. Rowe Price, Crescent Cove, and existing investors. Lambda then added a $500 million GPU-backed debt facility in April 2024. In February 2025, Andra Capital and SGW co-led a $480 million Series D at a $2.5 billion valuation, with notable new strategic investors including NVIDIA, Andrej Karpathy, ARK Invest, In-Q-Tel, Supermicro, Pegatron, Wistron, and Wiwynn. Lambda added a further $275 million credit facility in August 2025.
The Series E in November 2025 was Lambda's largest single round: TWG Global — a $40 billion investment firm co-founded by Thomas Tull and Mark Walter and anchored by Abu Dhabi's Mubadala Capital — led a $1.5+ billion raise at a reported $5.9 billion post-money valuation, following the multibillion-dollar Microsoft deal. In early 2026, Lambda added $350 million in pre-IPO convertible notes led by Mubadala Capital, priced at approximately a 20% discount to the expected IPO price, with financial penalties if the listing does not occur within one year. Investment banks Morgan Stanley, J.P. Morgan, and Citi have been retained ahead of the H2 2026 IPO.
How did Lambda get here?
Lambda's journey ran from a face-recognition API side project in 2012, through a GPU hardware phase, to becoming a multi-billion-dollar hyperscale AI infrastructure provider on the cusp of an IPO.
- 2012Founded — facial recognition APITwin brothers Stephen and Michael Balaban co-founded Lambda in 2012. After Facebook acquired Face.com and shut its API, Lambda launched an open-source facial recognition alternative, quickly amassing thousands of developer signups. The founders recognized that the larger opportunity lay in the underlying GPU compute powering these workloads.
- 2017–2018GPU Workstations and Lambda Stack launchLambda introduced sub-$20K deep learning GPU workstations in 2017 and launched Lambda Cloud and Lambda Stack in 2018, giving ML teams a pre-configured CUDA/PyTorch/TensorFlow environment and early cloud compute access.
- July 2021Series A — $15M equity; $87.5M valuationGradient Ventures, Bloomberg Beta, 1517 Fund, Razer, and Georges Harik backed the Series A alongside a $9.5M debt facility, enabling Lambda's pivot to a cloud-first model.
- March 2023Series B — $44M; $205M valuationLed by Mercato Partners; angels include Greg Brockman, Garry Tan, Adam D'Angelo, Jeff Hammerbacher, and Lukas Biewald. Lambda had reached $250M in 2023 annual revenue as demand for H100 compute exploded post-ChatGPT.
- February 2024Series C — $320M; $1.5B valuation (unicorn)Led by Thomas Tull's USIT, with B Capital, SK Telecom, T. Rowe Price, and Crescent Cove. Lambda's unicorn milestone, funding rapid H100 cluster expansion.
- February 2025Series D — $480M with NVIDIA and Andrej KarpathyAndra Capital and SGW co-led at a $2.5B valuation. NVIDIA's strategic investment, alongside hardware ODMs Supermicro, Pegatron, Wistron, and Wiwynn, validated Lambda's supply-chain relationships and access to next-generation GPU architectures.
- September 2025NVIDIA $1.5B leaseback deal — 18,000 GPUsNVIDIA signed a four-year, $1.5 billion leaseback agreement to rent 18,000 of its own GPUs from Lambda's infrastructure, making NVIDIA Lambda's largest customer and validating Lambda's scale.
- November 2025Microsoft multibillion-dollar deal + $1.5B Series ELambda announced a multi-year, multibillion-dollar agreement to deploy tens of thousands of NVIDIA GPUs for Microsoft, followed by a $1.5B+ Series E led by TWG Global at a ~$5.9B valuation. NVIDIA Exemplar Cloud status certified.
- May 2026Leadership overhaul; IPO preparationsLambda appointed Michel Combes (ex-Sprint CEO) as CEO, named John Donovan as Chairman, added Jerry Hunter, CFO Charles Fisher, and CLO David Connolly. Closed a $350M pre-IPO convertible note round led by Mubadala; Morgan Stanley, J.P. Morgan, and Citi retained. H2 2026 IPO targeted.
Who are Lambda's competitors?
Lambda competes across two tiers: GPU-specialist clouds (CoreWeave, RunPod, Crusoe) and the general-purpose hyperscalers (AWS, Google Cloud, Azure), each with different scale, price, and developer-experience trade-offs.
- CoreWeaveThe closest pure-play peer — NVIDIA-heavy and AI-only — now public (NASDAQ: CRWV, ~$64B market cap as of June 2026) after its March 2025 IPO, with deeper Kubernetes orchestration and enterprise SLAs; Lambda counters with lower per-GPU pricing, zero-egress policy, and a developer-first positioning.
- Amazon Web ServicesBroadest ecosystem and global footprint, but GPU availability is constrained and pricing is typically 40–100% higher per GPU-hour; enterprises already deep in the AWS ecosystem often pay the premium for integration rather than switching to a specialist cloud.
- Google CloudStrong on TPU access and TensorFlow-native tooling; competitive on H100 pricing but includes egress fees ($0.12/GB) and sells primarily into customers already in the Google ecosystem. Google's own AI research demand creates tension with external availability.
- Microsoft AzureDeeply integrated with OpenAI and the enterprise Microsoft stack; ironically also a Lambda customer under their multibillion-dollar supply deal — illustrating how hyperscalers are increasingly leasing AI infrastructure from specialists rather than building all capacity internally.
- RunPodPeer-to-peer GPU marketplace with highly competitive self-serve H100 rates ($1.99/hr PCIe) and spot instances, but more variable reliability and no enterprise SLAs; appeals to cost-sensitive individual researchers and small teams who tolerate interruption risk.
- Crusoe EnergySustainability-differentiated GPU cloud powered by stranded or renewable energy; growing fast among ESG-conscious enterprises and regulated-industry buyers where carbon footprint is a procurement criterion.
Lambda — frequently asked questions
