AIWiki
Malaysia
Back to all articles
Tools & Platformsneocloudgpu-cloudai-infrastructure

Lambda

4 min readUpdated October 2026
Lambda
Type
AI cloud infrastructure provider (neocloud)
Founded
2012, by deep learning engineers
Core products
GPU cloud instances, GPU clusters, superclusters
Key hardware
NVIDIA H100, H200, B200, GB200 NVL72
Funding
US$1.5B Series E (Nov 2025); ~US$5.9B valuation
Revenue
~US$760M ARR in 2025 (+79% YoY)
Status
Planning IPO; reported US$4B pre-IPO raise (Oct 2026)
Related
CoreWeave, neocloud, GPU cluster, NVIDIA
Lambda (also branded Lambda Labs) is a United States cloud-infrastructure company that provides on-demand GPU compute — virtual machines, clusters and fully assembled supercomputers — for training and serving artificial-intelligence models. Founded in 2012 by deep-learning engineers and grown from a maker of deep-learning workstations into one of the largest independent neocloud providers, Lambda competes with hyperscalers on price and queue times for frontier-grade NVIDIA hardware.[1]

History and business model

Lambda began as a small company selling and renting machines configured for deep learning at a time when configuring GPU servers was still specialised work. As demand for AI compute outstripped hyperscaler capacity in the generative-AI boom, Lambda expanded from single rented instances into large dedicated clusters and multi-year infrastructure contracts.

The company's growth accelerated sharply from 2025. In February 2025 it raised a US$480 million Series D led by Andra Capital with NVIDIA among the investors, valuing Lambda at roughly US$2.5 billion.[2] In November 2025 a US$1.5 billion Series E lifted estimated valuations to around US$5.9 billion, alongside a multibillion-dollar agreement with Microsoft to provide dedicated GPU infrastructure, and separate deals under which NVIDIA rents tens of thousands of its own chips back from Lambda, totalling around US$1.5 billion.[3][4] The company reported roughly US$760 million in annual recurring revenue for 2025, up 79% year on year, and in October 2026 was reported to be targeting a US$4 billion final raise ahead of a planned initial public offering.[5]

Key products

  • Lambda Cloud — on-demand GPU instances (H100, H200 and newer NVIDIA generations) for training and inference, positioned as faster and often cheaper to access than hyperscaler quota.[1]
  • Dedicated clusters and superclusters — reserved multi-GPU systems including NVIDIA GB200 NVL72 racks, sold to enterprises and model developers under long-term contracts.[3]
  • Lambda Stack — pre-built GPU server images with CUDA, PyTorch and deep-learning tooling, a holdover from its workstation era that remains popular with research teams.
Lambda's strategic bet is that scarce frontier compute remains scarce: as long as NVIDIA supply is constrained, an independent provider that secures large GPU allocations and resells them with simple pricing can grow faster than the hyperscalers' overall capacity.[4]

Why it matters for AI

Lambda is one of the clearest tests of the neocloud thesis — that purpose-built AI infrastructure companies can win a durable share of the market against AWS, Azure and Google Cloud. Its Microsoft and NVIDIA contracts show hyperscalers and chipmakers alike are willing to route capacity through independents, while its planned IPO makes it a public-market bellwether for AI-infrastructure spending beyond the mega-cap names.[5]

>See Also

🇲🇾Malaysian Context

🇲🇾 Relevance to Malaysia: Lambda does not operate a Malaysian cloud region, so local customers typically reach its GPUs the way they reach most foreign clouds — with international payment cards and cross-border data flows. For Malaysian startups and universities, that makes Lambda an occasional burst-capacity option rather than a default: day-to-day work more often runs on hyperscaler Malaysia regions (AWS ap-southeast-5, Oracle's Kulai region), YTL AI Cloud's local NVIDIA GB200 capacity, or GPU access arranged through MDEC and National AI Office compute initiatives.

Lambda nonetheless matters to Malaysia as a market signal: global demand for independent GPU capacity is what underwrites the data-centre build-out in Johor and the Klang Valley, and Lambda's IPO will be closely watched by Malaysian investors and by operators such as YTL Power positioning domestic GPU clouds against foreign rental pricing. Cross-border use also engages the Personal Data Protection Act — personal data sent to a US GPU cloud is a transfer outside Malaysia and requires the safeguards PDPA places on offshore processing.

References

  1. ↑Lambda. Investor information — founded by deep learning engineers. https://lambda.ai/investors
  2. ↑Reuters. (2025). AI cloud startup Lambda raises $480 million in new round, NVIDIA among investors. https://www.reuters.com/technology/artificial-intelligence/ai-cloud-startup-lambda-raises-480-million-new-round-nvidia-among-investors-2025-02-19/
  3. ↑CNBC. (2025). Lambda AI and Microsoft and NVIDIA deal. https://www.cnbc.com/2025/11/03/lambda-ai-microsoft-nvidia.html
  4. ↑TechCrunch. (2025). AI data center provider Lambda raises a whopping $1.5B after multibillion-dollar Microsoft deal. https://techcrunch.com/2025/11/18/ai-data-center-provider-lambda-raises-whopping-1-5b-after-multibillion-dollar-microsoft-deal/
  5. ↑Reuters. (2026). Nvidia-backed Lambda targets $4 billion raise ahead of planned IPO, WSJ reports. https://www.reuters.com/technology/nvidia-backed-lambda-targets-4-billion-raise-ahead-planned-ipo-wsj-reports-2026-10-06/
Back to all articles