Lambda, the Nvidia-backed provider of cloud computing for artificial intelligence, is raising up to $4 billion at a pre-money valuation of $14.5 billion in what could be its final private funding round before a planned initial public offering in 2027, according to The Wall Street Journal.
The round is being led by hedge fund and growth investor Coatue Management and private-equity group Blackstone. If completed at the top of the range, it would be among the largest private financings this year for a company that rents out graphics processing units, the chips at the heart of AI training and inference.
The raise arrives as Lambda's order book swells. A letter to investors reviewed by the Journal shows the company's contracted backlog grew from $15 billion in June to $50 billion in September, an increase that reflects both the intensity of demand for AI computing capacity and the outsized role played by a single customer.
One customer, one very large contract
Much of that jump stems from a $35 billion commitment from Anthropic, the AI developer behind the Claude family of models, which signed a deal with Lambda in late August. The agreement turned Lambda overnight from a mid-sized player in GPU cloud into one of the more significant suppliers to a frontier AI lab.
That concentration cuts both ways. A contract of that size gives Lambda visibility over years of revenue and makes it easier to raise both equity and debt. But it also means the company's valuation, which has climbed sharply since its last equity round, leans heavily on Anthropic's ability to keep paying, and on the continued growth of the frontier model business more broadly.
Investors appear willing to accept that risk. Reliable GPU capacity remains scarce, and companies that can secure chips, power and data-centre space, and then sign long-term contracts with major AI developers, have become some of the most sought-after assets in technology finance.
For Anthropic, contracts with several infrastructure providers spread risk and secure capacity at a time when demand for its models has been growing quickly and access to the latest chips is constrained. Frontier AI developers have increasingly signed multi-year, multi-billion-dollar agreements with hyperscalers and neoclouds alike, turning compute purchasing into one of their largest strategic decisions.
Lambda's earlier fundraising history shows how quickly its position has changed. In 2025 the company raised $1.5 billion following a multibillion-dollar agreement with Microsoft, at a time when it was one of several specialist cloud providers competing for workloads that the largest hyperscalers could not immediately accommodate.
Equity now, scrutiny later
For specialist cloud providers, often called neoclouds, demand is not the main constraint. The harder problem is financing the enormous upfront cost of meeting it. Building and equipping data centres requires billions of dollars in chips, power infrastructure and facilities, much of which is typically funded with debt secured against customer contracts.
Lambda raised an additional $1 billion of senior secured, fixed-rate financing just last week, adding to earlier debt facilities used to buy chips. But lenders have become choosier about which data-centre projects they fund and on what terms, as the volume of AI-related borrowing has risen and questions have grown about the long-term economics of GPU rental.
Depreciation sits at the centre of that debate. GPUs bought today may be overtaken by faster chips within a couple of years, so a provider's profitability depends on how long it can rent each chip at attractive prices. Long-term contracts with creditworthy customers reduce that risk, which is one reason investors have rewarded neoclouds that sign multi-year agreements with large AI developers.
Raising a large equity round now serves several purposes. It strengthens the balance sheet against which future debt can be raised, reduces dependence on lenders at a time of tightening terms, and sets a reference valuation for the IPO. It also gives Lambda access to capital before the quarterly scrutiny of public markets arrives.
The company had reportedly intended to list this year but pushed back its plans amid market uncertainty. A 2027 debut would see it join other Nvidia-backed neoclouds, such as CoreWeave and Nebius, that are already listed and that depend on the health of their share prices to finance ongoing data-centre construction.
The queue for the public markets
Lambda is not the only AI infrastructure company preparing to test public investors. British neocloud Nscale filed for a US IPO last month after securing $3.36 billion in convertible financing, and is expected to begin trading soon. CoreWeave's performance since its 2025 listing has become a bellwether for the sector, with its shares sensitive to every signal about AI demand, chip supply and financing costs.
Public investors will want to understand several things about Lambda before it lists: how diversified its customer base is beyond Anthropic, how its contracts are structured, the cost and maturity of its debt, and how quickly its GPUs depreciate as Nvidia releases new generations of chips. The economics of GPU rental depend heavily on keeping expensive hardware fully utilised over its useful life.
Founded in 2012, Lambda began by selling GPU workstations and servers to machine-learning researchers before building out its cloud business. Its evolution from a hardware vendor to a provider of contracted AI capacity mirrors the transformation of the wider industry, as the training and deployment of large models has turned computing power into one of the most strategic resources in technology.
Coatue and Blackstone did not immediately respond to requests for comment, and Lambda did not comment on the reported round. If the raise closes as described, it will underline how much private capital is still available for companies that can deliver AI computing at scale, and how closely the fortunes of infrastructure providers are now tied to a small number of frontier AI developers.
For the broader market, Lambda's round is another data point in a debate that shows no sign of resolution: whether the vast sums flowing into AI infrastructure reflect durable demand that will be paid for by real-world applications, or a build-out running ahead of the revenue needed to sustain it. The answer will shape not only Lambda's IPO but the financing conditions for every company in its sector.
Blackstone's participation is also notable. The firm has become one of the largest investors in data centres globally through its property and infrastructure businesses, and backing a GPU cloud operator directly gives it exposure to a different layer of the same AI build-out, one tied to the chips and customer contracts rather than to the buildings themselves.
