Lambda Chases $4 Billion, With Anthropic Underwriting the Bet

Lambda, the Nvidia-backed cloud provider that rents out graphics processors by the hour, is reportedly raising as much as $4 billion in what would be its final private round before going public. The report, first carried by The Wall Street Journal and relayed by DataCenterDynamics, puts the company at a $14.5 billion pre-money valuation, with Blackstone and Coatue Management leading. Lambda declined to comment, and none of the figures are confirmed by the company, so they should be read as a snapshot of a deal still in motion.

If the round closes near those terms, the jump is steep. Last November a $1.5 billion round led by TWG Capital valued Lambda at $5.9 billion post-money. The new figure would roughly double that in under a year, and management is said to be aiming at a public listing in 2027, subject to the market cooperating.

One customer, a very large share of the book

The detail that stands out sits in a letter to investors the Journal reviewed. Lambda's contracted backlog grew from about $15 billion in June to roughly $50 billion in September. Most of that increase traces to a single company: Anthropic, which signed a reported $35 billion deal with Lambda in late August.

That is the kind of number that flatters a valuation and quietly reshapes the risk. A backlog anchored to one buyer is only as durable as that buyer's ability to keep paying, and Anthropic is itself spending heavily to train and serve models like its recently shipped Haiku 5.5. The arrangement is a smaller echo of the circular financing that has drawn scrutiny across the sector, where chipmakers, clouds and model labs increasingly fund one another's growth.

Lambda has also been borrowing. The equity talks come shortly after the company announced $1 billion in senior secured fixed-rate financing earlier this month, layering debt onto the balance sheet ahead of the raise.

The money keeps flowing toward compute

Lambda's round fits a pattern that has defined the year. AI companies have absorbed a commanding share of venture capital, and the heaviest spending has gone not into applications but into the data centers and chips underneath them. Investors are betting that demand for inference capacity keeps climbing faster than supply, and that a neocloud with a fat order book is a safer place to park money than the labs burning through it.

Whether that holds depends on names like Anthropic, whose founders have made their own long-term intentions unusually public, down to a charity pledge tied to a future listing. For now, Lambda's pitch to the market is straightforward: the backlog is real, the customers are serious, and the capacity is already spoken for. The question a 2027 IPO will have to answer is how many of those contracts survive contact with a leaner year.