Lambda is in talks to raise up to $3bn in pre-IPO financing at a valuation of $12bn or more, according to Bloomberg, as the Nvidia-backed cloud provider prepares for a listing expected next year.
The raise follows a very different kind of financing earlier in its run, when Lambda borrowed $917mn against its GPUs to buy more chips from the company that also invests in it.
The business is straightforward to describe and expensive to run. Lambda rents out Nvidia accelerators and the surrounding infrastructure to companies training and serving AI models, competing with the hyperscalers on price and with a growing field of specialist neoclouds on availability.
Equity ahead of an IPO does something the leveraged loan could not. Debt secured against depreciating hardware is an unforgiving instrument, and a public market debut is easier to price when the balance sheet is not dominated by borrowings against chips whose resale value nobody can confidently forecast.
Terms are not finalised and the discussions remain confidential, so neither the investor list nor the final size has been reported. Lambda has not commented publicly.
Its relationship with Nvidia is the part worth pausing on. Nvidia is an investor, a supplier, and indirectly a source of demand, an arrangement it has replicated across the sector to the point where its equity commitments passed $40bn this year.
Almost every company receiving that money spends a large share of it on Nvidia hardware, which flatters demand in a way debt markets have started to price. Lambda is a clean example of the structure rather than an outlier within it.
Investors have not treated that circularity as disqualifying so far, on the reasonable view that a supplier taking equity in its customers is only a problem if the underlying demand is not real. The disagreement is about how much of it is.
Customer traction is real, though. Lambda signed a cloud agreement with Hudson River Trading to supply access to Nvidia chips, which is the kind of counterparty that does its own diligence on uptime and cost.
The wider question for any neocloud is what happens when the capacity crunch eases. These businesses earn their margins on scarcity, and the same customers renting GPUs today are building their own silicon or negotiating directly with hyperscalers for the years after.
Timing an IPO into that is delicate. Lambda would be listing into a market that has repriced AI infrastructure more than once, and where lenders have grown noticeably more attentive to who is actually funding the buildout.
Contract structure is where these companies live or die. Long-dated commitments from creditworthy customers make the debt cheap and the equity story simple, while spot rentals to startups leave a fleet of expensive hardware exposed to whatever next year’s funding market looks like.
Rising input costs are not helping the arithmetic either. Nvidia has notified customers of AI server price increases above 15% driven by memory shortages, which lands directly on the cost base of a company whose product is renting those servers out.
Public markets also apply a scrutiny private rounds do not, particularly to depreciation schedules. How quickly a company writes down GPUs determines its reported profitability, and there is no settled convention across the sector for how long an accelerator should be assumed to earn.
A $12bn valuation would put Lambda well below the largest neoclouds while placing it firmly among the companies investors expect to reach public markets. The listing has been anticipated since at least last year, when reports first suggested the company was preparing for it.
The competitive set has changed too, with CoreWeave already public and several rivals in the queue behind it. A listing that once looked like a novelty is now one of several, and comparable trading multiples exist for investors to price against.
Nothing here is settled. The round is in negotiation, the valuation is a target rather than a mark, and an IPO said to be planned for next year has plenty of time to become an IPO planned for the year after.
Get the TNW newsletter
Get the most important tech news in your inbox each week.
