Lambda on the hunt for 'another $800M' to fuel its GPU cloud
Why sell shovels when you can rent them
In the AI gold rush, if you can't be the one selling the GPUs then the next best thing could be to rent them. This week, we learned that Lambda is seeking $800 million in funding to do just that.
Founded in 2012, San Jose-based Lambda is no stranger to accelerated computing, having got its start building systems specifically for machine-learning R&D. It later expanded into colocation services before launching a GPU cloud service in 2018.
According to a Financial Times report, citing people familiar with the matter, term sheets for the impending funding round are expected by mid-July and JPMorgan is helping to coordinate the affair. The capital would be used to purchase additional Nvidia GPUs and associated network infrastructure and software, and to hire staff.
Leasing large quantities of GPUs, particularly for those training custom models, has become a lucrative business in recent years. As our sibling site The Next Platform recently determined, a cluster of 16,000 H100s costing $1.5 billion to purchase, deploy, and network would generate roughly $5.27 billion in revenue over the course of four years.
Of course, to play this game you still need a lot of capital to buy the GPUs in the first place — something that, so far, Lambda has had no shortage of success doing. Back in April, it secured half a billion dollars in debt financing to purchase "tens of thousands" of Nvidia's fastest accelerators, which served as collateral for the loan.
The debt financing came on top of the $320 million series-C funding round it announced back in February, the majority of which is also going toward Nvidia GPUs. The Register reached out to Lambda for comment regarding the reported funding round.
- SK hynix pumps billions into HBM chips to meet AI demand
- AI to boost datacenter capex by 28.5% and become the top server workload
- AI query optimization in IBM's Db2 shows you can teach a tech dinosaur new tricks
- Alibaba Cloud reveals its datacenter design, homebrew network used for LLM training
While $800 million may sound like a lot of capital for an upstart biz, it's far from the biggest figure we've seen in this field lately. Many AI outfits have seen their valuations skyrocket as hype over neural networks reaches new heights.
In May, CoreWeave, another bit barn peddling cut-rate GPU rentals, scored $1.1 billion in a series-C round. That same month, the cloud provider used its enormous collection of GPUs as collateral for a $7.5 billion loan backed by Blackstone, BlackRock, and others.
Meanwhile, similar operations such as Voltage Park and TensorWave have looked to recreate Lambda and CoreWeave's successes. However, it's not just AI infrastructure vendors that have seen their valuations take off.
Back in May, so-called data foundry Scale AI, which provides high-quality datasets used in AI training, saw its valuation touch $14 billion after VC firms and AI heavyweights, such as Nvidia, Amazon, and Meta, injected $1 billion into the firm.
Some feel we're heading into bubble-bursting territory. That's either wishful thinking – because the hype may only just be getting started – or a shallow acknowledgment that all good things eventually come to an end. Nvidia's market cap dipped $500 million last week and that was largely inconsequential: The GPU titan's stock price is holding steady today, being up nearly eight percent this past month and up more than 150 percent in the past half-year, and its market cap is sitting pretty north of $3 trillion. ®