The crush of traffic going into training and running AI has quickly turned into a major cost and resource headache for organisations. Today, Cast AI — a startup building tools to ease and optimise AI and other workloads with automation — is raising a major round of funding on the back of its strong growth and partnerships with major players in the space.
The company has raised $108 million — a Series C that it will be using both for more R&D as well as to expand its business both in core markets like the U.S., as well as elsewhere. We understand from sources that the round puts the company’s post-money valuation at “near unicorn” valuation — close to $900 million from what I understand.
“It’s all about GPU, compute and electricity,” said Yuri Frayman, Cast’s CEO and co-founder. “Our play is to ensure that we create efficiency, to be able to promote more workloads across GPUs. That is what we are about.”
(To put that valuation into some context, when Cast raised its last funding, $35 million, in November 2023, it was valued at $300 million post-money, per PitchBook data. The startup prior to this latest round raised just over $86 million.)
Cast AI is officially based out of Miami, Florida but “is heavily situated in Europe” and is described as “a European company” by Frayman, with most of its development out of Lithuania, as well as Poland, Romania and Bulgaria.
It has amassed 2,100 customers in the last three years of business. Companies like Akamai, BMW, FICO, HuggingFace, NielsenIQ, and Swisscom are among those using its technology to analyse an organization’s cloud and on-premise capacity, to find the optimal cost-performance ratio around how to distribute compute workloads across them. Frayman says that it integrates with all major cloud providers and anything else that a customer may be using.
At a time where companies are facing a shortage of processors to train and run AI models, the need for better resource allocation is a strong one. Cast, citing its own research, claims that on average only 10% of CPUs and 23% of memory are utilized, and the same extends out to GPU usage.
This Series C — both in size and participants — underscores what else it is working on, and whom else it is working with.
G2 Venture Partners and SoftBank Vision Fund 2 are co-leading the round, with Aglaé Ventures (LVMH chairman and CEO Bernard Arnault’s investment firm), and previous backers Hedosophia, Cota Capital, Vintage Investment Partners, Creandum, and Uncorrelated Ventures also participating.
Notably, Frayman pointed out that the oversubscribed round puts the company into the same portfolio stable as OpenAI and the AI infrastructure provider Crusoe Energy — two companies that are, with SoftBank, Oracle and others, working on the massive Stargate AI infrastructure project out of the U.S. Frayman said that his company counts a number of these companies as partners and customers already.
“We are partnering with Crusoe, where we’re inside their stack, and we are partnering with SoftBank to be able to facilitate the efficiency in their AI datacenters,” he said, adding that it is also part of the large project between OpenAI and SoftBank to build services in Japan. “We are partnering with the entire ecosystem,” he added.
Cast AI is talking and doing a lot with AI these days, but that was not exactly where the company got its start. Ukraine-born Frayman, who founded the company with Leon Kuperman and Laurent Gil in 2019, started his career in finance before pivoting to software development.
Back in 2006, he and Gil built what Frayman described to me as one of the “earliest machine learning startups” — Viewdle. There they build some of the earliest applications of using Nvidia’s GPUs to train its classifiers for image searches. “That’s how far back we go in terms of understanding the power of machine learning,” he said.
That company would eventually get acquired by Google.
Along with Kuperman the three founders later worked on a cloud-based cybersecurity startup, Zenedge, which was the inspiration for Cast: there, they struggled to keep cloud costs under control as they scaled up. (Zenedge was eventually acquired by none other than Oracle.)
The first use case for Cast came out of their experience with that resource struggle, and while it has always had “AI” in its name and ethos, it was about the application of it, specifically make cloud use and allocation more efficient for Kubernetes workloads.
Kubernetes applications are still at the heart of Cast, Frayman said, both in terms of revenues and ethos. (And if you go to its site, that is the prominent messaging there, too.) But it is the surge of activity around AI where all the buzz and growth are coming from at the moment, from customers and investors.
“Cast AI is setting a new standard for cloud efficiency at a time when infrastructure demands are surging,” said Tim Yap, Investment Director at SoftBank Investment Advisers, in a statement.
“Right now in the world, everyone is talking about AI agents,” said Carl Fritjofsson, general partner at Creandum. “Cast was was an AI agent before we started talking about that technology, you know. They’ve just been building these type of automation for a long time.”
