The crush of site visitors going into coaching and operating AI has shortly became a significant price and useful resource headache for organisations. Right this moment, Cast AI — a startup constructing instruments to ease and optimise AI and different workloads with automation — is elevating a significant spherical of funding on the again of its sturdy development and partnerships with main gamers within the area.
The corporate has raised $108 million — a Collection C that it is going to be utilizing each for extra R&D in addition to to broaden its enterprise each in core markets just like the U.S., in addition to elsewhere. We perceive from sources that the spherical places the corporate’s post-money valuation at “close to unicorn” valuation — near $900 million from what I perceive.
“It’s all about GPU, compute and electrical energy,” mentioned Yuri Frayman, Solid’s CEO and co-founder. “Our play is to make sure that we create effectivity, to have the ability to promote extra workloads throughout GPUs. That’s what we’re about.”
(To place that valuation into some context, when Solid raised its final funding, $35 million, in November 2023, it was valued at $300 million post-money, per PitchBook information. The startup previous to this newest spherical raised simply over $86 million.)
Solid AI is formally based mostly out of Miami, Florida however “is closely located in Europe” and is described as “a European firm” by Frayman, with most of its improvement out of Lithuania, in addition to Poland, Romania and Bulgaria.
It has amassed 2,100 prospects within the final three years of enterprise. Corporations like Akamai, BMW, FICO, HuggingFace, NielsenIQ, and Swisscom are amongst these utilizing its know-how to analyse a corporation’s cloud and on-premise capability, to seek out the optimum cost-performance ratio round distribute compute workloads throughout them. Frayman says that it integrates with all main cloud suppliers and anything {that a} buyer could also be utilizing.
At a time the place firms are going through a scarcity of processors to coach and run AI fashions, the necessity for higher useful resource allocation is a powerful one. Solid, citing its own research, claims that on common solely 10% of CPUs and 23% of reminiscence are utilized, and the identical extends out to GPU utilization.
This Collection C — each in measurement and members — underscores what else it’s engaged on, and whom else it’s working with.
G2 Enterprise Companions and SoftBank Imaginative and prescient Fund 2 are co-leading the spherical, with Aglaé Ventures (LVMH chairman and CEO Bernard Arnault’s funding agency), and former backers Hedosophia, Cota Capital, Classic Funding Companions, Creandum, and Uncorrelated Ventures additionally collaborating.
Notably, Frayman identified that the oversubscribed spherical places the corporate into the identical portfolio secure as OpenAI and the AI infrastructure supplier Crusoe Power — two firms which might be, with SoftBank, Oracle and others, engaged on the large Stargate AI infrastructure venture out of the U.S. Frayman mentioned that his firm counts a lot of these firms as companions and prospects already.
“We’re partnering with Crusoe, the place we’re inside their stack, and we’re partnering with SoftBank to have the ability to facilitate the effectivity of their AI datacenters,” he mentioned, including that it is usually a part of the big venture between OpenAI and SoftBank to construct providers in Japan. “We’re partnering with the complete ecosystem,” he added.
Solid AI is speaking and doing quite a bit with AI today, however that was not precisely the place the corporate bought its begin. Ukraine-born Frayman, who based the corporate with Leon Kuperman and Laurent Gil in 2019, began his profession in finance earlier than pivoting to software program improvement.
Again in 2006, he and Gil constructed what Frayman described to me as one of many “earliest machine studying startups” — Viewdle. There they construct a few of the earliest functions of utilizing Nvidia’s GPUs to coach its classifiers for picture searches. “That’s how far again we go when it comes to understanding the facility of machine studying,” he mentioned.
That firm would ultimately get acquired by Google.
Together with Kuperman the three founders later labored on a cloud-based cybersecurity startup, Zenedge, which was the inspiration for Solid: there, they struggled to maintain cloud prices below management as they scaled up. (Zenedge was ultimately acquired by none apart from Oracle.)
The primary use case for Solid got here out of their expertise with that useful resource battle, and whereas it has at all times had “AI” in its identify and ethos, it was in regards to the utility of it, particularly make cloud use and allocation extra environment friendly for Kubernetes workloads.
Kubernetes functions are nonetheless on the coronary heart of Solid, Frayman mentioned, each when it comes to revenues and ethos. (And for those who go to its website, that’s the distinguished messaging there, too.) However it’s the surge of exercise round AI the place all the thrill and development are coming from in the mean time, from prospects and traders.
“Solid AI is setting a brand new normal for cloud effectivity at a time when infrastructure calls for are surging,” mentioned Tim Yap, Funding Director at SoftBank Funding Advisers, in an announcement.
“Proper now on the planet, everyone seems to be speaking about AI brokers,” mentioned Carl Fritjofsson, basic accomplice at Creandum. “Solid was was an AI agent earlier than we began speaking about that know-how, you already know. They’ve simply been constructing these kind of automation for a very long time.”