Parasail says its fleet of on-demand GPUs is larger than Oracle’s entire cloud

Cloud infrastructure is dominated by a number of massive business gamers: AWS, Microsoft’s Azure, and Google Cloud. Whereas to some it might appear to be AI is headed in an identical route, the founders of Parasail assume AI infrastructure will look very totally different — and are betting their firm’s destiny on it.

Parasail works with dozens of suppliers to ship on-demand GPUs for firms and enterprises seeking to construct AI fashions and functions. Parasail offers prospects entry to {hardware}, together with Nvidia’s H100, H200, A100, and 4090 GPUs, at a fraction of the associated fee that incumbents cost, in response to the corporate.

“There’s principally three cloud distributors who run the web, and that isn’t precisely how the web is being rebuilt whenever you have a look at AI,” Tim Harris, one among Parasail’s co-founders and the CEO of Swift Navigation, informed Trendster. “It’s rather more fragmented. The compute is rather more fungible and fluid, so you possibly can really inherently run it in a extra horizontal nature, and that’s actually what we had been making an attempt to do.”

Harris added, “We didn’t desire a world the place AI was managed from soup to nuts by the hyperscalers.”

Harris and Parasail CEO Mike Henry had the concept for the startup just a few years again. Henry, the previous CPO of Groq, informed Trendster that he had spent a very long time interested by what it will take to construct AI infrastructure that might compete with Nvidia. He mentioned that when he realized AI infrastructure was being quickly constructed up by quite a few gamers, he noticed a chance for a horizontal transfer.

In accordance with Henry, the speedy clip of innovation taking place in AI {hardware} was making it troublesome for firms to maintain up.

“We needed to actually concentrate on, how will we make [this] so simple as doable for the shopper?” Henry mentioned. “They’re barely maintaining with the open-source mannequin releases alone.”

Harris and Henry received began on the corporate again in 2023, employed an engineering group, and commenced constructing in early 2024.

At the moment, there’s no scarcity of distributors seeking to assist enterprises and different firms construct and scale their AI merchandise. From hyperscalers like Nvidia and Microsoft to startups corresponding to Collectively AI and Lepton AI, prospects have loads of choices.

The Parasail founders don’t assume they’re all one and the identical.

Certain, all these distributors present GPUs and AI infrastructure, Henry mentioned, however he thinks the proprietary know-how Parasail has working below the hood helps it stand out. This tech is what connects Parasail’s GPUs from varied sources collectively.

Wednesday marks the official launch of its platform, however Parasail is already working with dozens of shoppers together with Elicit, Weights & Biases, and Rasa. The corporate has additionally raised $10 million from a seed spherical in 2024 that had participation from Foundation Set Ventures, Threshold Ventures, Buckley Ventures, and Black Opal Ventures.

To achieve significant market share, Parasail must go head-to-head with hyperscalers and guess on the demand for GPUs persevering with to develop. There’s cause to imagine that it’ll, however there are additionally indicators — like Microsoft canceling a few of its information heart contracts — that the predictions round wanted AI infrastructure could also be a bit overblown.

“We see actually no finish [to] the demand,” Harris mentioned. “It’s actually that prospects have a tough time doing it — have a tough time scaling AI. The fashions now are attending to a spot the place [companies] can simply seize open-source fashions and just about run them, however then having the ability to get entry to GPUs, entry to information facilities, all of the optimizations… we will try this with a click on of a button.”