Cloud infrastructure is dominated by several large industry players: AWS, Microsoft’s Azure, and Google Cloud. While to some it may look like AI is headed in a similar direction, the founders of Parasail think AI infrastructure will look very different — and are betting their company’s fate on it.
Parasail works with dozens of providers to deliver on-demand GPUs for companies and enterprises looking to build AI models and applications. Parasail gives customers access to hardware, including Nvidia’s H100, H200, A100, and 4090 GPUs, at a fraction of the cost that incumbents charge, according to the company.
“There’s basically three cloud vendors who run the internet, and that isn’t exactly how the internet is being rebuilt when you look at AI,” Tim Harris, one of Parasail’s co-founders and the CEO of Swift Navigation, told TechCrunch. “It’s much more fragmented. The compute is much more fungible and fluid, so you can actually inherently run it in a more horizontal nature, and that’s really what we were trying to do.”
Harris added, “We didn’t want a world where AI was controlled from soup to nuts by the hyperscalers.”
Harris and Parasail CEO Mike Henry had the idea for the startup a few years back. Henry, the former CPO of Groq, told TechCrunch that he had spent a long time thinking about what it would take to build AI infrastructure that could compete with Nvidia. He said that when he realized AI infrastructure was being rapidly built up by numerous players, he saw an opportunity for a horizontal move.
According to Henry, the rapid clip of innovation happening in AI hardware was making it difficult for companies to keep up.
“We had to really focus on, how do we make [this] as simple as possible for the customer?” Henry said. “They’re barely keeping up with the open source model releases alone.”
Techcrunch eventJoin 10k+ tech and VC leaders for growth and connections at Disrupt 2025
Netflix, Box, a16z, ElevenLabs, Wayve, Sequoia Capital, Elad Gil — just some of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don’t miss the 20th anniversary of TechCrunch, and a chance to learn from the top voices in tech. Grab your ticket before Sept 26 to save up to $668.
Join 10k+ tech and VC leaders for growth and connections at Disrupt 2025
Netflix, Box, a16z, ElevenLabs, Wayve, Sequoia Capital, Elad Gil — just some of the 250+ heavy hitters leading 200+ sessions designed to deliver the insights that fuel startup growth and sharpen your edge. Don’t miss the 20th anniversary of TechCrunch, and a chance to learn from the top voices in tech. Grab your ticket before Sept 26 to save up to $668.
San Francisco | October 27-29, 2025 REGISTER NOWHarris and Henry got started on the company back in 2023, hired an engineering team, and began building in early 2024.
Today, there’s no shortage of vendors looking to help enterprises and other companies build and scale their AI products. From hyperscalers like Nvidia and Microsoft to startups such as Together AI and Lepton AI, customers have plenty of options.
The Parasail founders don’t think they’re all one and the same.
Sure, all these vendors provide GPUs and AI infrastructure, Henry said, but he thinks the proprietary technology Parasail has running under the hood helps it stand out. This tech is what connects Parasail’s GPUs from various sources.
Wednesday marks the official launch of its platform, but Parasail is already working with dozens of customers, including Elicit, Weights & Biases, and Rasa. The company has also raised $10 million from a seed round in 2024 that had participation from Basis Set Ventures, Threshold Ventures, Buckley Ventures, and Black Opal Ventures.
To gain meaningful market share, Parasail will have to go head-to-head with hyperscalers and bet on the demand for GPUs continuing to grow. There is reason to believe that it will, but there are also signs — like Microsoft canceling some of its data center contracts — that the predictions around needed AI infrastructure may be a bit overblown.
“We see literally no end [to] the demand,” Harris said. “It’s really that customers have a hard time doing it — have a hard time scaling AI. The models now are getting to a place where [companies] can just grab open source models and pretty much run them, but then being able to get access to GPUs, access to data centers, all the optimizations… we can do that with a click of a button.”