Cloud infrastructure is dominated by several large industry players, including AWS, Microsoft’s Azure, and Google Cloud. For some, it may seem like AI is heading in a similar direction, but the founders of Parasail believe that AI infrastructure looks very different and risk the company’s fate.
Parasail works with a number of providers to provide on-demand GPUs for businesses and businesses looking to build AI models and applications. Parasail provides access to hardware, including NVIDIA’s H100, H200, A100, and 4090 GPUs.
“There are basically three cloud vendors that run the internet. Looking at AI, it’s not how the internet is being rebuilt,” Tim Harris, co-founder of Parasail and CEO of Swift Navigation, told TechCrunch. “It’s much more fragmented. Computing is much more reliable and fluid, so it can actually be essentially executed with a more horizontal nature.
Harris said, “We didn’t want a world where AI was controlled from soup to nuts by hyperschool.”
Harris and Parasail CEO Mike Henry had the idea for a startup a few years ago. Henry, a former CPO at GROQ, told TechCrunch that he spent a lot of time on what it would take to build an AI infrastructure that could compete with Nvidia. He said he saw an opportunity for horizontal movement when he realized that AI infrastructure was being built rapidly by many players.
According to Henry, the rapid clips of innovations happening in AI hardware make it difficult for businesses to keep up.
“We really had to focus on it. How do we make it as much (this) as possible for our customers?” Henry said. “They’re barely catching up to just the release of open source models.”
Harris and Henry started the company in 2023, hired an engineering team, and began construction in early 2024.
Today, there is no shortage of vendors trying to help businesses and other companies build and expand their AI products. From hyperschools like Nvidia and Microsoft to startups like AI and Lepton AI, customers have plenty of options.
The founders of Parasail don’t think they are all the same.
Certainly, all these vendors provide GPU and AI infrastructure, Henry said, but he thinks Parasail stands out from its own technology running under the hood. This technology connects parasail GPUs from a variety of sources.
Wednesday marks the official launch of the platform, but Parasail has already worked with dozens of customers, including Elicit, Weights & Biase and Rasa. The company also raised $10 million from the 2024 seed round and joined from Coreset Ventures, Threshold Ventures, Buckley Ventures and Black Opal Ventures.
To gain meaningful market share, parasail must dive straight with hyperscalar and GPU demand must continue to increase. There are reasons to believe that it will happen, but there are also signs that Microsoft is canceling some of its data center contracts – the predictions about the required AI infrastructure may be a bit exaggerated.
“I literally think that demand will never end,” Harris said. “It’s really true that customers are struggling to do that — they’re struggling to scale AI. The model is where (companies) grabbed open source models and get to a place where they can pretty much run, but they can access the GPU, access the data center and do all the optimizations.