Parasail raises $32 million to scale AI inference cloud

“Give me tokens. Just give me tokens. I want them fast. I want them cheap. I want them now.” This is the demand driving Parasail, according to CEO Mike Henry. The company provides a cloud computing service for companies running AI models for inference—defined as running AI models. To scale its operations, Parasail has raised a $32 million Series A funding round, coming a year after the startup emerged from stealth. The company currently orchestrates these workloads globally across 40 data centers in 15 countries.

Henry, a former executive at chipmaker Groq, designed the company’s strategy around “tokenmaxxing,” which refers to maximizing token throughput. Rather than committing to owning all its own silicon, Parasail primarily rents processing time from third-party facilities and buys additional capacity from liquidity markets. This brokerage model allows the company to route workloads dynamically to avoid demand peaks, driving down costs for developers. By focusing on cost-efficiency and flexibility for startups, Parasail aims to compete with larger cloud providers. According to Henry, Parasail’s infrastructure already generates 500 billion tokens a day as it competes with other cloud inference providers like Fireworks AI and Baseten.

This infrastructure is becoming critical as software developers shift their architectures. Andreas Stuhlmüller, CEO of Elicit—a startup using open models for research assistance that previously raised a $22 million Series A—noted that his company has moved toward open models. Stuhlmüller explained that sending 100,000s of requests to an API endpoint is difficult, prompting a shift toward using open models for initial tasks before querying models from providers like OpenAI and Anthropic. This high volume of queries is a primary driver for inference infrastructure. Samir Kumar, a partner at Touring Capital, which co-led Parasail’s Series A, projects that inference will represent at least 20% of the cost of building software in the future.

While some market observers express skepticism, Parasail’s backers argue that demand remains strong. Steve Jang, a partner at Kindred Ventures, which co-led the Series A round alongside Touring Capital, addressed concerns of an AI bubble. Jang stated that there is “no AI bubble,” asserting that inference demand is far outstripping supply.

Why it matters

As inference demand outstrips supply, startups like Parasail are building specialized cloud infrastructure to lower costs, betting that inference will become a significant portion of software development expenses.