Developers creating software utilizing generative AI models share a common demand: "Provide tokens. Simply provide tokens. I need them quickly. I need them affordably. I need them immediately." This urgent need is consistently heard by Mike Henry, CEO of Parasail, a company offering cloud computing services for businesses executing AI models for inference. Henry informed TechCrunch that Parasail processes an astounding 500 billion tokens daily, highlighting its significant capacity in this domain.
Mike Henry previously served as an executive at Groq, a chipmaker specializing in Large Language Models (LLMs), where he established their cloud service. This initiative stemmed from an early insight that AI software developers would require cloud processing tailored to their specific demands. Following a year since emerging from stealth mode, Parasail has successfully secured $32 million in Series A funding to expand these specialized services on a large scale.
Despite Henry's background in physical chip design, Parasail does not exclusively rely on proprietary hardware. While the company does possess some of its own GPUs, its primary strategy involves leasing processing time across 40 data centers in 15 countries worldwide and acquiring additional capacity from liquidity markets. This intricate orchestration occurs behind the scenes, effectively reducing the cost associated with inference requests for its clients.
Through intelligent workload allocation and strategic avoidance of peak demand periods, Parasail positions itself to compete effectively against companies that own their silicon, which often face limitations due to pre-existing customer commitments and workload capacities.
Parasail's growth potential is intrinsically linked to the ongoing expansion of open-source models and agents beyond leading research laboratories. Both Parasail's leadership and its investors attribute this trend to the escalating costs and operational complexities associated with utilizing services from prominent providers such as Anthropic and OpenAI.
Andreas Stuhlmüller, CEO of Elicit—a startup that recently secured $22 million in Series A funding to create a research assistant for scientific literature—observes the emergence of a hybrid architectural approach. His clients, including leading pharmaceutical companies, leverage Elicit's LLM-powered tool to efficiently review and analyze data from tens of thousands of scientific publications.
"We have increasingly adopted open models because transmitting hundreds of thousands of requests to a single API endpoint proves quite challenging," Stuhlmüller explained to TechCrunch. This shift is particularly critical as Elicit now employs agents to enhance its service, distributing tasks and operating more strategically across extended timelines. Open models manage the initial screening processes, thereby reducing overall operational costs, before a more advanced frontier model delivers the definitive answer.
The surge in model queries, driven by the increasing integration of agents into software development, is fueling investment in infrastructure providers like Parasail, which facilitate cost-effective inference. Samir Kumar, a partner at Touring Capital and co-leader of the recent funding round, informed TechCrunch that he anticipates inference costs will constitute at least 20% of future software development expenses.
Considering the competitive cloud compute landscape, Henry asserts that Parasail's specialized focus on inference (excluding training workloads) and its readiness to engage startup clients without demanding long-term commitments differentiate its service. This strategy positions Parasail distinctly from larger cloud computing providers targeting enterprise clients, and even from well-capitalized competitors in the cloud inference sector such as Fireworks AI and Baseten.
Naturally, focusing entirely on seed and Series B startups within the inherently unpredictable AI sector introduces a unique set of risks.
Steve Jang, a partner at Kindred Ventures and the other co-leader of the recent fundraising effort, believes that the economic realities of deploying AI models will necessitate the type of compute brokerage service offered by Parasail. He emphasizes that this demand precedes the widespread adoption of AI models for applications like content generation and robotics.
"Many perceived an AI bubble. There is no AI bubble," he stated to TechCrunch, underscoring that "inference demand is significantly surpassing current supply."
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.