Listen to this articleParasail has raised $32 million in Series A funding co-led by Touring Capital and Kindred Ventures to scale its serverless GPU orchestration platform for AI inference. The capital will expand global compute capacity, enhance scheduling algorithms, and accelerate adoption among startups building with open source models in a supply constrained market.
Parasail, a San Francisco-based AI infrastructure company founded in 2023 by Mike Henry (former CPO at Groq) and Tim Harris, has raised $32 million in a Series A round co-led by Touring Capital and Kindred Ventures. This funding comes roughly a year after the company emerged from stealth in April 2025, following a $10 million seed round in 2024 led by Basis Set Ventures with participation from Threshold Ventures, Buckley Ventures, and Black Opal Ventures.
What is Parasail’s technology?
The company operates a serverless AI deployment network that aggregates and orchestrates GPUs from dozens of global providers across 40 data centers in 15 countries. It delivers on-demand compute for machine learning inference workloads, emphasizing flexibility with no long term contracts, no vendor lock-in, and pay per token pricing. Key offerings include instant serverless inference that auto scales for experiments or production APIs, dedicated serverless pools for guaranteed throughput and low latency, fully reserved dedicated GPUs for custom models and enterprise privacy needs, and batch processing that delivers 80-90% cost savings for large offline jobs. The platform supports Nvidia H100, H200, A100, and RTX 4090 GPUs, dynamically matching workloads in real time while optimizing across rented capacity and liquidity markets to minimize costs and avoid peak demand bottlenecks.
Parasail generates 500 billion tokens per day and focuses exclusively on inference (training is not supported). Its proprietary orchestration technology connects fragmented compute resources horizontally, enabling developers to access hardware, data centers, and optimizations with minimal engineering overhead. This approach contrasts with hyperscalers like AWS, Azure, and Google Cloud by prioritizing fluidity in a market where compute is fungible and rapidly evolving, rather than centralized control. Pricing undercuts incumbent clouds significantly while maintaining production grade reliability and simplicity for scaling open source models.
The latest round reflects strong market tailwinds in AI inference demand, which investors describe as far outstripping supply amid the rise of agentic systems and open source models. Developers increasingly adopt hybrid architectures, using cheaper open models for initial tasks like high volume screening before routing to frontier models, driving what the company and its backers call “tokenmaxxing”: the imperative to deliver tokens faster, cheaper, and at massive scale as agents proliferate in software. Inference is projected to account for at least 20% of future software development costs, creating opportunities for specialized brokers like Parasail that serve seed and Series B stage AI startups without enterprise style commitments.
Recommended: How to Navigate the New York LLC Publication Process Without the Headaches
Traction includes customers such as Elicit (which processes over 100,000 scientific papers daily via high throughput screening), Weights & Biases (leveraging massive DeepSeek capacity), Rasa (deploying custom models with improved European latency), and Oumi (generating millions of responses for dataset construction via batch jobs). These deployments highlight Parasail’s ability to remove bottlenecks in real time processing, reduce costs for open model strategies, and simplify coordination for large scale experiments.
Strategically, the Series A capital will accelerate expansion of the global GPU footprint, enhance orchestration and scheduling algorithms, bolster engineering and go to market teams, and improve reliability across the distributed network. This builds directly on the seed round’s focus and positions Parasail to capitalize on the ongoing fragmentation of AI infrastructure, where rapid hardware innovation makes ownership impractical for most builders and demand shows “literally no end.” By betting on continued proliferation of open models and agent driven token consumption, the company aims to become a foundational compute layer that democratizes access and levels the playing field against hyperscaler dominance.
In the broader AI infrastructure landscape, Parasail’s model underscores a shift toward brokerage and optimization in a fluid, multi vendor environment. It competes with specialized inference platforms like Fireworks AI and Baseten while differentiating through extreme flexibility, cost efficiency, and a startup first customer base. The round signals investor conviction that inference will remain a high growth, supply constrained segment even as overall AI hype evolves, with Parasail’s horizontal approach enabling it to scale alongside the next wave of AI software development.
Please email us your feedback and news tips at hello(at)dailycompanynews.com
