Sapiom, an AI agent routing startup, secured $35M in Series A funding to reduce AI agent costs. The platform slashed Polsia's monthly Anthropic expenditure from $1.2M to $100K, demonstrating its potential to disrupt the AI model market.
TL;DR
- Sapiom raised $35M in Series A funding to optimize AI agent model usage, reducing costs for startups.
- The company's routing technology cut Polsia's Anthropic bill by 95%, from $1.2M to $100K monthly.
- Anthropic's investment in Sapiom signals a strategic shift towards volume-driven model usage.
What happened
Sapiom, a San Francisco-based startup, raised $35M in Series A funding led by Dragonfly, bringing its total funding to $50M in just 11 months. Notable investors include Accel, Gradient, Coinbase Ventures, and surprisingly, Anthropic, a major AI model provider.
Sapiom's platform intercepts AI agent model calls and routes them to the most cost-effective model based on task complexity, cost, latency, reliability, and company policy. The company launched three products alongside the funding announcement: Router, Agent Studio, and Runtime.
Sapiom has processed over 270M transactions since its launch six months ago, powering over 100,000 agent runs per day. The company serves open-weight models from its own data center in San Jose, charging for compute directly.
Polsia, an AI startup, reduced its monthly Anthropic bill from $1.2M to $100K after implementing Sapiom's routing technology. This represents a nearly 10x reduction in costs.
Why it matters
Sapiom's technology addresses the growing concern of escalating AI agent costs, which is leading to project cancellations and a shift towards cheaper models. The company's solution could help startups and enterprises deploy AI agents at scale more affordably.
Anthropic's investment in Sapiom, despite the potential reduction in its own revenue, indicates a strategic bet on the growth of the overall AI agent market. The company believes that cheaper inference will lead to more agent usage, benefiting all model providers.
The AI agent cost crisis is driving the emergence of 'Inference FinOps,' a discipline focused on governing AI spending. Sapiom's technology is part of a broader trend towards optimizing AI model usage and controlling costs.
Key facts
- Sapiom raised $35M in Series A funding, bringing its total funding to $50M in 11 months.
- The company's routing technology reduced Polsia's monthly Anthropic bill from $1.2M to $100K.
- Sapiom has processed over 270M transactions and powers over 100,000 agent runs per day.
- Anthropic is a notable investor in Sapiom, despite the potential reduction in its own revenue.
- Sapiom launched three products: Router, Agent Studio, and Runtime.
- The company serves open-weight models from its own data center in San Jose.
- Gartner forecasts that over 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs.
- At least 80 active competitors offer some form of model routing, including Amazon, Microsoft, and OpenRouter.
Context
The AI agent market is growing rapidly, but so are the costs associated with running these agents. As more companies deploy AI agents at scale, they are facing six- and seven-figure monthly bills for model inference.
This cost crisis is driving a shift towards cheaper models and the emergence of 'Inference FinOps.' Companies are looking for ways to optimize their AI spending and control costs, similar to how they managed cloud spending in the past.
Sapiom's technology is part of a broader trend towards optimizing AI model usage. The company's success could pave the way for more startups to enter the model routing market and drive innovation in AI cost management.
