San Francisco startup Sapiom secured $35 million in Series A financing, bringing its total funding to $50 million. The round was led by Dragonfly, with participation from Anthropic, Okta Ventures, Menlo Ventures and Array Ventures. Sapiom, founded 11 months ago, sits between AI agents and the large‑language models they invoke, directing each request to the most affordable model that can handle the task.
The company’s core product, called the Router, evaluates the cost and capability of available models before a call is made. By sending traffic to cheaper, open‑weight models hosted in Sapiom’s own San Jose data center, the platform avoids the markup that many rivals add. Since launch, the service has logged more than 270 million transactions.
A concrete example underscores the impact. Polsia, an AI‑driven operations startup, saw its monthly token bill from Anthropic soar to $1.2 million as its revenue grew from $100,000 to $10 million. After integrating Sapiom’s routing engine, Polsia’s spend fell roughly tenfold to about $100,000. Founder Ilan Zerbib told Semafor the reduction was “just unsustainable” at the higher price point and that cost is the new constraint on AI adoption.
Anthropic’s involvement as an investor raised eyebrows, but Zerbib frames the relationship as aligned rather than adversarial. While Sapiom helps customers spend less on frontier models, it still routes a portion of high‑value queries to those same models when performance demands it. The startup believes the volume of AI agents will explode, estimating “trillions of agents” will operate in the economy within three years, most of which will not need the most expensive models.
Industry analysts echo the cost‑concern narrative. Gartner predicts more than 40 % of agentic AI projects will be canceled by the end of 2027, citing escalating expenses as a primary driver. A recent KPMG survey of 2,100 executives found only 7 % could name a clear return on AI investment. Companies are beginning to cap AI spend, and chief technology officers are often forced to act as de‑facto CFOs for AI budgets.
Sapiom’s approach differentiates it from competitors like OpenRouter, which primarily act as middlemen without owning the underlying inference hardware. By charging for compute directly and leveraging its own racks, Sapiom aims to offer a more transparent cost structure. However, the routing market is becoming commoditized. Amazon Bedrock and Microsoft Azure now bundle routing capabilities, and open‑source routers move billions of tokens weekly without charge. Market trackers list roughly 80 active routing competitors.
Dragonfly’s Haseeb Qureshi joined Sapiom’s board, describing the challenge as an infrastructure problem rather than a dashboard issue. “Agents are becoming employees with no manager and no budget,” he said, “and increasingly, the CTO is the one acting as CFO, allocating real money with no visibility into where it goes.” The funding round gives Sapiom the runway to scale its data‑center operations and deepen its model‑selection algorithms, positioning it to capture a slice of the growing demand for affordable AI‑agent infrastructure.
Questo articolo è stato scritto con l'assistenza dell'IA.
News Factory APP - notizie agentiche per potenziare il tuo SEO e AEO.