BerriAI's journey reads like a compressed version of Silicon Valley's current AI infrastructure gold rush: rapid adoption, marquee customers, venture backing. And then, inevitably perhaps, the security headaches that come when you're suddenly routing hundreds of millions of API calls a day.
The San Francisco startup, which builds LiteLLM—an open-source gateway that connects developers to more than 100 different large language model APIs—has quietly raised $1.6 million in seed funding from Y Combinator, Gravity Fund, and Pioneer Fund. The company emerged from Y Combinator's Winter 2023 cohort, with the seed round closing sometime in 2023–2024, though the precise date remains unconfirmed. Y Combinator's directory lists the investors, and Gravity Fund published an investment thesis in mid-2023, but BerriAI itself never issued a formal announcement.
That low-key approach makes some sense. The team has been heads-down building what amounts to critical plumbing for the AI economy: software that handles routing, fallbacks, rate limiting, and budget management across a sprawling landscape of LLM providers. Think of it as a universal adapter for anyone trying to juggle OpenAI, Anthropic, Cohere, and dozens of others without rewriting code every time they want to switch models.
The traction has been real. LiteLLM has accumulated more than 50,000 stars on GitHub—a count that fluctuates somewhat, tracking between roughly 49,000 and 54,000 depending on the month. The company claims to route hundreds of millions of LLM API calls daily and reports $7 million in annual recurring revenue in job postings, though this figure hasn't been independently verified. Customers reportedly include NASA, Adobe, Netflix, Stripe, and Nvidia, alongside a roster of well-funded startups like Rocket Money, Samsara, and Lemonade.
BerriAI has also locked in distribution deals that matter. Vercel, the frontend deployment darling, announced support for LiteLLM in April 2026. LiteLLM now appears in marketplaces on both AWS and Microsoft Azure, the kind of enterprise channel presence that signals a maturing product.
But the company's growth arc collided hard with security reality.
When the Supply Chain Breaks
In March, a supply chain attack hit LiteLLM's releases on PyPI, Python's package repository. Attackers managed to poison a GitHub Actions security scanner, compromising the integrity of the very tool meant to catch such intrusions. The breach affected thousands of users, including Mercor, a recruiting platform that TechCrunch identified among the impacted parties.
That wasn't the end of it. Between spring and early summer, researchers disclosed additional critical vulnerabilities: a pre-authentication SQL injection flaw, an authentication bypass exploiting host header injection. The Cloud Security Alliance documented the issues in research notes. BerriAI patched the holes, but the damage—to trust, to momentum—had been done.
Litigation followed. The case White et al v. Mercor.io Corporation et al, filed in a Texas federal court on April 5, 2026, names "Berrie AI Incorporated d/b/a LiteLLM" among defendants, with filings extending into June. The details remain sparse, but ongoing legal exposure is rarely a selling point when you're courting enterprise customers wary of compliance risk.
Scaling Through Turbulence
To BerriAI's credit, the team hasn't pulled back. The company recently open-sourced a "LiteLLM Agent Platform" and secured a spot in CrowdStrike's Open Gateway Ecosystem, where co-founder and CEO Krrish Dholakia appeared among cited participants. Job listings emphasize reliability and performance engineering—the unglamorous work of making sure the infrastructure doesn't buckle as usage climbs.
One posting describes LiteLLM as "post-Series A," though no public announcement of a Series A round has surfaced. Whether that's aspirational language, an internal milestone, or simply imprecise phrasing is unclear.
For developers evaluating AI gateway options, the open-core model offers flexibility. Self-hosting is free; the commercial tier adds enterprise staples like single sign-on, audit logs, secret management, and multi-region control. Online discussions routinely compare LiteLLM with rivals like Portkey, Helicone, and OpenRouter, each jockeying for position in a crowded but fast-growing market.
The seed funding—modest by today's inflated standards—gives BerriAI runway to keep building. Whether the company can translate its GitHub popularity into durable enterprise revenue depends on many things, not least whether customers believe the security storm has truly passed. Trust, in infrastructure, is everything. And once shaken, it's hard-won back.
In an industry moving as fast as AI tooling, a year of both breakneck adoption and unwelcome scrutiny might just be par for the course. The question is whether BerriAI emerges stronger, or whether the next wave of competitors—armed with cleaner security records—overtakes them before they can consolidate.
