Startups & Funding

Fireworks AI's $17.5B Pitch: Most Companies Can Skip Frontier Models

Fireworks AI co-founder Chenyu Zhao says open-weight models trail the frontier by three to six months, defends two breach disclosures, and sizes up a market shifting from cost per token to cost per task.

Anthropic rival says companies can (mostly) skip the pricey models
Anthropic rival says companies can (mostly) skip the pricey modelsAlexas_Fotos / pixabay
By Elena Vasquez6 min read

Why it matters

  • Fireworks AI was valued at $17.5 billion by investors this summer and is reportedly in early talks for another funding round.
  • Co-founder Chenyu Zhao says open models are at most six months behind frontier models, probably closer to three.
  • Two fall security breaches exposed API keys and customer data; ~150,000 Uber drivers' personal data was stolen and Notion disconnected.
  • Palo Alto Networks agreed to buy rival Applied Compute, founded a year ago by ex-OpenAI researchers, for $4.5 billion in stock.
  • Zhao cited $10,000–$100,000 monthly token spend at 10,000–100,000 users as the point where enterprises should consider open source.

Fireworks AI, valued at $17.5 billion by its investors this summer, is telling enterprises they can skip frontier models from Anthropic and OpenAI for most of their workloads — and its co-founder says open-weight models are now at most six months behind, probably closer to three.

Chenyu Zhao, who spent a decade at Google and managed a 50-person engineering team on Google Cloud Vertex AI, made the case on stage at the Slush startup conference in Helsinki. The timing was pointed. Anthropic went public the same day in the largest IPO on record, with investor enthusiasm for the AI lab running high. In Helsinki, the topic of the day was open-source everything.

The contrast frames a real fork in the AI market. Enterprises are hunting for ways to avoid paying frontier prices, and a cluster of well-funded companies is chasing that money. Fireworks runs AI models for other businesses. It specializes in open-weight models, which anyone can download, and helps customers tune them on their own data so they run faster and cheaper than the models of Anthropic, OpenAI and other heavyweights — and are harder for rivals to copy. Its customers have included Uber, Shopify and Notion.

Zhao argues the economics force a reckoning once a product reaches scale.

"Frontier models are a household brand, and before you find product-market fit, using the best model is low risk. But once you pass 10,000 or 100,000 users and you're spending $10,000 to $100,000 a month on tokens, it's time to think about open source."

How Close Are Open Models to the Frontier?

Not long ago, the honest answer would have been grim. Zhao says a few years ago he'd have said open models trailed by a year to a year and a half. Then came DeepSeek and what he calls "a waterfall of fantastic open models." His current estimate: at most six months behind, probably closer to three.

His pitch to CEOs who have never heard of Fireworks is ownership, not raw capability.

"I describe Fireworks as a specialized intelligence platform. OpenAI and Anthropic are building general intelligence. We use open models, which are often smaller, cheaper and faster, and let companies add their own data to specialize them. That way you own your AI rather than being beholden to giant AI conglomerates. If you're just using Opus or Sonnet, that's not a differentiator."

What Did Satya Nadella's Warning Change?

Microsoft CEO Satya Nadella warned in July that companies using proprietary AI models pay twice — once in money and again in know-how, absorbed by the frontier labs through the very prompts customers use to get answers. Zhao sees validation rather than disruption.

"We think we'll live in a two-model world, with frontier and open models in parallel," he said. Businesses want the best intelligence for high-level planning, but they don't need a frontier model for basic arithmetic. That logic is driving interest in routing — sending each request to the cheapest capable model — and a shift in how buyers measure spend: cost per task, not just cost per token.

Can Fireworks Survive Two Security Breaches?

The company has taken its first black eye. It disclosed two security incidents this fall. In one, hackers reached API keys and tokens stored as secrets for a limited number of accounts. In the other, attackers breached its cloud provider, made their way into a development system, and took customer data including billing records and some datasets.

The fallout was concrete. Notion said AI chat transcripts from fewer than 1% of its workspaces were exposed — and disconnected from Fireworks. Uber said hackers stole the names, addresses, birth dates and driver's license numbers of about 150,000 drivers through a vendor that TechCrunch identified as Fireworks. Cursor also went public about the impact.

Zhao wouldn't name other affected customers. "It was a small subset of our customers that had this data impact," he said.

The breaches expose a tension with Fireworks' core promise: zero data retention for inference traffic. Zhao says there's no indication that promise was compromised. The exposure came through a different door.

"Notion was using batch prediction, where you upload a dataset that's stored at rest, so it's not a zero-retention product by nature. We're moving those services to an online flow so we keep data only as long as we need it."

On the broader threat picture, Zhao said every company is probably seeing an unprecedented frequency of attacks. He attributed much of the motivation behind the Fireworks incident to "LLM jacking": hunting through providers for API keys and secrets to sell tokens on illicit marketplaces. Fireworks now has to win back the confidence of customers like Notion and Uber.

Where Does the Money Go From Here?

There's capital chasing everyone in the category. Rivals Together AI and Baseten have announced massive funding rounds in recent months. And on Wednesday, Palo Alto Networks said it will buy Applied Compute — a startup founded just a year ago by former OpenAI researchers that, like Fireworks, helps companies train and run custom AI models on their own data. The price: $4.5 billion, all of it in stock.

Fireworks itself is reportedly in "early talks" for another round, though Zhao declined to comment.

He left a decade at Google to build it. The decision, he said, came down to people: "When you meet great co-founders, it's obvious within five minutes that they're really smart. I was sold after talking to Lin [Qiao] and Dima [Dzhulgakov]."

What About Nvidia — and Anthropic's IPO?

Jensen Huang left Nvidia last week, and Nvidia has backed Fireworks since 2024. Zhao called it "a blow to the industry" and said the company feels for Huang and his family. But he doesn't expect the relationship to change: "Nvidia has always been a strong supporter of open source, and I don't see that changing. We expect it to keep investing in Nemotron, and we're excited to keep partnering with them."

On Anthropic's record-setting IPO, Zhao sees two paths: one where the lab cements its place in an oligopoly, another where it starts losing traction. His own money is on the long game.

"On the record: I'll probably call my broker tonight and buy a few shares. This is a long-term bet. In 10 years AI usage will be at a completely different scale, and that takes an incredible amount of capital, which is why these companies are going public."

That's the wager underneath the whole market: whether AI demand grows so large that frontier capital wins by default, or whether enough of it trickles down to cheaper, specialized, open-weight infrastructure. With Palo Alto Networks paying $4.5 billion for a one-year-old competitor and open models closing to within months of the frontier, the second scenario is no longer a fringe bet.

Source: techcrunch.com

Share this article:

More from Elena Vasquez

Elena Vasquez

Show full bio

Market editor covering technology and product news in artificial intelligence and machine learning at AI In Context.

795 articles

Related articles

  1. Anthropic and OpenAI Ship New Models With the Same Pitch: More for Less
  2. Anthropic's IPO Bet: $80 Billion in Losses Against a Trillion-Dollar Story
  3. AI Models Keep Cheating on Tests, and Researchers Are Quitting
  4. OpenAI Opens Applications for Grove Cohort 2
  5. OpenAI Launches Frontier, an Enterprise Platform for AI Agents

« Previous articleNext article »