Sunday, 22 November 2026
Fireworks AI targets enterprises with open-source models amid Anthropic IPO

Image: © TechCrunch

Fintech & AIFriday, November 20, 2026 at 12:53 AM UTC

Fireworks AI targets enterprises with open-source models amid Anthropic IPO

Fireworks AI co-founder Chenyu Zhao argues that businesses can largely avoid expensive frontier models by using open-weight alternatives, even as the company navigates recent security incidents and a competitive market.

By Salala Times Desk
Key points
  • Fireworks AI specializes in tuning open-weight models for enterprise clients such as Uber, Shopify, and Notion.
  • The company disclosed two security incidents involving exposed API keys and customer data from a cloud provider breach.
  • Uber reported that approximately 150,000 drivers had personal information stolen through a vendor identified as Fireworks.
  • Notion has disconnected from Fireworks following the exposure of AI chat transcripts from fewer than 1% of its workspaces.
  • Fireworks holds a $17.5 billion valuation and is reportedly in early talks for a new funding round.

As Anthropic launched the largest initial public offering in history, a contrasting narrative was gaining traction at the Slush startup conference in Helsinki. Fireworks AI, a company that runs artificial intelligence models for other businesses, is positioning itself as a cost-effective alternative to the expensive frontier models offered by industry heavyweights like OpenAI and Anthropic.

The company specializes in open-weight models, which are available for download and can be fine-tuned by customers using their own proprietary data. This approach aims to deliver faster and cheaper performance while making it more difficult for competitors to replicate the resulting intelligence.

Chenyu Zhao, co-founder of Fireworks, stated that enterprises are increasingly gravitating toward platforms like his to avoid the high costs associated with frontier models. He argued that unless a business requires the most advanced generalized intelligence, these premium options are often unnecessary.

Security Incidents and Customer Impact

In one instance, hackers accessed API keys and tokens stored as secrets for a limited number of accounts. In a separate breach, attackers penetrated the company’s cloud provider, accessed a development system, and obtained customer data, including billing records and datasets.

Uber reported that hackers stole the names, addresses, birth dates, and driver’s license numbers of approximately 150,000 drivers through a vendor identified as Fireworks.

Notion, another customer, stated that AI chat transcripts from fewer than 1% of its workspaces were exposed. The company has since disconnected from Fireworks. Zhao noted that while some customers were impacted, he could not publicly name specific entities affected by the data exposure.

Strategic Positioning and Market Context

He estimated that open models are now only three to six months behind frontier models in capability, a significant improvement from a year or more ago. This gap has narrowed due to the emergence of models from DeepSeek and other providers.

Palo Alto Networks recently announced a $4.5 billion stock acquisition of Applied Compute, a startup founded by former OpenAI researchers. Rivals such as Together AI and Baseten have also secured major funding rounds recently.

Future Outlook

Fireworks, which was assigned a $17.5 billion valuation by investors in the summer, is reportedly in early talks for another funding round. Zhao declined to comment on the status of these discussions.

Regarding the recent security issues, Zhao attributed much of the activity to LLM jacking, where attackers hunt for API keys to sell tokens on illicit marketplaces. He maintained that Fireworks’ promise of zero data retention for inference traffic had not been compromised, though he acknowledged that batch prediction services, which store data at rest, are being moved to online flows to better align with zero-retention principles.

Zhao also addressed the recent departure of Jensen Huang from Nvidia, noting that Nvidia has been a strong supporter of open source since backing Fireworks in 2024. He expressed confidence that Nvidia would continue to invest in its Nemotron project and partner with Fireworks.

On the valuation of Anthropic’s public debut, Zhao offered a personal perspective, stating he would buy shares as a long-term bet on the scale of future AI usage. He suggested that while Anthropic may cement its place in the industry oligopoly, the long-term demand for capital-intensive AI development remains strong.

Sources
Based on reporting by TechCrunch.

More in Fintech & AI

View all →