Fireworks AI bets enterprises will skip frontier models

Original: Anthropic rival says companies can (mostly) skip the pricey models

Why This Matters

Open-weight model platforms are emerging as a credible, cheaper alternative to frontier AI vendors.

At Slush 2026 in Helsinki, Fireworks AI co-founder Chenyu Zhao argued most enterprises don't need expensive frontier models from Anthropic or OpenAI, pushing open-weight alternatives tuned on customer data for speed and cost savings.

On the same day Anthropic went public in what TechCrunch describes as the largest IPO on record, Fireworks AI was making the opposite pitch at Slush in Helsinki: most companies don't need frontier models at all. Fireworks runs open-weight AI models for businesses—including Uber, Shopify, and Notion—customizing them on proprietary data to run faster and cheaper than offerings from Anthropic or OpenAI. Co-founder Chenyu Zhao, a former Google Cloud Vertex AI engineering manager, argues frontier models are both overpriced and unnecessary for the majority of enterprise use cases. He also flagged a growing concern: that frontier labs quietly absorb customer know-how through prompts—a worry Microsoft CEO Satya Nadella raised publicly in July.

Fireworks' pitch isn't without baggage. The company disclosed two security incidents this fall. In one, API keys and tokens were accessed for a limited number of accounts. In a second, attackers breached a cloud provider, reached a development system, and stole billing records and customer datasets. Notion confirmed AI chat transcripts from under 1% of its workspaces were exposed and has since cut ties with Fireworks. Uber said hackers stole names, addresses, birth dates, and driver's license numbers for roughly 150,000 drivers via a vendor TechCrunch identified as Fireworks.

Despite the incidents, investor appetite remains strong. Fireworks was valued at $17.5 billion in summer and is reportedly in early talks for another funding round, though Zhao declined to comment. Rivals Together AI and Baseten have also announced large rounds recently. On Wednesday, Palo Alto Networks said it would acquire Applied Compute—a one-year-old OpenAI alumni startup in the same space—for $4.5 billion in stock.

Source

techcrunch.com — Read original →