Conference Presentation, Fireside Chat
Fireside Chat with Lin Qiao, CEO & Co-Founder of Fireworks AI | RAISE Summit 2026
- Matt Miller anticipates closing the first fund by June 30th (relative to the stated reference point) to initiate investment in Fireworks.
- Lynn projects a shift toward a future dominated by millions of specialized, single-application models rather than a few dominant general AGI models, with open model quality reaching a threshold for solving multiple problems simultaneously starting last year and intensifying this year.
- Organizations are expected to allocate between 5% and 20% of their inference budgets to continuous model training while experiencing a 5 to 10-fold reduction in standard inference costs, as customized models tuned for specific tasks become five to ten times cheaper than API wrappers, improving long-term cash flow.
- The operating landscape will require intelligent routing and a blend of models utilizing private data, necessitating that no single model handles all tasks, with new model launches occurring weekly to maintain leaderboard superiority and adapt to evolving applications and base models.
- Companies must establish product-metric-based evaluation sets to continuously optimize quality, while the lack of constant data feedback creates a risk of forced assumptions during growth, requiring rapid operational understanding to make correct judgment calls.
- As the technology frontier expands, the outlook suggests every company will own the underlying technology to build upon, necessitating automated agents to manage information overload during scaling.