Fireworks AI has raised a $1.5 billion Series D round, with Menlo Ventures among the investors, as the inference platform reports annualized revenue of $1 billion and rapidly growing token volumes1.
The round positions Fireworks as one of the most heavily capitalized companies in the AI inference layer. Menlo Ventures described Fireworks as "the runtime for specialized intelligence," framing the company's opportunity around a production tier where speed, cost, and control matter as much as raw model capability.
Fireworks' daily token volume has nearly tripled since late last year, climbing from 15 trillion to 43 trillion tokens per day. The company's annualized revenue recently reached $1 billion.
Menlo's investment thesis centers on a structural shift in inference demand. According to the firm, models gained the ability in 2026 to sustain long-running tasks without constant human oversight, unlocking new classes of economically valuable work and broadening the types of inference workloads enterprises require. Open-source model usage on OpenRouter has expanded more than 10x since the start of the year, per Menlo's account.
Fireworks' platform supports custom and open-source models running alongside frontier models in production environments. Several customer relationships illustrate the platform's positioning. Vercel reportedly achieved 40x latency improvements for its v0 users by partnering with Fireworks on reinforcement fine-tuning and speculative decoding. Cursor trained Composer 2, described as a frontier-level coding model, on Fireworks' platform. Factory, another customer, is reportedly using Fireworks to deliver up to 15x more work per dollar spent through open-source model options.
The company's founding team includes former leaders from Meta's AI infrastructure. CEO Lin Qiao previously led PyTorch at Meta. Dmytro Dzhulgakov was one of PyTorch's core maintainers and a senior leader within Meta's AI organization. The founding team also included former leaders of Meta's ads infrastructure, News Feed machine learning, and ranking systems, as well as the former AI lead of Google Vertex. George Hu, who previously helped scale Salesforce 50x to $5 billion before leading Twilio's 10x growth journey, is also part of the leadership.
ANALYSIS The $1.5 billion round and $1 billion annualized revenue figure place Fireworks squarely in the top tier of AI infrastructure companies by both valuation signal and commercial traction. The customer examples — Vercel on latency, Cursor on model training, Factory on cost efficiency — each represent a distinct inference shape, reinforcing the thesis that production AI workloads are fragmenting beyond a single model tier.