The founders of Fireworks pose for a photo at the startup’s headquarters in San Mateo, California. From left in the front row: Chenyu Zhao, CEO Lin Qiao and Benny Chen. Back row: James Reed, Pavel Garbacki, Dmytro Ivchenko and Dmytro (Dima) Dzhulgakov.
Fireworks
The cost of the latest artificial intelligence models is causing increasing concern among financial managers, and they have begun directing their employees to consider open source alternatives.
This strengthens the cloud startup Fireworks, with which it competes Amazon And Google to host models that developers can integrate into applications. The NvidiaThe company he backs said Thursday that it had annual revenue of more than $1 billion, five times what it did last year, and that it has now raised a $1.5 billion round of funding at a valuation of $17.5 billion.
“We’re seeing super linear demand,” Fireworks co-founder and CEO Lin Qiao told CNBC in an interview at the company’s headquarters in San Mateo, California. “This is a unique opportunity to have a market like this.”
Fireworks are much smaller than Anthropocene And OpenAIeach valued at over $800 billion by investors this year, and doesn’t even come close to the top names in tech, whose market caps are in the trillions. However, the startup’s sales milestone suggests that companies are not entirely satisfied with the models from the top labs.
The success also provides new evidence that Amazon, Microsoft and Google do not completely dominate cloud computing.
Shares of user-friendly cloud infrastructure providers DigitalOcean are up 149% so far this year as growth has accelerated. CoreWeavewhich rents out Nvidia graphics processors, or GPUs, raised $1.5 billion in an initial public offering last year and is now worth $42 billion.
By managing the computing infrastructure for models, Fireworks operates in the inference cloud market alongside startups such as Baseten and Together AI. It has also started providing GPUs for training AI models such as Neocloud’s CoreWeave, Lambda and I won’t.
Instead of going it alone, Fireworks has started forming alliances.
In March, it announced a partnership with Microsoft, which offers its own foundry service for running open models. The agreement allows customers of the Windows and Office company to access models via Fireworks, which relies on computing power from more than 20 providers, including Microsoft.
“Through Microsoft, we can achieve a much broader reach,” Qiao said.
Fireworks offers developers an easy way to adopt models from Chinese companies like DeepSeek. MiniMax And Z.ai. OpenAI open weight models released last year are also available. The idea is for customers to bring in their own data, which Frontier Labs doesn’t have, and refine models until they deliver cutting-edge performance for specific tasks, Qiao said.
While Anthropic and OpenAI provide “general intelligence,” Fireworks can unlock “specialized intelligence,” she said.
The argument may sound familiar to those who follow the discourse of technology figureheads.
Microsoft CEO Satya Nadella wrote in a blog post on Sunday: “A company should be able to use a model without sacrificing the knowledge that makes it unique.”
Nadella said Palantir CEO Alex Karp’s comments on CNBC earlier this month.
Technical customers “want to know that they own the means of production,” Karp said. “It doesn’t transfer to anyone else.”
Dollars and cents also matter. Cryptocurrency exchange operator Coinbase has introduced cheaper models where it makes sense, wrote CEO Brian Armstrong in a June X post.
“Our cost is five to 10 times cheaper compared to the closed model of the same quality,” Qiao said.
A former one Meta Director Qiao and six of her co-founders founded Fireworks in 2022. The company employs around 200 people. Qiao expects the number of employees to reach 600 by the end of 2026.
“This year we will really step on the gas,” said Qiao.
Firecrackers hired Salesforce Chief Executive George Hu was named president in April. The startup plans to assemble an impressive sales team after years of customers signing up. The new money will also help Fireworks obtain more GPUs and hire more engineers.
Developers are increasingly relying on Fireworks to process requests.
According to Qiao, Fireworks currently processes 40 trillion AI tokens per day. Google announced in May that its AI models process about 19 billion tokens per minute for developers, which translates to more than 27 trillion per day. OpenAI announced in March that its developer tools were processing 15 billion tokens per minute, which would mean about 22 trillion per day. Each token corresponds to approximately three quarters of a single word.
Last year, about half of Fireworks’ revenue came from AI coding startup Cursor, which has become less dependent on OpenAI and Anthropic and developed a custom model called Composer.
“We are much more diversified right now,” Qiao said. In June with Elon Musk SpaceX agreed to acquire Cursor in a $60 billion all-stock deal, with the transaction expected to close this quarter. Fireworks’ other customers include: Elastic, GitLab And MongoDB.
Atreides Management, Index Ventures and TCV led Fireworks’ new round. Nvidia participated, as did Evantic and Lightspeed Venture Partners.
REGARD: CEO of OpenAI: Chinese open source models are becoming very good
Choose CNBC as your preferred source on Google and never miss a moment from the most trusted name in business news.
https://www.cnbc.com/2026/07/16/fireworks-nvidia-cloud-ai-startup-value.html
