Jensen Huang and Lisa Su have just invested in a top female student from Fudan University.

According to ZDXX on October 31st, on October 28th, Fireworks AI, a US AI inference unicorn, announced the completion of a Series C financing round of $250 million (approximately RMB 1.775 billion). The total financing amount has risen to $327 million (approximately RMB 2.321 billion), and its valuation has reached $4 billion (approximately RMB 28.4 billion), more than seven times the $500 million valuation a year ago.

This round of financing was led by Lightspeed Venture Partners, Index Ventures, and Evantic. Existing investors such as Sequoia Capital, NVIDIA, AMD, MongoDB, and Databricks continued to participate. This financing is divided into two parts: a primary market financing of $230 million (approximately RMB 1.635 billion) and a secondary market financing of $20 million (approximately RMB 142 million), both participated by the same group of investors. Previously, Fireworks AI had received multiple investments from Sequoia Capital and NVIDIA.

Founded in 2022, Fireworks AI was co - founded by Lin Qiao, the CEO and co - founder who was a former senior engineering director at Meta, along with several former core members of Meta's PyTorch. Its core business is to provide a platform for developers to help them run and customize open - source models.

Notably, three of the co - founders are Chinese. Lin Qiao, the CEO and co - founder, graduated from Fudan University with a bachelor's and master's degree in computer science and then obtained a doctorate in computer science from the University of California, Santa Barbara. The other two Chinese co - founders are Benny Chen, a former software engineer at Meta, and Chenyu Zhao, a former senior software engineer at Google.

▲ Chinese co - founders of Fireworks AI (Source: Fireworks AI)

Fireworks AI does not directly purchase or develop server hardware. Instead, it builds infrastructure based on cloud - based GPU clusters and provides customers with access interfaces through APIs.

Its major customers include well - known enterprises such as Samsung, a global mobile phone giant; Uber, a globally renowned digital travel platform; Shopify, a leading global cross - border e - commerce platform; and Upwork, the world's largest freelance trading market. Its annualized revenue has exceeded $280 million (approximately RMB 1.99 billion).

Fireworks AI will use the new financing to recruit more than 150 new AI researchers, engineers, product, sales, and marketing employees, as well as to purchase GPUs to drive AI models.

01. A Chinese female entrepreneur's startup with a 7 - fold valuation increase in a year

The founding team of Fireworks AI is extremely impressive. In 2022, Lin Qiao, a former senior engineering director at Meta, and six co - founders established Fireworks AI in Redwood City, California. The co - founding team consists of six engineers from Meta's former PyTorch project and one former Google AI engineer. Before starting the business, Lin Qiao had led a team of over 300 engineers at Meta and had 24 years of industry experience.

▲ Founding team of Fireworks AI (Source: Fireworks AI)

During her time at Meta, Lin Qiao's greatest achievement was leading the development and deployment of PyTorch in Meta's data centers, mobile devices, and AR/VR devices, providing support for Meta's series of applications. Currently, PyTorch has become one of the world's mainstream open - source machine - learning frameworks.

Based on the successful experience of PyTorch, Lin Qiao led the team to establish Fireworks AI, targeting the core pain points of AI application implementation - the computational cost and efficiency of inference, and returning control to developers.

Fireworks AI has been established for only three years. It completed its seed - round financing in the year of its establishment, Series A financing in 2023 with investors including Sequoia Capital, NVIDIA, and AMD. In 2024, its valuation reached $552 million (approximately RMB 3.924 billion) in the Series B financing. The recent new round of financing has skyrocketed Fireworks AI's valuation to $4 billion, more than seven times the valuation a year ago.

Currently, the Fireworks AI platform processes over 10 trillion tokens daily and serves more than 10,000 enterprise customers.

▲ Partners of Fireworks AI (Source: Fireworks AI)

Compared with traditional cloud service providers, Fireworks AI offers large models with lower costs, higher performance, and support for customized open - source models.

Fireworks AI provides access to hundreds of cutting - edge open - source models covering text, image, audio, and multi - modal fields, supporting model fine - tuning, reinforcement learning, and evaluation.

▲ Large model library of Fireworks AI (Source: Fireworks AI)

All functions are built on its ultra - high - speed inference engine. The official website of Fireworks AI shows that compared with other service providers, the performance of the Fireworks AI platform can be increased by 40 times, and the cost can be reduced by 8 times.

In terms of function pricing, Fireworks AI adopts an extremely flexible pricing method, including serverless inference, model fine - tuning customization, and on - demand GPU deployment.

▲ Pricing of some functions of Fireworks AI (Source: Fireworks AI)

"The mission of Fireworks AI is to provide the best toolchain to help application developers achieve the highest quality, speed, and cost - effectiveness with the help of generative AI," said Lin Qiao. "We do not develop dedicated semiconductors ourselves. Instead, we allow developers to virtually access advanced chips to run open - source AI models and help them optimize the inference process and customize AI models."

02. The rapid development of the inference market, but enterprises are still cautious about adopting inference platforms

This financing comes at a time when the focus of competition in the AI industry is gradually shifting from training to inference. As tools like ChatGPT become more common in daily use, the demand for AI model inference is also increasing. Tech giants such as NVIDIA, Amazon, and Qualcomm, as well as AI chip startups like Groq and Cerebras Systems, are all actively promoting chip products dedicated to AI inference tasks.

Anoushka Vaswani, a partner at Lightspeed Venture Partners, said: "Looking back at the development of the inference market in the first three years, its scale has grown from zero to over $5 billion (approximately RMB 35.495 billion)."

Sualeh Asif, the Chief Product Officer of Cursor, said: "Fireworks has helped us achieve high - performance operation of the Fast Apply and Copilot++ models. In terms of performance, they have outperformed other competitors we have evaluated. Fireworks has also helped us achieve task - specific acceleration and new architectures."

Analysts point out that although AI companies like Cursor are already using Fireworks AI to deliver AI models to users, most enterprise customers are still more cautious when adopting platforms like Fireworks AI because the platform requires customized AI engineering support.

AI engineering requires a high level of professional skills. Enterprises must master how to combine trained models, datasets, and inference platforms. Many enterprises are still training in - house technicians or looking for external talents to fill this gap.

Chirag Dekate, an analyst at Gartner, a market research and consulting firm, said: "Currently, about 80% of enterprises have not reached this advanced AI engineering stage."

03. Conclusion: The soaring valuation of Fireworks AI and the fierce competition in the AI inference market

The soaring valuation of Fireworks AI reflects the rapid growth of the AI inference market. While other companies are still screening customers based on cost, Fireworks chooses to support developers, empowering them to independently build, deploy, and scale AI and control everything in their own way.

However, inference platforms like Fireworks AI are all facing new risks. Enterprises are still cautious when adopting these platforms, and the entry of large cloud service giants such as Amazon, Microsoft, and Google is intensifying the competition.

This article is from the WeChat official account “ZDXX” (ID: zhidxcom), author: Wang Xinyi, published by 36Kr with authorization.