Fireworks AI CEO explains why AI's infrastructure can't keep up with rampant demand

The demand for artificial intelligence is experiencing an unprecedented surge, and it shows no signs of slowing down. Lin Qiao, the CEO of Fireworks AI and a former engineer at Meta, is at the helm of a company valued at $4 billion that is currently processing a staggering 15 trillion AI tokens daily. Qiao predicts that this is just the beginning of a massive growth trajectory in token consumption. In a recent interview, Qiao highlighted the rapid expansion of Fireworks AI's inference cloud platform, which has seen an increase from 13 trillion tokens just a few months prior and from 10 trillion in late 2025. These tokens, which simplify words and inputs into manageable numerical formats, are crucial for processing and pricing AI model usage. With her extensive background in building PyTorch, an open-source framework vital to modern AI development, Qiao understands the importance of a robust infrastructure. The widespread adoption of AI is not limited to tech companies; industries across the board are integrating AI tools into their operations. Qiao shared examples ranging from finance departments automating forecasts to legal teams developing internal AI solutions, and even gig workers utilizing generative AI for on-demand music creation. Her own daughter exemplifies this trend, using multiple AI platforms for various tasks. "That's the world we're living in," Qiao stated, emphasizing that AI tools are becoming ubiquitous. However, this rapid growth is putting pressure on the entire technological ecosystem. Qiao pointed out that the demand for GPUs is high, with prices climbing and power infrastructures facing strain as companies rush to expand their AI capabilities. "The whole system is saturated," she remarked, highlighting bottlenecks that affect everything from semiconductor production to energy availability. Despite the presence of major cloud providers like Amazon, Google, Microsoft, and Oracle offering GPU rentals, Qiao believes Fireworks AI's unique value lies in its ability to navigate the complexities of AI deployment. Enterprises often find it challenging to keep pace with the fast-evolving landscape of models and hardware. Fireworks AI aims to alleviate this burden by optimizing performance and managing infrastructure, enabling clients to focus on their core business needs. Qiao's experience with both PyTorch and Fireworks has reinforced her belief that as AI technologies become more accessible, their adoption will accelerate rapidly. With current token usage trends indicating a continuing surge, the future of AI appears poised for explosive growth.