There’s a certain poetry to speed. The kind that turns raw computation into rhythm, where latency isn’t just a metric, it’s the beat. Fireworks AI just hit that tempo with a $250M Series C at a $4B valuation, and the tech world just turned its head. The round was co-led by Lightspeed Venture Partners, Index Ventures, and Evantic, with Sequoia Capital doubling down. That’s not just capital, it’s conviction.
Fireworks AI was born in 2022 from a crew that doesn’t just understand artificial intelligence infrastructure, they built it. Lin Qiao, former Senior Director of Engineering at Meta and once the driving force behind PyTorch, now leads as CEO. Alongside her, Dmytro Dzhulgakov holds the CTO post, a Meta alum and PyTorch core maintainer who speaks fluent GPU. Add in Dmytro Ivchenko, Chenyu Zhao from Google’s Vertex AI, Benny Yufei Chen, James Reed, and Pawel Garbacki, and you’ve got a lineup that could out-engineer gravity.
Their mission? To give enterprises control of their AI stack, full throttle, no training wheels. Fireworks isn’t another black-box API; it’s the open-source inferno that lets teams build, fine-tune, and deploy generative models on their own terms. With 10T+ tokens processed daily, serving 10K+ orgs, the company’s growth is less “hockey stick” and more “rocket stage separation.” Revenue just crossed $280M ARR, profitability is in the rearview, and they’ve scaled inference to match Google Search traffic.
This isn’t a prototype playground, it’s the real deal. Fireworks built FireAttention, a proprietary inference engine hitting 250 tokens/sec on NVIDIA B200s. Think 15x faster, 4x lower latency, 99.99% uptime. Numbers that make even the silicon blush. And while others brag about GPUs, Fireworks orchestrates 8 cloud providers across 18 regions, BYOC included.
Series C funds will expand that firepower 3–4x, bringing 150+ new AI researchers and engineers into the fold. Expect more breakthroughs in post-training alignment, inference optimization, and the kind of compound AI systems that make autonomy feel inevitable. Partnerships with NVIDIA, AMD, MongoDB, Databricks, AWS, and Oracle aren’t press-release filler, they’re accelerants.
What Lin Qiao and Dmytro Dzhulgakov have built isn’t just infrastructure; it’s an anthem for every developer tired of waiting for inference to catch up with imagination. Fireworks AI didn’t just raise capital, they raised the ceiling on what open AI can be.
