Fireworks AI raises $250M at $4B valuation to power application-specific inference at Google-scale token volumes

situation because Non is going public today and the IPO's timing might be yeah I just want to be cognizant of his timing. So okay so we will have the production team working on let's bring in let's bring in Lynn from Fireworks AI into the TV panel show. Welcome to the show, Lynn. How are you doing?

Hi, thanks for having me. Uh, thank you for taking the time. Big day. Uh, would love to get your introduction to the company, how you're framing your business relative to everything else that's going on in AI. It's such an exciting time. Um, and then um, we'd love to go into the news. Yeah, definitely.

So, Fireworks, we are a AI inference platform. Yeah. Uh, with application specific inference. Okay, so here's how we think about inference differently that we do not think about inference as we buying compliances at home. For example, we buy a refrigerator, use that for 10 years and then upgrade to the next one. Yeah.

Um we believe model should be continuously becoming smarter and smarter and learning and adapt into the application. So every single application developer, they should have their own model. They shouldn't use offtheshelf um you know any model other people use and become a model API wrapper.

Instead the model should be part of their product design. Yeah. And the model is like we raise child it will continue to learn and be smarter and be more specialized in solving particular problem really really well.

So that's kind of inference platform we provide with um smarter model faster model and a much more costefficient model for every single application. So I mean you're competing with am I crazy? Are you competing with like AWS, Azure? Like these are the biggest hyperscalers.

They are somewhat model agnostic but serve up inferences APIs. Like Lynn's competing with everyone. Is this the craziest? He's coming for everybody. But but I mean what? Take me a a click deeper on the actual strategy.

Is it just that there's so much demand that you can go in go and satisfy the the extra demand or do you think you're building something that will just be completely differentiated from the hyperscalers forever? Right. So um we uh currently we are running very high traffic.

Uh we are processing more than 180 requests per second. That is an order of magnitude of uh it seem in the ballpark of Google search traffic volume. We process more than 10 trillion tokens per day.

Um okay based on Google's earning cloud that's similar to their [laughter] um that is um that is actually the same as uh their um token um processed uh from Gemini. Wow. So open process has a large volume of data.

Um and that means um that just means um open model has the future and there are so many startups and uh enterprise they are reinvent the new user experience and they want to have full control of of the model as their IP um and have the model co-develop co-evolve co- adapt into their product and we see that as a future and that's how we differentiate from product focus point of view and we do not provide this inference one size fits all we provide inference one size fits one specific tailored towards application.

Are you GPU rich? Are you GPU poor? Is this uh part of the uh part of your strategy to uh have GPUs available when your companies need them? Is that a differentiator?

Yeah, I I would like to say we design our product for GPU poor because um interestingly uh we are at the beginning of escurve of huge uh AI native product explosion. Yeah.

But product market fit today doesn't mean a viable business in the sense that uh many of the uh we we talk we work with so many companies they just couldn't scale to millions of developers or billions of consumers they're going to scale into bankruptcy [laughter] fundamental their infrastructure is so expensive even if they build the best product has so much value they just couldn't make a viable business.

Yeah. Scaling to bankruptcy is an iconic line. It's a great line. I'm not going to name. There's one company that really comes to mind. I won't name them. [laughter and gasps] Um, give us the news. Give us the news from today on the fundraising side and get that gong ready again. What happened? We congratulations.

We raised our CC. Uh, yes. Give us some more details. Who who came in? How much? What is the valuation? What else are you sharing about the round? Yeah, we raised 250 million co-led by light speeded and index. Boom. Love it. with fantastic participation from Aventic and Sequoia. Fantastic.

You can just call it a quarter billion now. You're getting into the big number, the really big number range. So, right, it brought us our valuation to $4 billion. Four billion. So, really excited for all the customers, partners, and investors. And and quickly, how old is the company? We're three years old.

Wow, that is impressive. Congratulations. Uh well, thank you for everything that you're doing to power the AI revolution, helping people avoid scaling into bankruptcy. That would be the company model. That's honestly one of the greatest lines. Uh we're going to be we're going to be using that line quite a lot.

So, thank thank you for that. And uh great to meet you. Congrats to the whole team. Congratulations, Milestone. Thank you so much.