Dragonfly reposted this
We’re thrilled to announce our investment in Prodia! Like many others, Dragonfly has been watching the galaxies of crypto and AI collide; we suspect that this intersection will produce some of the most generational companies of this era. And we think that Prodia perfectly fits this mold. Their image generation API has generated over 300 million images in the cheapest and most performant way possible. On the surface, according to Bloomberg, the generative AI market is set to expand to $1.3 trillion by 2030. And yet, this demand requires, in turn, a tremendous amount of chips, but the market is currently fragmented on the supply side. Interested parties must typically go to a large vendor like AWS or GCP today, sign multi-year contracts, and have to pay per compute hour. Many companies overpay because they have to include projections of expected demand – high demand means they have to pay above market rates for these GPUs at peak time, including when demand is normal most of the time. If there is a spike in usage, it will often temporarily make the customer’s service unusable or slow down inference. Conversely, if there is not enough usage, the customer is wasting money by letting GPUs run empty as they have already paid for the supply. Prodia swiftly sidesteps this issue by charging per generation, meaning all their performance benefits go straight to the consumer, alongside unrivaled optimization on their backend to juice speed. It’s cheaper because the customer doesn’t have to spend engineering efforts to deploy nodes, maintain servers, and design complicated servers to reduce empty GPU capacity. It's the equivalent of serverless compute for AI — paying only for your workload, rather than for the underlying hardware, and allowing the vendor to manage the load balancing and hardware efficiency. Ultimately, the Prodia team has done this before: they previously worked on Storj, one of crypto’s OG projects in storage. Ultimately, we’re incredibly grateful to be on this adventure with them and can’t wait as they unveil more over the ensuing months.
We raised $15M and couldn't be prouder of the team and honored to have started Prodia with Shawn Wilkinson & Monty Anderson. Excited to get closer to building an AI-powered future that is more accessible, democratic, and open. Reflecting on our journey, it feels like destiny. We've been experimenting with generative AI for years. After the release of Stable Diffusion, we launched an app to simplify access and were blown away by the response. Over 100K users signed up in under 2 months! Think of the scene in Silicon Valley with GPUs on fire—that was our reality, with some hardware still running in our basement. Getting GPUs from traditional cloud providers was nearly impossible. Fortunately, I was working with two experts in distributed systems. We just launched our own distributed GPU cloud because we had to. From that point on, our mission became clear: to ensure a world with open and equal access to AI, a distributed compute system must exist. A future where you can contribute to the network or access the best models is the future we envision. Thank you to our amazing investors, would not be possible with you. Anirudh P. Patrick Dai Jeremy Voss Matthew Roszak Balaji S. Timothy Chen Jermaine Brown Allen Nance Curtis Cheng Erik Huddleston Amit Vasudev Lauren Feld Zachary Ginsburg Sajid Rahman Ryan Condron Xan Ditkoff Nicolas Pinto Thomas France Shawn Lim Salvador Gala Etiënne V. HashKey Group Leon Marshall Marlon Williams Jason Kam Richard Dai, CFA Kenzi Wang Nicky Montana Matthias Buschle Jason Choi Benson Y. And also all the folks I couldnt find on Linkedin :) Today, we are proud to call some of the largest image generators in the world our customers. With a recently launched low-latency API, we’ve achieved best-in-class generation times. Try it here: https://sdxl.lol/