coindesk.com
Hyperbolic, one of the newer contenders in the race to apply blockchain tech to artificial intelligence, has raised $7 million in a seed funding round led by Polychain Capital and Lightspeed Faction.
The AI boom has strained the global supply chain for graphics processing units (GPUs) and other hardware used to train AI models. Surging hardware prices have squeezed all but the best-resourced AI outfits, making it difficult for startups and researchers to access AI alongside Big Tech companies like Microsoft, Google and Meta.
Hyperbolic's blockchain-based cloud platform aims to make AI hardware affordable for a wider swath of builders.
Read more: Math Olympian in Shadow of John Nash Tries to Solve Blockchain, AI Trust Dilemma
"Currently, AI companies are reportedly spending over 80% of their capital on compute resources, without a cost-effective solution on the market, and causing an industry-wide bottleneck due to cost and distribution challenges," Hyperbolic explained in a statement shared with CoinDesk. "Hyperbolic’s solution addresses this issue by creating a scalable system to aggregate global GPU compute and leveraging blockchain technology to ensure their network of nodes operates in a verifiable and secure manner."
Hyperbolic's first product will be an AI inference service that allows builders to use state-of-the-art AI models "at a fraction of the cost," according to the company. (After an AI model is trained, it makes "inferences" based on user requests, such as when ChatGPT responds to user queries.)
In the long term, the firm plans to build a "GPU marketplace" that offers developers and researchers cheaper access to cloud computing hardware for training their own AI models.
The seed round included participation from Chapter One, LongHash, Bankless Ventures, Republic Digital, Nomad Capital, CoinSummer Labs and Third Earth Capital. It also included angel investors like Balaji Srinivasan, NEAR's Illia Polosukhin and Polygon's Sandeep Nailwal.
Hyperbolic enters a growing category of startups raising big bucks to build at the intersection of AI and crypto.
Eventually, Hyperbolic co-founder Jasper Zhang thinks blockchain projects like his own can be used for more than just addressing AI accessibility hurdles. The pitch from Zhang – a mathematics olympiad winner in China and Russia with a doctorate degree from the University of California, Berkeley – isn't dissimilar from that of other crypto founders who hypothesize a radical new future where blockchains transform how AI is built and monetized.
"There will be a lot of other applications building on Hyperbolic," he said in an interview with CoinDesk. "An example could be a GPU futures exchange: Let's say you tokenize the future usage of a specific GPU and invite traders to come and trade it."
Zhang also imagines Hyperbolic could be used for "revenue sharing" – changing how AI models are monetized. "You could tokenize the revenue of an AI model, and then the token holder can help the AI model-builder to bootstrap his project."