Blockchain-based, AI compute protocol Gensyn raises $43 million in a16z-led round

London-based Gensyn has raised $43 million in a Series A funding round. The company offers a blockchain-based protocol that connects buyers and sellers of compute power, offering a cryptographic verification system to provide proof that the machine learning work sent has been completed correctly.

The startup’s $43 million Series A round was led by A16z Crypto, with CoinFund, Canonical Crypto, Protocol Labs, Eden Block, Maven 11, and a number of angel investors including machine translation pioneer Oriol Vinyalas participating.

“We’ve entered a new industrial revolution. Steam, electricity, and the internet respectively powered the previous three. If it’s not already clear from the way you’ve been querying information with ChatGPT, generating code with Copilot, or the rapidly-shortening timelines for human-level Artificial General Intelligence: ours is a revolution of intelligence - and it’s fueled by computational power,” said Gensyn Co-founder Harry Grieve.

If you ever plugged into UC Berkeley’s SETI@home or the Covid-days Folding@home project(s), you’ve essentially got the gist of what Gensyn is working on - harnessing the compute power present in large-scale networks.

Now if the subject of vast amounts of compute power for extended periods of time is ringing any bells, then yes, you’re spot on when you point to anything AI related, and more specifically, the development thereof.

With the thesis that prohibitively costly access to cloud computing platforms including AWS and Azure is a major bottleneck holding back the true potential of AI development, three years ago, Gensyn set out on a journey to cut the middlemen out of the process and provide to power of compute to any and all that wanted it.

According to Grieve, “Creating a decentralised network of devices that anyone can access to train machine learning models cuts out the margins of intermediaries (e.g. AWS) and sharply increases supply. This has a significant, deflationary effect on the price of compute; in turn, that allows for bigger models for the same price and, for many, the ability to simply have a seat at the table for training large models. It means better models. It means the right to build your own models and not be beholden to the economic and technical influence of incumbents. It means computational liberty. It means the world will be unrecognizable in 5 years.”