Compiled by: Golden Finance

Note: On December 10, 2024, the Crypto x AI project Hyperbolic announced the completion of a $12 million Series A financing, led by Variant and Polychain Capital. Together with Hyperbolic’s previous $7 million seed round led by Polychain Capital and Lightspeed Faction, as well as earlier pre-seed financing, Hyperbolic’s total funding reached $20 million.

What kind of project is Hyperbolic? Please refer to Golden Finance’s previous report “Understanding Crypto x AI Rising Star Hyperbolic in One Article”.

At the same time, Variant published an article on its official website explaining why it led the investment in Hyperbolic. Let us follow Variant to see Variant’s investment logic in Hyperbolic.

Transformations in computing platforms tend to evolve in pairs or groups, often with new hardware, application and distribution innovations that push each other forward. Examples: personal computers, the Internet, and the World Wide Web. Mobile devices, social networks and cloud computing.

We believe that a new powerful combination is emerging at the intersection of crypto (technology) and AI. AI requires massive coordination of GPUs, while cryptography uses incentives to pool resources. AI is probabilistic, while cryptography is deterministic. We believe cryptography can solve two of the most pressing problems facing AI: cost and trust (especially the cost of trust).

Let’s explain this in detail:

Running models is currently extremely expensive. The root cause is often put down to supply issues, with the biggest tech companies hoarding resulting in a shortage of GPUs. But that’s not the case; GPUs are plentiful in data centers, mining farms, PCs, and local machines around the world. Instead, GPUs appear scarce because their supply is fragmented and uncoordinated. So what we're really facing is a coordination problem in a decentralized network of GPU vendors, which makes them expensive.

The decentralized network of GPUs has lower original costs, but brings a new problem: the cost of trust. How can you trust that a model run by a network of different actors is working correctly? Traditional solutions in the crypto space have been to have every node perform the same computation, thereby introducing significant overhead, or to reduce the computational burden entirely.

But for AI models, this doesn’t work because having each node perform the same calculation is too slow, and reducing model size also reduces quality. No doubt there are validation issues in centralized environments too (e.g. how do you know if ChatGPT is giving you GPT-4o or GPT 3.5? ), but OpenAI’s reputation could underpin trust more cheaply, albeit lacking cryptographic rigor. This verifiability may not matter when asking about cookie recipes, but it certainly matters when asking about the presence of malignancy in medical images. As AI takes on more and more important jobs in society, the cost of trust will only increase. Crypto networks are a step ahead in this regard, as they must solve the verification problem in order to reduce costs.

This is Hyperbolic’s turn.

Hyperbolic is the first participant we have seen to solve the problem of trust cost in decentralized GPU networks. One of the team's key innovations that makes this possible is sampling-based verifiable machine learning (spML). It uses a random sampling protocol called Proof-of-Sampling to guarantee verifiability in a decentralized network of GPU providers (assuming all parties behave economically rationally) while maintaining the ability to run the largest, highest quality AI models required efficiency. Hyperbolic enables verifiably running models at lower cost without sacrificing performance or quality.

Early market reaction supports this. Hyperbolic is one of the only platforms to host the Llama 3.1 405B base model in BF16 format, a large open source model that is comparable in quality to OpenAI's proprietary GPT-4o model, but running Llama 3.01 405B on Hyperbolic is much faster than GPT-4o model using OpenAI is 10x cheaper. Integrations with leading AI platforms such as Hugging Face’s Gradio, OpenRouter and Quora’s Poe underscore Hyperbolic’s commitment to bringing the highest quality models to the AI ​​community. Well-known AI developers like Andrej Karpathy have used Hyperbolic to run open source models because it is able to run higher quality models, is cheaper, and has a better user experience than competing products.

But Hyperbolic is more than just a formidable Web2 competitor, it will be unrivaled in meeting the needs of Web3 applications. Currently, Web3 applications are forced to make a Faustian bargain when integrating AI: in order to obtain the required performance, they must rely on a centralized source of AI inference, which directly contradicts the project's decentralized ethos and reintroduces Oracle problem. Because Hyperbolic will provide decentralization along with performance and quality, Web3 applications will be able to take advantage of it without sacrificing either party.

We believe that the team is focused on first building a platform that is accessible to all users (not just Web3 users)Competitive products are the right thing to do. GPU supply is mercenary and will follow demand without hindrance, so it's critical to attract demand first to build the necessary stickiness. We expect inference demand will continue to attract GPU supply and achieve the economies of scale needed to compete in the market over the long term. A less than perfect analogy to explain this is what Amazon does with AWS. Amazon first focused on building demand for computing through products that users loved (like its Marketplace), providing the computing supply needed to support that demand, and ultimately achieving economies of scale that allowed it to launch AWS and sell it more cheaply than its competitors. Better ways to provide computing services to third parties. After establishing core demand and supply on the Hyperbolic network, we believe the team will have the ability to scale at all levels of the AI ​​stack, including training, data sourcing and pre-processing.

Hyperbolic’s founders are the strongest team we have encountered in this space. They all have deep expertise in crypto and artificial intelligence, which makes them well-equipped to deal with the challenges of AI models. It has unique advantages in the decentralized computing market.

On the crypto side, Jasper Zhang, CEO and co-founder of Hyperbolic, is a mathematical expert with expertise in proof verification of distributed systems. Jasper is a multiple-time Math Olympiad winner, earned his PhD in mathematics from Berkeley in less than two years (becoming the fastest person to complete a five-year PhD in the school's history), and was previously a quantitative analyst at Citadel and researcher at Ava Labs.

In terms of AI, Hyperbolic CTO and co-founder Yuchen Jin is an expert in machine learning and distributed systems. Yuchen holds a PhD in computer science from the University of Washington on a prestigious scholarship and manages a team of engineers at OctoAI building optimization solutions for AI models.

We are excited to announce that today we led Hyperbolic’s Series A funding round. We’re excited to support Jasper, Yuchen, and the rest of the Hyperbolic team on their journey to make AI more accessible, verifiable, and open.