Author: TechFlow

 

If the crypto business is a bit virtual, what will happen if it is combined with the real technology industry?

As the concept of AI chips becomes more and more popular, "encryption chips" have gradually become the focus of VCs.

According to CoinDesk, yesterday a startup called Fabric announced the completion of a $33 million Series A financing round, led by Blockchain Capital and 1kx, with participation from Offchain Labs, Polygon and Matter Labs.

Previously, the project had raised $6 million in a seed round led by Metaplanet.

This company has targeted the encryption hardware sector:

Fabric said the new funds raised will be used to build computing chips, software and encryption algorithms; at the same time, the company's business roadmap shows that it mainly wants to create a new chip processing unit called "Verifiable Processing Unit (VPU)" specifically for processing cryptography. The new chip is expected to go into production later this year and ship in the fourth quarter.

When talking about hardware, you may think of DePIN, but obviously Fabric's business does not point to DePIN. It is more like being independent of the encryption narrative, providing computing hardware/resources for the basic layer of the encryption algorithm level, just like the CPU provides hardware support for computers.

Fabric said in a press release that the VPU is "the first custom silicon chip to use an instruction set architecture specific to cryptography," meaning that "any cryptographic algorithm can be broken down into mathematical building blocks that are natively accelerated and supported by the chip."

According to this understanding, it seems that all current encryption infrastructure (L1/L2, ZK, smart contracts, FHE), etc. can benefit from the computing power of this chip, which is hardware that enables infrastructure.

When a group of VCs began to shift from "rolling infrastructure" to "rolling hardware", what exactly can this VPU bring to the crypto industry?

What exactly is VPU?

Although Fabric has not yet released a white paper on its business, we can get a general idea of ​​the specific functions of this VPU from public information.

Skipping various technical explanations and descriptions, the editor will use a more popular way to help you quickly understand VPU. The key is to understand what is missing in the development of Web3 today.

Blockchain or Web3 is basically based on cryptography technology:

Every operation of the blockchain, from simple transfers to complex smart contract executions, requires a large amount of cryptographic calculations.

Existing hardware devices, such as the CPU and GPU we are familiar with, can handle these tasks, but the efficiency is not ideal. The CPU is like an all-around athlete, good at various tasks but mediocre in cryptography. Although the GPU performs well in parallel computing, it was originally designed to handle graphics rendering, not complex cryptographic operations.

Therefore, it seems more reasonable to create a processor unit specifically for cryptographic calculations.

The VPU can therefore be understood as a true "crypto-processor", combining the best features of a GPU and an ASIC to create a component dedicated specifically to cryptographic purposes.

Traditional CPUs are like a Swiss Army knife that can do a lot of things but are not very efficient at handling specific tasks.

ASIC (Application Specific Integrated Circuit) is like a well-made scalpel, which performs well in specific tasks but lacks flexibility. VPU cleverly finds a balance between the two, a bit like an intelligent surgical tool that can be quickly adjusted according to different surgical needs.

According to the information given on the official website, this flexibility comes from the "cryptography-specific instruction set architecture".

Sounds complicated? Think of it as a recipe book specifically designed for cryptography, where each recipe is a common cryptographic operation, such as elliptic curve operations, hash functions, or zero-knowledge proofs.

The VPU is able to understand and quickly execute these "recipes" directly without having to convert them into more basic instructions like traditional processors.

Where might it be used?

This design makes the VPU perform well in cryptographic tasks. For example, some easy-to-think-of application scenarios can be:

  • Quickly verify the validity of transactions when dealing with complex smart contracts;

  • When performing zero-knowledge proofs, a VPU might be able to complete calculations in milliseconds that would take a traditional CPU seconds or even minutes to complete.

  • When processing large-scale data, the VPU can achieve near real-time encryption and decryption.

In another application known as the Holy Grail of cryptography and also at the forefront of encryption technology, VPU can significantly improve computing efficiency:

  • Accelerate FHE key generation and basic operations: Reduce key generation time from hours to seconds, and basic operations from seconds to milliseconds.

  • Support for large-scale FHE data processing: The time for statistical analysis on large encrypted data sets can be reduced from hours to minutes.

  • Optimizing FHE model training: In privacy-preserving machine learning, it is possible to reduce training time from days to hours.

In addition, in the specific applications of public chain networks that we are familiar with, VPU can also expect to significantly improve node performance:

  • Accelerate block verification and consensus process: It is possible to reduce verification time from hundreds of milliseconds to tens of milliseconds, and consensus time to sub-seconds.

  • Improve the efficiency of smart contract execution: The execution time of complex DeFi contracts may be reduced to one-tenth or less of that of traditional CPUs.

In decentralized identity systems and privacy-related fields, VPU provides more efficient identity authentication:

  • Rapidly generate and verify zero-knowledge proofs: Completed in milliseconds, enabling real-time, trustless authentication.

  • Supports complex multi-factor authentication: Processes multiple encrypted biometric data simultaneously to provide safer and faster identity authentication.

Top students from prestigious universities drop out of school to take on important roles

Since we are dealing with chips and cryptography, professional matching is definitely very important.

Public information shows that the two founders of Fabric are both top students from prestigious universities, and judging from their names, they also have Chinese backgrounds.

MICHAEL GAO is from MIT and is also the champion of the U.S. Mathematical Olympiad. He was previously an architect at an AI startup chip company backed by Bill Gates. Now he has dropped out of school to work on a new project.

On top of that, his bio says Bitcoin OG.

Another TINA JU has done some research in biology and mathematics, as shown in her profile, and public information shows that she also graduated from the prestigious Stanford University.

In addition, the company has an experienced professional team consisting of dozens of GPU and AI chip architects, software and compiler experts, and senior cryptographers, but they all seem to be older than the founders.

This also has the same characteristics as some crypto-related projects that came out of prestigious universities that we have seen before: young students take the lead as founders, with professional teams supporting them from behind.

From virtual to real?

In the context of the cryptocurrency and Web3 fields being dominated by software innovation and financial models for a long time, Fabric's breakthroughs in hardware, especially innovations in cryptographic chips, are still a good entry point.

However, the primary problem is the uncertainty of market demand. A large part of the growth of the crypto industry comes from speculation and hype. In such an environment, the actual demand for high-performance cryptographic computing may not be as great as expected.

Software development can adapt to market changes relatively quickly, while hardware R&D requires a longer cycle and greater capital investment. If the direction of industry development changes suddenly, specialized hardware may face the risk of a sharp drop in demand.

Whether Fabric can succeed in its journey from virtual to real remains to be seen.