Demand for AI has been skyrocketing ever since ChatGPT debuted in late 2022. In fact, the demand has been so great that we are facing several challenges...
In addition to boatloads of additional energy, AI also requires vast amounts of GPUs (Graphics Processing Units) to train AI models and make inference calls on them.
Blockchain technology, and the underlying token incentives, can be harnessed to aggregate and coordinate these scattered resources.
Training vs. Inference
Training an AI model requires processing vast amounts of data over a span of weeks or months. Inference, on the other hand, is simply querying an existing model, like you would do through ChatGPT or Grok.
The massive AI data centers (of xAI, Google, or Meta) that house hundreds of thousands of GPUs are typically used to generate the AI models.
A decentralized network of GPUs is often more efficient at handling the inference requests, however, compared to the GPUs housed in these big tech data centers.
And there are plenty of underutilized GPUs spread around the world that could be used to serve these inference requests, often at a much lower cost than their centralized counterparts.
In fact, if you have a GPU in your desktop computer, you could lend it to a decentralized GPU network like Akash, and start earning the project's token as compensation.
Akash's Blockchain
Akash was one of the earliest DePIN (Decentralized Physical Infrastructure Network) projects that focused on aggregating scattered GPUs for the primary use of AI inference.
Akash built their own sovereign blockchain based on the Cosmos SDK. However, they are actively looking for an alternative, with Solana being the number one contender.
If the move is confirmed, Akash would be joining other decentralized GPU networks on Solana including Render, Aethir, and Nosana.
Burn-Mint-Equilibrium Token Model
Akash's latest mainnet update (on their Cosmos-based chain) changed their tokenomics model to Burn-Mint-Equilibrium (BME), which could be considered more efficient.
Consider how Bitcoin emits a set amount of BTC to miners every 10 minutes on average, regardless of demand for the token. Since Bitcoin's launch, however, a lot of innovation has occurred.
Helium was the first DePIN project to implement the Burn-Mint-Equilibrium (BME) tokenomics model in which tokens are minted and awarded to network contributors, but also burned based on usage of the underlying service.
In the case of Akash, AKT tokens will now be burned proportional to the demand for GPU power, reducing token supply, and theoretically increasing its value.
You can track the number of AKT tokens being burned on Akash' stats webpage.
Until next time...
Plenty of DePIN projects have dropped the standard token emissions model in favor of BME, in order to link the token supply to real-world demand.
Only time will tell if this novel tokenomics model becomes the new standard, but it has certainly been gaining traction in the DePIN space.
If you learned something new from this article, be sure to check out my other posts on crypto and finance here on the Hive blockchain. You can also follow me on X or InLeo for more frequent updates.
Sources
- What Burn-Mint Equilibrium Means for Akash
- Akash BME Announcement on X