Unit 3 · Level 5 · AI × crypto
Decentralized compute & data
AI is hungry for compute, and GPUs are scarce and expensive. Decentralized compute networks (like Render or Akash) use tokens to pay strangers for their idle GPU power; data networks pay contributors for training data. The bull case: an open marketplace undercutting big cloud providers. The bear case: serious AI labs need reliable, fast, co-located clusters, which scattered hobbyist GPUs struggle to provide.
Free to play. No ads, no token, no account needed to start.
What you get asked
What role does the token actually play in a decentralized compute network?
The token coordinates a marketplace: buyers pay it, providers earn it. Which means its long-term value leans on real usage. That's a testable claim, not a vibe.
What is the strongest HONEST bull case for decentralized compute?
Idle hardware exists, cloud compute is expensive, and that gap is a real arbitrage for price-sensitive workloads like rendering. Note the bull case is about economics, not magic.
And the strongest honest BEAR case?
Training big models needs thousands of GPUs with ultra-fast interconnects in one place. Distributed networks fit rendering and some inference: a real but narrower market than 'replacing the cloud'.
The graduate's test for any compute or data network: is there real paying ___, or just token speculation?
Networks usually publish usage stats: jobs completed, fees paid by actual customers. If fees are near zero while the token is worth billions, the market is pricing a story.
Match each check to what it reveals about a compute/data network
Same toolkit you used on DeFi protocols, pointed at AI networks. Sectors rotate; the diligence questions never do. 🐜
The rest of this unit
Agents with wallets, decentralized compute, narrative coins, and the deepfakes hunting your keys.