If ETH had continued using PoW, it might have had one of the largest global GPU power pools. While it wouldn’t directly become an OpenAI-level large model training cluster—because the core of AI competition has shifted from sheer GPU quantity to high-end GPUs, VRAM capacity, high-speed networks, and software ecosystems—it still wouldn’t stop it from becoming a massive distributed AI computing network. Imagine if the millions of ETH mining GPUs back then didn’t exit but instead connected to the AI computing market via protocols. The Ethereum ecosystem could have formed a decentralized AI cloud powered by global miners. Kind of like a decentralized version of AWS GPU Cloud. This is an interesting historical "what if": ETH’s transition from PoW to PoS improved energy efficiency and capital efficiency, but in some sense, abandoning PoW also meant giving up the potential for a decentralized AI computing network. @VitalikButerin Ethereum originally had one of the largest distributed GPU networks globally. If combined with the computing demands of the AI era, it might have become a foundational infrastructure for decentralized AI in the future.