Vitalik Buterin said the failure scenario he worries about most is not a technical collapse of Ethereum, but a crypto industry that turns into a place for nothing but speculation. If the sector ends up as a market with only token trading and no meaningful applications, he said, it could simply fade out as people lose interest.
Infrastructure improved, but applications still lag
Speaking in Chiang Mai, Buterin said Ethereum made major progress on scaling over the past year. The network’s gas limit rose from 30 million to 60 million, with a target of 300 million this year. He also pointed to zkEVM deployment and better wallet infrastructure. On the technical side, his view was clear: the base layer and surrounding stack have moved forward.
What troubles him is the gap between those gains and what users actually do onchain. He said the ecosystem once carried wider ambitions around DAOs and decentralized applications that could change how people coordinate. In his telling, crypto found financial traction, but lost direction on governance. Token-based voting in DAOs remains flawed, and the broader application layer has not matched earlier expectations.
Why SocialFi runs into the same wall
Buterin described SocialFi as stuck in an awkward place. The core problem, he said, is structural: when social activity and financial incentives are tied too tightly together, the financial side tends to overpower the social one. Users stop showing up for high-quality content and start optimizing for rewards, which leads to spam, low-value posts, and attention farming.
He contrasted that with Substack, where leading writers are known for substance and consistency. In some crypto-native social platforms, by contrast, top spots often go to accounts driven by volume tactics or hype. For him, the difference lies in curation and community building, not in whether a platform can attach a token to posting. He also said Farcaster’s move toward wallets reflects a search for scale, as wallets appear more likely than pure social products to reach mass adoption.
Prediction markets work, but many markets are too shallow
Asked about applications that did break out, Buterin said prediction markets were the one category in 2025 that could plausibly claim strong growth. Even there, he sees a limit. Markets focused on near-term sports outcomes or short-window Bitcoin price moves may function well as products, but that does not mean they create much social value.
He said he is more interested in longer-horizon designs such as futarchy, where communities vote on goals and prediction markets help decide the best path to reach them. He cited MetaDAO as one project exploring that direction. On Polymarket, Buterin said he personally made $70,000 last year using $440,000 in capital, mainly by identifying markets that had entered what he called a “crazy mode” and betting that extreme outcomes would not happen.
Oracle failures can turn bad information into onchain payouts
One of Buterin’s sharpest warnings was aimed at oracle design. He described a market tied to the war in Ukraine, where the outcome depended on whether Russian forces controlled a key train station in a city. The contract relied on updates from ISW’s public map and social media output. A map change briefly turned an event the market had priced at roughly 5% into a 100% outcome. Even though the update was later reversed, funds may already have been paid out.
His point was that many Web2 information sources were never built with the security assumptions required for onchain settlement. He grouped oracle approaches into two broad models: centralized providers, where users trust a company, and decentralized token-voting systems such as UMA. But he said token voting can break down if large holders coordinate to force an outcome, because smaller participants are pushed to follow voting power rather than truth. He added that Chainlink remains the standard choice across DeFi, though its setup is also complex and relatively centralized.
Ethereum’s role in AI is as an open, permissionless layer
On AI, Buterin pushed back on the idea that crypto and AI must be combined for its own sake. He described Ethereum first as a decentralized world computer with a key property: permissionless access. Humans, companies, and AI agents can all hold assets, transact, and potentially take part in governance under the same basic rules.
He highlighted several areas where the overlap does make sense. Crypto can serve as a bank account for AI agents that cannot open traditional accounts. AI systems can act as participants in prediction markets. Blockchains may also help verify whether content was made by a human or generated by AI. On coding tools, he said he usually works through OpenRouter and uses mainstream models including ChatGPT, DeepSeek, and Gemini.
The bigger danger is a powerful stack with no meaningful use
When asked about underestimated risks facing ETH, Buterin did not point to consensus failure or hacks. He said the more serious problem would be a future where the ecosystem ships thousands of applications, yet none of them carry real social meaning. If the strongest decentralized technology ends up producing only games, toys, or gambling venues, that would be a deeper loss.
He also said neither AI nor quantum computing should be viewed as a classic 51% attack threat to Ethereum. In a proof-of-stake system, that kind of attack is about capital and coordination, not raw computing power. Quantum systems threaten cryptographic signatures rather than the consensus model itself, while AI could help strengthen Ethereum by improving formal verification and finding software bugs.
What he wants builders to focus on next
Buterin pointed to three areas he most wants developers to build. The first is decentralized social networks where users are not locked into a platform and can move their identities and relationships. The second is smarter DAOs, built around clear goals and governance structures rather than token issuance plus basic voting. The third is decentralized stablecoins.
On stablecoins, he said the more interesting direction may not be fiat pegs at all. A stable asset could instead track real-world value such as CPI or energy prices, aiming to preserve purchasing power rather than mirror a national currency. He added that Libra had a strong underlying idea, but its implementation became too tied to corporate control. In his view, the decentralized version of that concept is still worth pursuing.
That brought him back to the same concern running through the conversation: if crypto fails to build open systems with real use, the future digital order may be dominated by centralized AI. For Buterin, one of crypto’s purposes is to keep another path open.