The team that built MegaETH is taking its engineers into AI inference, the computing work of running trained models to answer requests, and says a public marketplace for it will open on its blockchain in December.
The MegaETH Foundation became MEGA on Tuesday, and the new entity said it is committing $30 million to products built on MegaETH, starting with inference.
“We realized that staying a chain-only company felt wrong and would have meant ignoring the biggest opportunity in front of us,” MegaETH co-founder Shuyao Kong wrote on X. Kong said MEGA is not turning into an AI company, writing, “MEGA is a crypto-native company, and proudly so.”
Get Unchained’s crypto news in your inbox with the free Unchained Daily newsletter.
One Token, a Wider Mandate
In a separate thread, Kong said the new MEGA entity “has no equity and one token: $MEGA.” Its board will be tasked with decisions such as working with the community to update tokenomics, carrying out programmatic token buybacks and investing further in product lines, according to Kong.
MegaETH “remains general-purpose and permissionless,” the project posted, and teams building on it will keep getting technical support. Kong said the team is putting resources into four upgrades and hopes to finish decentralizing the network by the end of the first quarter of 2027. They are fast finality, which shortens the time before a transaction counts as settled, a rotation of the sequencer (the node that orders transactions) so no single team produces blocks, a market where operators compete on connection speed to that sequencer, and real-time censorship resistance.
Why It Matters
With no equity in the new entity, the MEGA token is the one asset tied to MEGA’s products, so how its board handles buybacks and tokenomics will shape what holders get from the inference push. The plan also ties MegaETH’s decentralization roadmap to a product deadline, since the project said the December launch requires pushing the chain further.
Inside the Whitepaper
A whitepaper by Yilong Li and Lei Yang sets out the thinking. It argues that buyers can stick with a familiar, pricier inference provider because they cannot check whether a cheaper one serves the model it advertises or handles their data safely. The paper proposes tools to make providers interchangeable on model authenticity, privacy and security, and floats tradable forward contracts for inference capacity, pointing to Osaka’s 18th-century Dojima Rice Exchange as a precedent. It says a blockchain can keep shared records and handle payments but cannot on its own prove inference was delivered.
MegaETH said in a teaser post that its strongest engineers were moved onto a “special project” in early summer, after its accelerator program wrapped.
MEGA has not said which models or providers the market will carry, or what role the MEGA token will play in it.
Related Listen: Why Programmatic Token Buybacks Are ‘Kind of a Red Flag’
