MegaETH Cuts 20% of Staff. The AI Explanation Doesn't Survive Contact With the Org Chart.

One in five. That's the share of staff MegaETH's parent company let go in its latest restructuring, and it's the only figure worth arguing about this week. A network that sold itself on raw throughput just removed a fifth of the people who build throughput.

Co-founder Shuyao Kong confirmed the cut. Her account was direct. The affected roles were tied to the core blockchain, the wallet, and the product layer around them. Builders. The people you'd normally protect while a chain is still climbing toward real usage.

There's an AI line attached to the announcement too. Restructuring plus an artificial intelligence push, packaged as one story. They aren't one story. More on that in a minute.

What Actually Got Cut

MegaETH has spent two years selling a specific promise. Sub-second blocks. Six-figure transaction throughput. A design that treats the EVM as something to optimize rather than inherit. The pitch lands. Dragonfly led a $20 million seed round back in 2024, and the project has kept a loud, technical, very-online presence ever since.

Loud and technical doesn't pay salaries.

Here's the structural problem. MegaETH has no token in circulation. No token means no treasury marks to lean on, no emissions to compensate contributors with, no liquid asset to borrow against. What exists is a runway, and runway is measured in months at current burn. Cut 20% of headcount and you buy time back. Sometimes a lot of it. Not speculation. Arithmetic.

That's the boring read, and boring reads tend to be correct. A team of a few dozen people at Bay Area and Singapore comp runs into the tens of millions annually once you aggregate salaries, tooling, audits, and infrastructure. Twenty percent off that line is real money. Real money is what extends a pre-token project's life.

So the cut itself isn't the scandal. The composition is the signal.

The AI Story Doesn't Hold

Look at the roles. If AI tooling genuinely absorbed a fifth of your output, you'd cut where output is fungible. Support. Content. Growth. Parts of the org where a model plus a prompt can plausibly replace a human and nobody outside the building notices.

But the cuts landed on the blockchain team. The wallet team. The people writing the client.

Here's the thing about AI coding assistants. They're excellent at scaffolding, boilerplate, test harnesses, the glue code nobody enjoys. They're bad at consensus edge cases, state management under load, and the fifteen months of debugging that separates a demo from a network. Any founder who's shipped infrastructure knows this. Cutting protocol engineers and calling it an AI efficiency play is a framing decision, not an engineering one.

The AI explanation is doing work it hasn't earned.

There's a harder question underneath it. If performance is the entire product, why is the performance team the first thing that shrinks? You can rebuild a wallet in a quarter with three good people. You can't rebuild the team that understands your execution layer's pathological cases. That knowledge doesn't live in a repository. It lives in heads, and heads walk out the door with a severance check.

History rhymes here. Every infrastructure cycle produces a wave of teams that raised on a thesis, spent two years on the thesis, then discovered the thesis needed another two years. The ones that survive cut surgically. The ones that don't cut the muscle and keep the marketing.

Which one is this? The disclosure doesn't say. That's the problem with a 20% number and no org chart attached.

Where The Margin Went

Zoom out and the pressure makes sense. Layer-2 economics have compressed hard since blobs got cheap. Fee revenue across the major rollups sits in the low single-digit millions per quarter on aggregate, split across a dozen chains competing for the same marginal user. Basis points of margin, not percentage points.

MegaETH's answer was always speed as a differentiator. Sequencer-level performance that unlocks order books, perps, and games that can't run on a 2-second block. Good thesis. Expensive thesis. And the market that was supposed to pay for it hasn't arrived at the scale the 2024 vintage of pitch decks assumed.

So who wins here? Short term, competitors with tokens and treasuries. A chain with a liquid asset can fund contributors through a drawdown without touching headcount. A chain without one has exactly one lever. This is that lever being pulled.

Who loses? The engineers, obviously. But also the users who were promised a mainnet timeline. Restructurings of this shape almost always push ship dates. Watch the commit cadence over the next sixty days. That's the tell, not the blog post.

What To Watch

The cut is defensible. The framing isn't. A team that tells you AI replaced its protocol engineers is either overstating the AI or understating the damage. Neither reads well.

What I want to see is whether MegaETH treats this as a consolidation or a retreat. Consolidation means the remaining team ships, the wallet stays on the roadmap, and the performance claims get validated on a public network with real load. Retreat means quiet scope cuts, a sliding mainnet date, and a pivot toward whatever's fundable next quarter.

Kong's disclosure was honest about the roles. That's worth something. Plenty of founders would have called it an efficiency initiative and moved on.

But honest disclosure of a bad decision is still a bad decision. Twenty percent is a big number. The next one is bigger if the chain doesn't ship. And the market doesn't grade on effort.