MegaETH Cut 20% of Its Staff and Sent $30M Toward AI. Its TVL Fell 96%.

I keep a running note file of projects that go quiet in interesting ways. MegaETH just earned a line in it.

Here's the shape of it. The company cut roughly 20% of its staff. It's also planning to steer about $30 million from its treasury into an AI marketplace that's supposed to go live in December. And the on-chain numbers, the ones that tell you whether anyone actually wants to use the network, went the other direction entirely.

Value locked in the MegaETH system sits near $8.8 million. Earlier, that figure was above $240 million. That's a drop of about 96%.

Layoffs alone wouldn't get my attention. Startups cut headcount all the time and most of them survive it. But a 96% collapse in deployed capital plus a $30 million bet on a new product category plus a chain that hasn't finished proving its core pitch? That's a different story. That's a company asking its users to believe two things at once.

What 20% actually costs a chain

Headcount math is easy to report and hard to interpret. Twenty percent of a team is a number. Which twenty percent is the actual question.

An execution layer needs a very specific mix of people. You need engineers who tune throughput. You need people who think about finality and what happens when a transaction is ordered but not yet settled on Ethereum. You need security reviewers, developer tooling folks, and someone who answers the phone when the sequencer starts acting strange at 3am. A sequencer, in simple terms, is the part of a rollup that puts transactions in order before they're posted to the base chain. It's the machinery that makes the network feel fast. You can't outsource that and you can't pause it for a quarter.

So when a team sheds a fifth of its staff while standing up a brand new AI product, the honest question isn't "did they cut too deep." It's "what got deprioritized." If the answer is performance tuning or developer support, that's a real cost. If it's marketing, nobody outside the company will ever notice.

Then there's the money. Roughly $30 million out of treasury, aimed at a marketplace that doesn't exist yet. That's the bigger bet here. The layoffs save maybe a few million a year. The AI allocation is an order of magnitude larger than the entire value currently locked on the network it's supposed to serve. Sit with that for a second. The new product is being funded at more than three times the size of the thing it's attached to.

Here's why the plumbing matters. A treasury isn't free capital. It's either raised money or token-denominated runway, and spending it on a new business line means the core network has a shorter window to prove itself. Users and holders should want a clear answer on how the two efforts connect, not an assumption that AI success magically lifts the chain.

The number that needs caveats, and still stings

I'll be fair to MegaETH on the TVL figure, because TVL is a slippery metric and anyone who treats it as gospel is being lazy. Locked value moves when token prices move. It moves when a rewards program ends. It moves when three large liquidity providers decide to rotate somewhere else in a single afternoon.

But 96% isn't normal rotation. A decline that size means incentives stopped doing the work, or the apps on top weren't sticky, or both. The useful follow-up questions are boring and specific. Are there applications with repeat users who show up without a yield subsidy? Can an independent developer ship something on this chain without a full-time guide from the core team? Is the throughput claim holding under real load?

Those answers say more about the scaling thesis than any announcement about a new business category.

Timing is doing MegaETH no favors

The change comes at a time when Ethereum itself is under pressure. Ether ETFs have bled roughly $1.85 billion in outflows, which tells you institutional appetite for the base asset has cooled. That's the backdrop. Every layer 2 is now competing for a shrinking pool of patient capital, and the ones with unproven traction feel it first.

And yes, there's a real case for crypto plus AI. Machine-to-machine payments need settlement rails. Agents need verifiable payment controls, which is the exact problem Circle has been publicly working through. The intersection is legitimate.

But a trend isn't a business plan. An AI marketplace has to answer plain questions. Who's the buyer? What service are they purchasing? What's the take rate? Why does this specific chain make the experience better instead of merely possible? If the answer boils down to "AI is hot and we've a treasury," that's not a pivot, that's a hedge.

There's one more commitment worth pinning to the wall: a decentralization milestone reportedly targeted for March 2027. Timelines like that decide who controls upgrades and critical infrastructure. A new product line can't quietly replace the need to explain governance, operational resilience, and what protections users actually have. If AI becomes the headline for the next fifteen months, that governance date is the thing to hold them to.

My read: watch the customers, not the roadmap

Here's my honest take. The layoffs are defensible. The direction might even be right. What I don't buy is the sequencing. You don't launch a $30 million marketplace before you've shown the world a chain with durable demand, unless the marketplace is genuinely the demand.

Who wins here? Teams with live apps and paying users, because they don't need to explain themselves. Who loses? Anyone holding a token whose value story just got rewritten from "fast execution layer" to "AI marketplace parent company" without a shareholder vote. That's not a crime. It's just a much harder thing to price.

For everyday users, nothing changes overnight. The chain still runs. Transactions still settle. But builders deciding where to deploy next quarter are watching this closely, and builders are the only real leading indicator a rollup has.

So here's what I'd actually do with this information. Ignore the roadmap graphics. Ignore the AI press cycle. Watch two dates. December, when the marketplace either ships with named customers or slips quietly. And March 2027, when the decentralization commitment either lands or gets restated. In practice, those two data points will tell you more about MegaETH than every announcement between now and then.

One more thing. When a team points more treasury at a new product than the total capital currently sitting on its network, the treasury has quietly become the product. That's worth remembering the next time someone quotes a valuation.