Fusaka, the Upgrade That Outgrew Its Own Roadmap

On December 3, 2025, Ethereum shipped Fusaka, the upgrade the ecosystem had waited on since the sharding dreams of 2017. Two months later, on February 3, 2026, Vitalik Buterin posted that the rollup-centric vision Ethereum had organized itself around for half a decade "no longer makes sense." Those two events are the same story, and it is one of the strangest in crypto: an upgrade so successful it undercut the worldview that designed it. Ten months on, with the dust settled and the charts honest, here is the full picture, including the part where the price ignored the fundamentals entirely, because that part has the better lesson for traders.

What Fusaka actually did

The centerpiece is PeerDAS, peer data availability sampling, and it is the mechanism Ethereum had been chasing for eight years. Before it, every node had to download every blob of rollup data to verify it was available, which capped how much data the chain could safely carry. With PeerDAS, nodes sample small chunks and verify availability probabilistically, the way you check a shipment by inspecting random boxes instead of opening all of them. That one change made much higher blob counts safe, and the follow-up "blob parameter only" forks in December and January cashed the check: the blob target rose to 14 per block with a maximum of 21, several times the pre-Fusaka capacity. Alongside, the block gas limit stepped up to 60 million with a new per-transaction cap, so the base layer itself got roomier, not just the data lanes.

The results showed up fast. Layer 2 data posting got dramatically cheaper, L1 fees collapsed to frequently under ten cents and occasionally to a cent in quiet hours, daily transactions jumped roughly half again, and active addresses climbed about 60% in the months after activation. For the first time since fees became Ethereum's defining complaint, the base chain was competing on cost with the chains built to undercut it.

The irony: it killed the story that justified it

Recall the official plan, covered in our Ethereum series: the base layer stays slow and maximally secure, and rollups do the scaling. Fusaka was built as that plan's crown jewel, cheaper data for the rollups. Except it worked so well that the base layer stopped being slow, and that broke the plan's premise from the inside. Vitalik's February post named both cracks: Ethereum's L1 was now scaling aggressively on its own, and the L2s had decentralized far slower than promised, with only three majors reaching even Stage 1 by the end of 2025 and most still running with centralized training wheels. His sharpest line drew the boundary: high-throughput chains connected by multisig bridges are not "scaling Ethereum." The new assignment for L2s, in his telling, is to justify themselves on something other than scaling: privacy, app-specific performance, ultra-low latency. The post drew real backlash from teams whose business model was the old story, which is what happens when a roadmap becomes an industry. And the shift was already visible elsewhere: ENS scrapping its own planned L2 to stay on mainnet, the story our ENS post told, was this same tide reaching one protocol early.

The price told a different story

Now the trader's half, because the chart did not read the changelog. The sequence followed a pattern old enough to have a rhythm: big upgrade, initial pump, slow bleed, loud FUD, and the eventual reversal that embarrasses the FUD. ETH popped several percent on activation day from around $3,000, volume spiked, and the coverage called it structurally bullish, which it was. Then came the bleed: through late December and January the price ground down while the fundamentals kept improving, new addresses more than doubling since launch while the candles stayed red.

Was it engineered panic? The honest answer our myths post would give: possible, unnecessary. Sell-the-news is the default behavior after every major upgrade, Shanghai, Dencun and Pectra included, because the crowd buys the rumor for months and has nothing left to buy once it ships; the celebration is the exit liquidity. Add January's usual risk-off tone, a strong Bitcoin bleeding the alts through a weak ETH/BTC pair, and the pattern explains itself without a villain. Meanwhile the on-chain record showed large holders accumulating into exactly the dips retail was panic-selling, which is the opposite of distribution and the oldest tell there is. The crowd was positioned long into the event, and as the futures post keeps saying, the market does not hunt you, it hunts crowds.

Ten months later, the scoreboard

What held: L2 activity kept growing even as the narrative deflated, staking kept climbing, developer activity stayed high, and the fee-burn thesis keeps compounding quietly now that cheap blockspace is filling with actual usage. What changed: the roadmap now runs through the base layer. The next named upgrade, Glamsterdam, is an 18-change bet on L1 itself, block-level access lists and enshrined proposer-builder separation, with developers aligned on pushing toward a 200 million gas limit once those land, roughly tripling base throughput again. The rollup-only era produced the tools; the L1-scaling era is spending them. Whether that vindicates or buries the old roadmap is a debate for the forums. The chain, meanwhile, just got cheap enough to use, which was the point all along.

Why a trader should care

Three durable lessons. First, protocol upgrades are scheduled volatility: the dates are public months ahead, the crowd positions into them, and the sell-the-news resolution is frequent enough to plan around, so fading the celebration and buying the despair at real levels is a setup, not a superstition. Second, Fusaka is the cleanest recent case of the divergence rule from the DeFi post and the ENS story: fundamentals improving while price bleeds is normal, not contradictory, and the trader who checks which one the leverage is positioned against reads the next move better than the one who argues the changelog at the chart. Third, keep upgrade calendars on your event list the way CPI and FOMC already are, because a known date that reprices a major is exactly the event-risk sizing case from the playbook. The technology and the chart run on different clocks. Respect both, and never expect one to keep the other's time.