Ethereum is improving, but ETH holders aren't smiling.

Author: Baihua Blockchain

BitMine holds 6.06 million ETH, with unrealized losses exceeding $4 billion. Over the past week, this largest corporate holder of Ethereum spent $45.9 million repurchasing its own stock, while spending only $24.5 million on acquiring tokens.

Repurchasing its own shares yields greater returns than betting on its own ethos.

Yet, 1.37 million ETH are queued on-chain waiting to be locked into the network, facing a 24-day wait. The ETH/BTC ratio has slipped to 0.029, approaching a three-month low. Traditional channels are withdrawing while on-chain capital is being locked up. Never has Ethereum been so user-friendly, yet ETH refuses to trade at a premium.

Bleeding While Queuing

In October, U.S. spot Ethereum ETFs recorded net outflows exceeding $1.2 billion, led by BlackRock's ETHA. Since January this year, inflows have occurred in only two months, with outflows dominating the rest.

The rules are explicit: ETFs cannot stake, meaning holdings generate no yield. Treasuries offer solid risk-free returns; who would willingly hold a fund that only depreciates without producing profit?

Corporate treasuries have also hit the brakes.

BitMine holds approximately 6.06 million ETH, accounting for nearly 5% of the circulating supply, nearing its self-imposed cap. Most of it is locked in staking to earn interest. With an average cost basis of around $3,500 against a current price near $2,300, the paper loss exceeds 40%.

Ranking second, SharpLink holds about 889,000 ETH, equally mired in unrealized losses. Together, the two entities account for nearly 6% of the network's total circulating supply.

Management's calculus is straightforward: the stock trades at just 0.88 times book value. Using cash generated from staking to buy back discounted shares is far more lucrative than continuing to accumulate positions in the open market.

Year-to-date in 2026, BitMine has spent nearly $400 million on share repurchases, vastly surpassing its token acquisitions during the same period. The company still buys ETH, but the pace has noticeably decelerated.

Rational on a micro level. But the spot market's largest buyer is stepping back.

On-chain data paints a different picture.

In late October, roughly 1.37 million ETH queued to activate staking, facing waits exceeding 24 days; withdrawals numbered only about 13,000, clearing within a day. The line for entries spills into the streets, while the exit gate sees no one.

The Ethereum protocol can process a maximum of approximately 57,600 ETH in and out daily. A queue of 1.37 million indicates entry demand drastically outstrips channel capacity. Even with the annualized yield diluted to 2.61%, capital still opts to sacrifice liquidity in exchange for on-chain interest.

Early October brought another stress test. Due to a security incident, MetaMask voluntarily withdrew about 17,000 validators, temporarily pushing the withdrawal queue to 850,000 ETH. The network smoothly absorbed the shock within three weeks, keeping the total staked amount stable above 43.7 million ETH.

This was not a panic exit, but a temporary fluctuation triggered by a security incident. The network swallowed the impact without hiccuping.

ETH's price isn't dictated by the most steadfast holders, but by whoever is selling at the margin.

Fees Are Lower, Now What?

Transaction fees on the Ethereum mainnet fell by over 60% year-over-year in Q3. In October, the average transaction cost was $0.25. During the network congestion of 2021, the same operation routinely exceeded $53. The fee to send an ETH three years ago could now cover over 200 transfers.

How cheap has it gotten? ENS doesn't want to build an L2 anymore.

Ethereum Name Service (ENS) originally planned to launch a dedicated Layer 2 network called NameChain, moving domain registration and resolution off-chain. In February 2026, the team announced abandoning that route, opting instead to fully deploy ENS v2 directly on the mainnet.

With mainnet interactions costing just a few cents, building a private L2 entails disproportionate costs: maintenance expenses, cross-chain bridge security risks, and fragmented liquidity. Staying on the mainnet allows the domain system to integrate directly with lending, trading, and identity protocols without waiting for cross-chain confirmations.

ENS is the first major protocol to abandon an L2 due to cheaper mainnet costs, and certainly not the last.

The benefits are obvious, but the costs hide within the tokenomics.

Ethereum features an automatic burn mechanism: the base fee of each transaction is directly destroyed by the network. When rates were high, more ETH was burned than newly issued, making ETH scarcer with usage. This fueled the market's once-celebrated narrative of a hyper-deflationary currency, with Ethereum briefly entering deflation during the 2021 peak.

Now it's the opposite. The base fee has dropped to just 0.06 Gwei, requiring 13.63 Gwei to match new issuance—a difference of two orders of magnitude. The network sees a net inflation of approximately 2,954 ETH daily, creating nearly $7 million in sell pressure at current prices.

The deflation engine has stalled.

Previously, the busier Ethereum became, the scarcer ETH grew. Today, user numbers remain high, but the burn contribution per transaction approaches zero.

The transmission chain between network usability and coin valuation is broken. The better Ethereum becomes to use, the wider this gap widens.

The Tides Are Receding for L2s

In February 2026, Vitalik Buterin published an article characterizing the L2 trajectory, stating that branded sharding no longer makes sense.

What is branded sharding? It refers to individual projects building their own L2s while sharing Ethereum's security. It sounds like a division of labor, but in practice, it brings countless issues.

Vitalik was blunt: many L2s claim tens of thousands of TPS, but migrating assets from the mainnet relies on multi-sig wallets controlled by a handful of individuals. Nominally an Ethereum expansion, they practically build centralized external databases. After running for years, decentralization for most rollups has advanced little, and permissionless fraud proofs remain elusive.

Once the mainnet itself becomes cheaper and more scalable, the necessity of L2s lacking independent security guarantees comes into question.

A reshuffle promptly followed.

Kroma shut down services, Kinto went completely offline, and Blast's TVL plummeted by 97% after its airdrop tide receded.

Top-tier protocols are also contracting. Besides ENS returning to the mainnet mentioned earlier, Synthetix and prediction market Trueo have scaled back deployments on long-tail L2s, refocusing on the mainnet. Aave hasn't migrated entirely but is gradually shutting down underperforming small deployments, concentrating core liquidity on the mainnet and a few top-tier L2s.

The direction is clear: those that can return do, while those that can't contract. Mainnet fees have become so low that flagship applications stay put, ironically weakening the rationale for L2s. The pros and cons of low fees ultimately point to the same contradiction.

200 Million Gas Stranded

Ethereum's next major upgrade, dubbed Glamsterdam, aims to raise the block gas limit from 60 million to 200 million, tripling mainnet throughput. It was successfully activated on the Sepolia testnet on October 6.

Then it halted.

Plans to launch it on the Hoodi testnet on October 26 were abruptly paused, with no new timeline provided.

The issue stems from a security audit. ChainSecurity's report dated October 13 noted that EIP-8037 in the upgrade alters how the VM processes failed transactions, potentially exposing vulnerabilities in 13 contracts already live on the mainnet.

In simple terms, these contracts should automatically roll back to protect user funds upon transaction failure. The new rule changes the resource refund mechanism during rollback, potentially disabling this safeguard. Once deployed, blockchain code cannot be altered; development teams cannot patch third-party contracts retroactively and must instead revise the proposal itself.

The delay for Hoodi means the December mainnet upgrade window is closed, leaving 200 million gas without a schedule. Ethereum chose security over speed.

Ethereum has the confidence to afford the wait. The mainnet hosts over half of the network's stablecoin supply and more than $20 billion in tokenized real-world assets, including treasury products from BlackRock and Franklin Templeton. However, Q3 stablecoin market cap continues to bleed, and the encroachment by low-cost networks like TRON is visibly creeping. The moat remains, but the water level is dropping.

The foundation is also adjusting. Vitalik stated he intends to transform it into a smaller ship, focusing on security and core R&D while reducing direct ETH dumping.

The network improves, but the token model lags behind, and security cannot be compromised.

Evidence of usability keeps mounting, but the price appreciation has yet to follow.