Investing in Ethereum

On March 12, 2026, BlackRock listed the iShares Staked Ethereum Trust on Nasdaq under the ticker ETHB, seeded with $107 million, and for the first time a mainstream American brokerage account could hold ether and collect the network’s staking reward without touching a wallet. The fund stakes between 70% and 95% of its ether through Coinbase Prime and Figment, keeps 18% of what that earns, and pays the rest out monthly in cash. Nineteen months earlier the same firm had launched a plain ether fund, ETHA, that held the coin and forfeited the yield entirely, because the SEC had not yet decided whether staking was a securities offering. It decided on May 29, 2025, and again, jointly with the CFTC, on March 17, 2026: it is not.

That is the wrapper story, and it is a good one. The asset story underneath is harder. Ether closed at $4,831 on August 22, 2025, twenty dollars above its November 2021 high, then fell to $1,566 on June 26, 2026 on our own tape, a 67% drawdown in ten months against 53% for bitcoin. Ether has underperformed bitcoin in every calendar year since 2022, and the mechanism is not a mystery: the network deliberately moved its economic activity onto cheaper side-chains in March 2024, and the fees that used to be burned to make ether scarce went with it.

So the reader arrives with two questions the ETF prospectus does not answer: is the yield real, and who pays it; and if the fee base has moved to other chains, what exactly are you buying. This guide answers both with the arithmetic shown. The hub’s flagship, Investing in Crypto, covers the asset class, fraud and custody; Investing in Bitcoin covers the other coin, and Investing in Crypto Through ETFs and Treasury Companies covers the wrappers as a group. This one stays on ether.

What you own when you own ether is different from what you own with bitcoin, and most of the guide follows from the difference. Bitcoin is a ledger with a fixed supply and one job. Ethereum is a general-purpose computer anyone can deploy a program to, and every use of it, a stablecoin transfer, a token swap, a loan, an NFT mint, pays a fee in ether. Ether is the fuel, the unit the fee meter reads in, and, since September 2022, the collateral that secures the machine.

That gives ether three economic lines bitcoin lacks. Issuance: the protocol pays new ether to validators, the parties who lock ether to order transactions and vouch for the chain. Fees: since August 2021 the base portion of every fee is burned, destroyed rather than paid to anyone, which shrinks supply when the chain is busy; the tip goes to the validator. MEV, maximal extractable value: what a validator earns by choosing the order of transactions, which specialised builders bid for through a market called MEV-Boost. Add issuance, tips and MEV, divide by the ether staked, and you have the staking yield.

Bitcoin’s investment case is scarcity plus adoption. Ether’s is closer to an equity claim: the network earns fees, part of those fees retire supply, and stakers are paid a share of the rest. Which means ether can be analysed, and found wanting, in a way bitcoin cannot. From early 2024 to 2026 the fees fell by about 95%, and that is the tension this guide keeps returning to.

What ether has returned, against bitcoin and against equities, comes before any of the upside is sold. The published number is enormous and true: Coin Metrics’ daily series starts at $1.20 on August 8, 2015, and the December 31, 2025 close was $2,968, a 2,470-fold gain, almost all of it in 2017, which returned +9,034% as the network became the venue for the initial-coin-offering boom. Nobody reading this bought in 2015, so the honest question is what ether has done since it became an asset an outsider could reasonably own.

From the end of 2017 to the end of 2025, ether went from $742 to $2,968, four times your money, about 18.9% a year compounded. The path is the problem. It fell 94.0% from a $1,380 close on January 13, 2018 to $83 on December 15, 2018; rose 58-fold to a $4,811 close on November 8, 2021; fell 79.4% to $993 on June 18, 2022; recovered to a new high of $4,831 on August 22, 2025, twenty dollars above the 2021 peak, which is a nominal round trip to nowhere across four years; and then fell 62% by February 2026 on Coin Metrics’ data and 67% by June on ours.

Three drawdowns of 62% or worse in eight years, and a holder who bought the January 2018 top spent three years under water, was made whole only in early 2021, and has since sat through two more drawdowns of 60% or worse.

Against bitcoin

The comparison that matters most to a crypto allocator is the one ether has been losing. Bitcoin’s calendar returns over the same years, from the same Coin Metrics files, were +1,337% in 2017, −73.5% in 2018, +94.4% in 2019, +304.9% in 2020, +59.7% in 2021, −64.4% in 2022, +155.5% in 2023, +121.2% in 2024 and −6.3% in 2025. Ether beat bitcoin in 2017, 2020 and 2021, the three years its fee base was growing, and has lost in each of 2022, 2023, 2024 and 2025, the years the fee base was shrinking or being redirected. In 2026 to date on our tape ether is down 16.4% against bitcoin’s 10.3%. Four straight losing years against the asset it is usually bought alongside is the single most important fact in this guide, and the section on layer 2s is the explanation.

Against equities

Over the three years to December 31, 2025, ether returned +148%, about 35% a year, against a 2025 total return of +17.9% for the S&P 500 (S&P Dow Jones Indices, via SlickCharts) and a 2022 loss of 18.1%. Over the eight years from 2017 the two are closer than the headline suggests. Ether’s 18.9% a year came with two drawdowns above 79%. The S&P’s roughly 14% (calendar total returns 2018 to 2025, S&P Dow Jones Indices via SlickCharts and ChartRow) came with two of 25% or more, 34% in March 2020 and 25% in 2022. And ether’s realised volatility runs roughly three to four times equity volatility, so the risk-adjusted excess return since 2018 is modest and depends entirely on the start date.

Two more gaps between the marketing number and the real one. The staking reward adds two to four points a year in ether terms from 2022, but it is paid in ether, so in 2025 a staker earned about 3% and still finished down 8%. And ether is the survivor of the 2017 smart-contract platforms; the peers it is measured against, EOS, Cardano, NEO, Tezos, mostly went to near zero. Being the survivor is the case for ether. It is not evidence about the next platform.

Our view of the current cycle rests on numbers we recorded ourselves: the last twelve months from Invest Alternative’s own data store. Our crypto series is fed by CoinGecko once a day; the ether series has 374 observations from August 29, 2025 to September 8, 2026, and the crypto sub-index, which is rebased to 100 on September 2, 2025 and carries a 9.7% weight in our composite, is built from the same feed. These are our figures, one data point a day, and not a market-wide index.

Ether opened our tape at $4,510.74 on August 29, 2025, peaked at $4,707 on September 13, and then fell for nine months: $3,802 at the end of October, $2,968 on December 31, $1,931 at the end of February 2026 after a 14% single-day fall in bitcoin on February 6, and a low of $1,566 on June 26, 2026, a 66.7% peak-to-trough decline. Bitcoin on the same tape fell 53.0%, from $124,740 on October 7, 2025 to $58,566 on July 1, 2026. The rally since has been sharper in ether: from the June low to September 8 ether is up 58.5% to $2,481.66 and bitcoin 34.2%. Twelve months on, September 8 to September 8, ether is down 42.4% and bitcoin 29.3%.

The ratio on our tape

Dividing our two series gives an ETH/BTC ratio of 0.0401 on August 29, 2025, a high of 0.0405 on September 13, a low of 0.0258 on June 7, 2026, and 0.0316 on September 8. So even inside a drawdown that bitcoin led, ether lost a further 36% against it from the September high to the June low, and has recovered about two-fifths of that since. On our sub-index, which blends bitcoin, ether, Solana and XRP, the reading is 59.05, a 41.5% one-year decline that is still 45% above the June 7 low of 40.66.

What averaging in did

A lump sum into ether at the October 6, 2025 close of $4,513 on our tape was down 45% on September 8, 2026. Twelve purchases of $1,000 on the sixth of each month from October 2025 to September 2026 (highest $4,513, lowest $1,582 on June 6) averaged about $2,302 per ether and were up 7.8% on the same day. The lump sum wins in a rising market; in an asset that has fallen 60% or more four times, the schedule is what keeps you in the trade.

The yield that distinguishes ether from every other large crypto asset comes out of proof of stake, and its size is set by rules you can check. Until September 15, 2022, Ethereum was secured the way bitcoin still is: miners burned electricity to add blocks and were paid about 14,896 new ether a day for it in August 2022 (Coin Metrics). On that date, in the Merge, the network switched to proof of stake: anyone who deposits 32 ether into the staking contract runs a validator, software that proposes and attests to blocks, and is paid new ether for doing it honestly. Issuance fell to 1,611 ether a day in October 2022, a 90% cut, because validators have no electricity bills to cover.

The reward has three parts, and only one is fixed by the protocol. Base issuance is a formula: the total paid to all validators grows with the square root of the amount staked, so each validator’s share shrinks as more ether joins. With 39 to 43 million ether staked the base rate is roughly 2.6% to 2.8% a year; it was above 4% in 2023 when half as much was staked. Tips, the priority fees users pay to get into a block, add a few tenths of a point in a quiet year, and MEV-Boost, the market in which block builders bid to order a validator’s block, added roughly half a point to a point in 2025 and 2026.

Public trackers put the base rate at about 2.78% in May 2026, with 3.3% to 4% all-in for a well-run solo validator (Datawallet, KuCoin Research). By early September 2026 it had slipped further: beaconcha.in’s ETH.STORE reference rate, the seven-day realised return of an average validator, printed 2.61%.

How much is staked

About 42.9 million ether, roughly 35% of the 122 million or so in circulation, was staked by early September 2026 across some 910,000 validators (validatorqueue.com and beaconcha.in, September 2026). The climb has been steady: 35 to 36 million through 2025, then 38.9 million, 32% of supply, across some 897,000 active validators in May 2026 (Datawallet; Coin Metrics gives 121.72 million supply on May 23, 2026).

Demand to enter has at times exceeded the protocol’s rate limit: on May 20, 2026 the entry queue held 3.59 million ether with a 62-day wait (Datawallet), and in late July it still held about 2.5 million ether, a 43-day wait, with the exit queue near empty (validatorqueue.com data via Cointelegraph, July 2026). That queue is the mechanism by which the yield compresses; every validator that clears it takes a slice of a pot that grows more slowly than the number of slices.

The yield is real. It is paid by the protocol in new ether plus fees, not by a company that can go bankrupt. The two honest caveats are that about 90% of it in 2026 is issuance, the staker’s share of a dilution imposed on every holder who does not stake, and that it is denominated in ether: 3% a year on an asset that fell 67% in ten months is a rebate on a loss, not income.

The claim that ether is “ultrasound money”, a shrinking supply, was the centre of the 2021 to 2023 investment case, and the supply mechanics show it is no longer true. The London upgrade of August 5, 2021 introduced EIP-1559, under which the base fee of every transaction is burned; when the chain is busy, the burn exceeds issuance and supply falls. Coin Metrics’ fee series shows 9,995 ether a day in May 2021 and 7,799 in May 2023, against post-Merge issuance of 1,600 to 2,400 a day, and between the Merge and April 2024 supply fell from 120.52 million to about 120.07 million.

Then the fee base went. The Dencun upgrade of March 13, 2024, described in the next section, moved the layer-2 networks off the expensive part of the main chain, and the fees fell off a cliff: 2,264 ether a day in April 2024, 516 in June 2025, 298 in August 2025 at the price high, 123 in December 2025, 269 in May 2026. Issuance, meanwhile, rose as more ether was staked, from about 2,500 a day in 2024 to 2,808 a day in May 2026.

The net is a supply that has grown every month since April 2024, by about 2,700 ether a day in the spring of 2026 (March 1 to May 23 on Coin Metrics’ supply series), or roughly 0.8% a year. Supply stood at 121.72 million on May 23, 2026, 1.0% above its level at the Merge; the corporate treasury BitMine cited “122.0 million” in its September 7, 2026 holdings release, which is consistent with the same run rate.

Two things about that chart. First, 0.8% annual inflation is low, about the same as bitcoin’s 0.83% over the twelve months to May 2026 on the same Coin Metrics files; the point is not that ether is being debased but that the deflationary case, which drove much of the 2021 buying, has inverted and stayed inverted for more than two years. Second, the burn is a function of gas prices (gas is the unit in which Ethereum meters computation, and its price rises with congestion), not transaction counts: the main chain processed 2.14 million transactions a day in May 2026, nearly double the 1.13 million of August 2022, and paid about a sixth as much fee (269 against 1,738 ether a day). Usage is fine. The price of usage is what collapsed.

Why the fees left, where they went and whether they come back is the heart of the guide, because the answer decides whether ether is worth owning at all. Ethereum’s main chain processes on the order of ten to twenty transactions a second, and in 2021 a busy day meant a $50 token swap. The community’s answer was not to make the main chain faster but to build layer-2 networks, separate chains such as Arbitrum, Optimism and Coinbase’s Base that execute transactions cheaply on their own servers, bundle the results, and post a compressed record back to Ethereum for security. The layer 2 pays Ethereum for the posting; the user pays the layer 2; the difference is the layer 2’s margin.

Until March 2024 that posting was expensive, because the layer 2s wrote their data into ordinary Ethereum transactions at ordinary gas prices, which is why the fee base held up through 2023 even as users moved. The Dencun upgrade of March 13, 2024 introduced “blobs”, a separate, cheap data lane built for exactly this. Layer-2 posting costs fell by more than 90% overnight, a Base transaction fell to a few cents, and Ethereum’s fee revenue went with it: 5,324 ether a day in March 2024, 2,264 in April, 1,212 in May. On quiet days the blob lane’s revenue to the base chain has been measured in single-digit ether (Yellow Research, 2026).

Where the money went

The layer 2s now do most of the work. Their combined transaction count ran at five to ten times the main chain’s through the first quarter of 2026, and Base alone has processed more transactions in a day than Ethereum itself (Yellow Research, 2026). L2BEAT, the reference tracker, counted 73 active rollups (the technical name for these networks, which roll many transactions into one posting) securing more than $48 billion in the spring of 2026. Arbitrum One and Base are the two that matter, together about three-quarters of the total, each between $10 billion and $17 billion depending on the month and the measure, and in L2BEAT’s late-July snapshot Base ($11.6 billion) had edged ahead of Arbitrum One ($10.6 billion) (L2BEAT data as summarised by Eco, SpotedCrypto and Everstake, 2026).

Follow Base’s margin. A Base transaction costs a cent or less for a simple transfer and typically under $0.05 (Base documentation, 2026), and Coinbase, which runs the sequencer, the server that orders Base’s transactions, keeps the difference between what users pay and what it pays Ethereum for blob space. Coinbase does not break Base out in its results: the sequencer take sits inside “other transaction revenue”, which was $47 million in the second quarter of 2026 and, in the company’s words, “largely driven by lower Base revenue” (Coinbase Q2 2026 shareholder letter, July 30, 2026).

Analysts working from the same disclosures put Base’s gross sequencer revenue near $19 million in the fourth quarter of 2025, about $200,000 a day (Talos, 2026); one tracker’s 2026 figure of about $185,000 a day (SpotedCrypto) is the same order of magnitude. The direction is beyond dispute: the margin on Ethereum’s usage accrues to the layer-2 operator, and Coinbase is a listed company with shareholders, not an ether holder.

This is the value-accrual debate. The bull case is that Ethereum is doing what it set out to do: it is the settlement and security layer, the layer 2s are its customers, and as their volume grows the blob lane will fill and price rise until the base chain earns again, the way a port earns from the ships rather than the cargo. The bear case is that blob space is deliberately kept abundant, that the layer 2s are in a price war with each other and with rival chains, and that a settlement layer whose customers pay it single-digit ether a day is a utility, not an asset. On the 2024 to mid-2026 fee numbers the bear case has been winning, and ether’s price against bitcoin says the market agrees.

Each upgrade that has shipped or is scheduled changes something a holder is paid for or exposed to. Ethereum now aims to fork roughly twice a year, each fork named for a city and a star.

Pectra, activated May 7, 2025, matters most for stakers. It raised a validator’s maximum effective balance from 32 to 2,048 ether, so a large staker runs one validator instead of 64 and rewards above 32 ether compound instead of sitting idle; the active validator count, about 897,000 in May 2026 and some 910,000 in September, has grown far more slowly than the ether staked, because consolidation replaces many 32-ether validators with fewer, larger ones (the average validator held about 47 ether in September against 43 in May, on the figures above). It also introduced account abstraction for ordinary wallets (EIP-7702), letting a wallet pay fees in a stablecoin or batch actions in one signature, and doubled the blob target from three to six per block, which lowered layer-2 costs further and did nothing for the fee base.

Fusaka, activated December 3, 2025 (CoinGecko, Everstake), introduced PeerDAS, under which validators verify small samples of the layer-2 data rather than downloading all of it, so blob capacity can be raised in steps without heavier hardware; a series of “blob parameter only” forks does exactly that through 2026. Each step is more cheap data for layer 2s, good for throughput and neutral at best for the per-unit price of blob space on which the fee-return thesis rests.

Glamsterdam is targeted for the second half of 2026; as of August it was in devnet and testnet hardening, and EIPsInsight’s published schedule carries the Sepolia and Hoodi test networks on September 21 and October 5 and mainnet on November 4, 2026 (EIPsInsight; Everstake; the Ethereum Foundation had not announced a mainnet date as of September 9). It moves the MEV-Boost market into the protocol itself (ePBS) and adds block-level access lists for parallel execution, a base-chain scaling upgrade. What a holder should take from the roadmap is simple: the network is being engineered for more throughput at lower unit prices, and the fee-return thesis needs demand to grow faster than the capacity being added, which since 2024 it has not.

The price is set at the margin by a few large, identifiable balance sheets, and 2025 added a new kind. Start with the stakers, the 43 million ether described above: about a third of it sits in liquid staking pools, of which Lido’s stETH is by far the largest at 8.89 million ether, 61.7% of a $25.6 billion liquid-staking market on June 15, 2026 (Bitcoin.com News, June 2026), and between 21% and 23% of all staked ether depending on the tracker (Datawallet puts it at about 23%; Lido’s own half-year figures show its share slipping from 23.9% to 21.2% by June 30, 2026). Exchange staking programmes, Coinbase’s and Binance’s above all, are the next-largest operators.

Then the funds: the US spot ether ETFs, launched July 23, 2024 and described in the next section, are the marginal buyer the market watches, because Farside Investors’ flow tables are public by the next morning.

Then the corporate treasuries, the 2025 novelty. BitMine Immersion Technologies, a former bitcoin miner chaired by the strategist Tom Lee, began buying ether in mid-2025 and by September 7, 2026 reported holding 5,929,198 ether, 4.9% of the entire supply by its own count, of which 5,067,309 was staked, with total crypto, cash and securities of $15.7 billion (company release, September 8, 2026, via PR Newswire and an SEC 8-K exhibit). SharpLink Gaming, the second-largest, held about 870,000 to 890,000 ether in August and September 2026 (868,699 on The Block’s treasuries tracker; 888,938 on CoinGecko’s), and the listed companies together held about 7.77 million ether, 6.5% of supply, as of August 18, 2026 (CoinGecko treasuries data).

These companies raise equity, buy ether with it, stake it, and report the yield as revenue; the stock trades at a multiple of the ether it holds, the “mNAV”, and when the multiple falls below one the incentive runs the other way. BitMine publishes its holdings weekly; we could not find a reliable published mNAV for it as of September 9, 2026. Investing in Crypto Through ETFs and Treasury Companies has the mNAV arithmetic and the 2026 unwind for the bitcoin versions; the ether versions carry the same structure with a smaller cushion.

Who sells? In the 2025–2026 drawdown it was ETF holders redeeming through the winter (the monthly figures are in the next section), leveraged traders liquidated in the October 10, 2025 cascade and again in February, and treasury companies whose premiums went away. Who bought at $1,566 in June 2026 is less visible, but the entry queue in May was the longest of the year: the stakers kept staking. That is a structural bid bitcoin does not have, and a structural overhang: 43 million ether that can be unstaked, through a rate-limited exit queue, if the yield ever stops being worth the price risk.

Staking terms now separate the spot ether ETFs, for most readers the right wrapper, because a fund that does not stake charges a fee to forgo the asset’s only income. How a spot crypto ETF is built, creates and redeems shares and tracks its coin is the ground of Investing in Crypto Through ETFs and Treasury Companies; the short version is that the SEC approved the products in two orders, May 23 and July 22, 2024, and that they are grantor trusts, so you own a pro-rata share of ether in cold storage and are taxed as if you held the coin. BlackRock’s ETHA and Fidelity’s FETH charge 0.25% a year, Grayscale’s mini trust (ticker ETH) 0.15%, and the legacy Grayscale ETHE 2.50%. None staked at launch, because the SEC had sued two exchanges over staking programmes in 2023 and the issuers would not test it.

The permission came in three steps. On May 29, 2025 the SEC’s Division of Corporation Finance stated that protocol staking, including through a custodian, is not a securities offering. Grayscale switched staking on in ETHE and its mini trust on October 6, 2025, and on January 5, 2026 ETHE paid the first staking distribution by a US ether ETP, $0.083178 a share, about $9.4 million in total for rewards earned between October 6 and December 31 (Grayscale free-writing prospectus filed with the SEC, January 5, 2026).

On March 12, 2026 BlackRock launched ETHB, the iShares Staked Ethereum Trust, as a separate fund rather than converting ETHA: it stakes 70% to 95% of its ether through Coinbase Prime and Figment, passes 82% of the rewards to holders as a monthly cash distribution, and charges a sponsor fee of 0.25%, waived to 0.12% on the first $2.5 billion for twelve months from March 12, 2026 (iShares product page; CoinDesk, March 2026). On March 17, 2026 the SEC and CFTC issued a joint interpretive release confirming that staking a non-security digital commodity such as ether does not trigger registration.

Fidelity filed an amended registration statement, made public in August 2026, to let FETH, then a fund of about $898 million, stake up to all of its ether and pay 85% of gross rewards to shareholders quarterly (CoinDesk, August 12, 2026); as of September 9, 2026 that amendment was pending. Of the others, 21Shares’ TETH began paying staking distributions on January 9, 2026 ($0.010378 a share, 21Shares notice), Bitwise withdrew its staking proposal in September 2025, and the remaining issuers had not announced effective staking terms as of September 9, 2026.

Reading the terms

Three numbers now differ from fund to fund and live in the prospectus and nowhere else: whether the fund stakes, what share of its ether it stakes (some must stay liquid for redemptions), and what cut of the reward the sponsor keeps. On a 3% gross yield, ETHB’s 82% pass-through on, say, 85% of assets delivers about 2.1% in cash before the sponsor fee; Fidelity’s proposed 85% on up to 100% delivers 2.55%; a plain ETHA delivers nothing and charges 0.25%. Over four years the gap between a staking and a non-staking fund is roughly 8% to 10% of the position, larger than the entire fee difference between any two funds. The distribution is ordinary income (see the tax section), so in a taxable account the after-tax gap narrows to perhaps 5% to 6%; in an IRA it does not narrow at all.

Flows

Ether ETF flows turned negative in November 2025 with the price, $1.42 billion out that month and $616 million in December on Farside’s tables, after a 2025 that had still netted $9.69 billion of inflows and set the record single day, $727 million on July 16, 2025 (Farside data via KuCoin News; CoinDesk, July 17, 2025). The outflows ran five straight months to March 2026, through the ETHB launch (CoinDesk, June 5, 2026), and an eight-week run of weekly outflows ended only in July (Phemex, July 2026).

The group returned to sustained inflows in August 2026, which Farside’s tables show as the strongest month since August 2025, with about $816 million to $824 million in the week ending August 28 alone and twelve consecutive positive days to September 1 (Farside Investors data as reported by KuCoin News, Crypto Briefing and BigGo Finance, August–September 2026).

Two cautions: the ether funds are about a sixth the size of the bitcoin funds, $15.6 billion of net assets on September 7, 2026 against $101 billion (SoSoValue), so a $200 million day is large here and ordinary there; and part of the inflow since March 2026 is holders rotating from non-staking funds into staking ones, a wrapper change, not new money. ETHB had drawn about $775 million of cumulative net inflow by September 4, against $12.9 billion for ETHA since launch and a cumulative $5.4 billion of outflow from legacy ETHE (SoSoValue).

Outside a fund, the cheapest way to hold and stake ether carries the most operational risk, so the routes below are ranked with both columns visible: the published cost and what can go wrong. Investing in Bitcoin covers exchange fee tiers, hardware wallets, multisig, inheritance and the physical risk of holding keys in full; here the focus is what changes because ether can be staked.

Solo staking

The protocol route: 32 ether (about $79,000 at our September 8, 2026 price), a computer with a good SSD and a home connection, the two client programs, and the discipline to keep them online. Nobody takes a cut, you earn the full reward, and you hold the keys. The costs are a few hundred dollars of hardware, electricity, and slashing: a validator that signs two conflicting blocks, which honest software does not do, loses part of its stake and is ejected, while one that goes offline earns a small negative until it returns. The best economics on the list, and for most readers a second job.

Liquid staking: Lido and Rocket Pool

Deposit any amount of ether with Lido and you receive stETH, a token whose balance grows daily with the staking reward and trades one-for-one with ether in normal conditions; Lido keeps 10% of rewards. Rocket Pool’s rETH works the same way with a 14% cut and a more decentralised operator set. Both are smart contracts, so the risks are the code, the operators, and the market for the token: a liquid staking token is a receipt, and when everyone wants to redeem receipts at once the receipt trades below the thing it represents.

In June 2022, during the Celsius and Three Arrows unwinds, the stETH discount, which had opened at 2% to 3% after Terra’s collapse in May, hit a record 8% on June 13, and Three Arrows sold stETH into it at 5.6% to 5.9% below ether (CoinDesk, June 14, 2022; crypto.news, June 2022); it stayed several points wide for weeks. Protocol withdrawals were impossible until the Shapella upgrade of April 2023; today they run through a queue, which limits how far the discount can go but does not remove it. Lido’s other risk is concentration: one protocol behind 21% to 23% of all staked ether is a tail risk for the network, and the token you hold would be the first thing repriced.

Exchange staking: Coinbase and Kraken

The simplest route: buy ether on Coinbase, press stake, and Coinbase runs the validators and pays the reward less its 25% commission on ether (35% on most other assets; Coinbase Help, 2026); its cbETH token is the same programme in wrapped form, with the commission taken from the rewards before the token accrues them. Kraken offers similar programmes where state rules permit; its bonded-staking commission is tiered by balance, from about 30% down to 10% at the largest accounts, and Kraken’s own worked example puts a $1.5 million balance in the 20% tier (Kraken support pages, 2026), so a $25,000 account should expect the top of that range.

Your ether sits in the exchange’s omnibus wallet, so counterparty risk sits on top of price risk, and a quarter of the yield is gone before it reaches you. The right route for one reader only: the one who would otherwise not stake at all.

Buy on an exchange, hold in a hardware wallet

Coinbase Advanced charges 1.20% for a market order and 0.60% for a limit order at the entry tier (under $1,000 of trailing thirty-day volume), falling to 0.40% and 0.25% between $10,000 and $50,000; Kraken Pro charges 0.80% and 0.40% at entry on the cross-platform tiers it introduced on July 9, 2026. Withdrawing to your own wallet costs a network fee that in 2026 is well under a dollar: Coin Metrics’ May 2026 figures work out to 269 ether of fees across 2.14 million transactions a day, about 0.00013 ether or 27 cents per transaction at that month’s price. A hardware wallet from Ledger or Trezor runs $79 to $169 (published prices, 2026). From the wallet you stake into Lido or Rocket Pool directly, keep the keys, and collect 86% to 90% of the reward. This is investor C in the worked example.

Restaking

Restaking protocols, EigenLayer the largest, let staked ether be pledged again to secure other services for an extra reward. That is leverage on the same collateral, exposed to a second set of slashing rules belonging to services with short track records, for yields paid mostly in the new protocols’ own tokens and points. Nothing in this paragraph belongs in a non-specialist’s first four years.

Every yield you will be offered on ether either has a payer or does not, and “decentralised finance” is both the source of every crypto yield and the reason most of them end badly. The flagship guide tells the Terra, Celsius and Stream Finance stories; the rule here is the same, applied to what an ether holder actually sees.

Staking, covered above, is the protocol paying validators. Payer: the protocol, from issuance and fees. Real, denominated in ether, about 2.6% to 2.8% base in 2026.

Lending is Aave, Compound, Morpho and their peers: you deposit ether or a stablecoin, a borrower posts more collateral than he takes out, and the rate is set by how much of the pool is in use. Lending USDC on Aave paid 4% to 10% through 2024 and 2025; the borrower was usually a trader levering a long, so the rate is highest at the top of the market and near zero at the bottom. Payer: the borrower. Real but cyclical, and the risk is not the borrower, who is liquidated automatically, but the code, the price oracle, and in the newer protocols the “curators” who choose what collateral is acceptable. In November 2025 Stream Finance’s $93 million loss cascaded into Morpho and Euler vaults that had accepted its token as collateral, and depositors who had never heard of Stream ate it. Investing in Stablecoins and On-Chain Yield, on this hub, covers the dollar side of that market, from the reserve reports to the sUSDe basis trade.

Liquidity provision is depositing ether and a stablecoin into an exchange pool such as Uniswap and earning the trading fees. Payer: traders. Real, but the position is short volatility: a sharp move either way leaves you holding more of the asset that fell, and the fees often do not cover it. This is called impermanent loss and it is neither impermanent nor small.

Basis and funding yields come from selling futures against spot and collecting what leveraged longs pay to stay long; Ethena’s USDe packaged it and paid double digits in 2025. Payer: other traders. Real until they leave, which they did in the fourth quarter of 2025, when funding went negative and USDe’s supply fell 70% from its October peak (Coin Metrics).

Points, airdrops and token emissions pay depositors in a protocol’s own future token for showing up. Payer: the protocol, in a currency it prints. Sometimes valuable, usually not, always taxable at receipt, and the basis of every “20% on ether” pitch of 2024 to 2026.

Your own principal is the fifth answer, the one that ends careers; the flagship guide’s Terra retelling is the reference case. If the yield is denominated in the thing being issued to pay it, you are the yield.

Ether’s costs are lower than any physical asset’s in this series, and the one that matters is not a fee at all. Round-trip transaction cost: 0.50% to 2.40% through an exchange depending on order type and tier, effectively zero through an ETF at a spread of a cent or two. Carry: the ETF sponsor fee of 0.12% to 0.25% a year for the mainstream funds (2.50% for legacy ETHE, which nobody should still own), against nothing for self-custody beyond a one-time $79 to $169 device. Network fees for moving or staking ether from your own wallet: under a dollar per transaction at 2026 gas levels, though a busy day can multiply that by a hundred.

Staking commission: from 0% solo to 30% at the top Kraken tier, per the chart in the ETF section, all taken from the reward rather than the principal.

The cost that matters is the one no schedule prints: the opportunity cost of not staking. A holder of plain ETHA, or of ether sitting unstaked on an exchange, forgoes roughly 2.6% to 2.8% a year in ether, about 11% of the position over four years, more than ten times the sponsor fee. The second is tax drag on the yield itself, which the next section prices. Everything else is small enough that it should not decide the route; custody risk and what you will actually do should.

Ether is the one crypto asset most readers will receive income from, and income is taxed differently from gains, so the US treatment is set out precisely. The foundation, which Investing in Bitcoin covers in full, is IRS Notice 2014-21: crypto is property, so every sale or swap is a capital gain or loss, long-term at 0%, 15% or 20% federally after a year (plus the 3.8% net investment income tax above $200,000 single or $250,000 joint), short-term at ordinary rates up to 37% inside a year. Ether is not a collectible under section 408(m), so the 28% rate on art and cars does not apply.

There is no wash-sale rule for crypto held directly as of September 2026 (section 1091 covers securities, and the bills that would extend it had not passed; Investing in Bitcoin has the detail), but there is for the ETFs, which are securities.

Staking rewards

Revenue Ruling 2023-14 (July 2023) holds that staking rewards are ordinary income at their fair market value on the day you gain dominion and control of them, and that value becomes your basis. So a staker is taxed twice on the same ether: as income at receipt, and as capital gain or loss between that basis and the eventual sale. For stETH, whose balance grows daily, that is 365 small income events a year and a crypto tax package to compute them; for Coinbase staking, a Form 1099-MISC once rewards exceed $600 in a year; for a staking ETF, a cash distribution on your brokerage 1099.

The taxpayer challenge to the income treatment, the Jarrett cases, has run since 2021; the first suit was mooted when the IRS refunded the money, and the second, over the 2020 tax year, is set for trial on September 29, 2026 (McDermott Will & Emery; Bloomberg Law). Until a court says otherwise, income at receipt is the rule.

Reporting

Form 1099-DA reached exchanges for the 2025 tax year: gross proceeds only, on forms issued in early 2026, and cost basis from the 2026 tax year for “covered” assets, meaning ether acquired on or after January 1, 2026 and held at the same broker until sold. Ether transferred in from a wallet, or bought earlier, is reported without basis, and the IRS assumes zero unless you show records. Revenue Procedure 2024-28 ended the universal-wallet method from January 1, 2025: basis is tracked per wallet and account, so the hop from Coinbase to your Ledger to Lido needs a record at each step. The DeFi broker rule was repealed under the Congressional Review Act on April 10, 2025, so Lido, Aave and Uniswap report nothing; the obligation is yours. The clean structure is a staking ETF inside a Roth IRA, where the ordinary-income treatment of the reward is irrelevant.

Four investors put $25,000 into ether at our September 8, 2026 price of $2,482 and hold four years. All see the same path: a 50% drawdown to about $1,240 in year two, then a recovery to $3,700 at the end of year four, a 49% gain or 10.5% a year, which is modest by ether’s standards and aggressive by anyone else’s. Gross staking yield is 3.0% a year in ether throughout. Everyone pays 15% federal long-term capital gains plus 5% state, 20% combined, and 32% federal plus 5% state, 37%, on ordinary income. Figures are rounded; the arithmetic is shown so the desk can check it.

Investor A: ETHA, no staking, taxable account

Buys $25,000 of shares at a $5 spread. The 0.25% sponsor fee comes out of the ether over four years, about $238 on an average balance near $23,750. End value $25,000 × 1.49 = $37,273, less fees and spreads, about $37,026. Gain $12,026, tax $2,405, net $34,621, a 38.5% after-tax return. In a Roth IRA she keeps the full $37,026.

Investor B: Coinbase Advanced, then Coinbase staking

Places a limit order at 0.60%, paying $150, and holds 10.0135 ether. Stakes on Coinbase, which keeps 25% of the 3% reward, so her ether compounds at 2.25% a year: 10.9455 ether after four years, 0.932 ether of rewards. Those rewards, valued on receipt along the price path, are about $2,209 of ordinary income, taxed at 37%, $817. At $3,700 the position is $40,499; she sells at 0.60%, $243, for proceeds of $40,256. Gain $40,256 − $24,850 − $2,209 (the rewards’ basis) = $13,197; tax $2,639. Net $40,256 − $817 − $2,639 = $36,799, a 47.2% after-tax return, with her ether in Coinbase’s custody throughout.

Investor C: exchange to hardware wallet to Lido

Same limit order, $150; withdraws to a Trezor for about $5 of network fee; stakes into Lido, which keeps 10%, so her stETH grows at 2.7%: 1.126 ether of rewards, about $2,672 of ordinary income across 1,460 daily accruals, tax $989. End value $41,208; sends back and sells at 0.60%, $10 of network fees, proceeds $40,951. Gain $13,434, tax $2,687. Less $150 for the device and steel backup: net $37,125, a 48.5% after-tax return, with the keys, the seed phrase, the transfer log and the four-year spreadsheet as her responsibilities.

Investor D: ETHB, staked ETF, taxable account

Buys $25,000 of shares. Assume the sponsor fee reverts to 0.25% after the first-year waiver, $238 over four years, and that the fund stakes 85% of assets and passes 82% of the 3% reward through as monthly cash: about 2.1% a year on the starting 10.07 ether, not compounded because it is paid out, valued at each year-end price on the same convention as B and C. Distributions total about $1,981, taxed at 37%, $733. End value of the shares $37,026, gain $12,026, tax $2,405. Net $37,026 + $1,981 − $733 − $2,405 = $35,869, a 43.5% after-tax return. In a Roth the same investor keeps $39,007, the best figure on the page.

Three things the chart says. The staking reward is worth $2,000 to $2,700 on $25,000 over four years, the largest line after the price itself; a holder of a non-staking fund is paying roughly 2% of the position a year for a brokerage ticker, which the arithmetic no longer justifies now that staking funds exist.

Self-custody wins in a taxable account by $326 over Coinbase and $1,256 over the staked ETF, the price of the sponsor’s 18% cut, the un-staked buffer and the loss of compounding; whether $1,250 over four years is worth holding your own keys is a question about you, not the fund. And the tax on the reward, $730 to $990 in every taxable case, is why the Roth version of the staked fund is the best number on the page: the ordinary-income treatment is the one cost a wrapper removes entirely.

For a reader starting from nothing, the flagship guide’s rules on sizing (1% to 5% of investable assets, a number you can lose entirely) and on bitcoin first apply unchanged; ether is a second holding, and these steps assume you have read the fee section and still want it.

1. Decide the split before the price. Inside a crypto allocation, ether above a third is a view that fees return to the base chain; below a fifth it is a diversifier that has diversified nothing since 2022. Pick a share, write it down, rebalance to it quarterly.
2. Use the staking fund in the retirement account. A staking ETF in a Roth IRA is the highest after-tax number in the worked example and the least work. Read the prospectus for three terms: stakes or not, what share, what cut. Prefer an 80%-plus pass-through; reject anything above 0.30% in fees.
3. In a taxable account, choose between two routes and no others. Either the staking ETF, accepting the sponsor’s cut for the brokerage statement and the wash-sale rule; or an exchange purchase by limit order, withdrawn to a hardware wallet bought direct from the manufacturer, and staked into Lido or Rocket Pool from there. Not exchange staking at a 25% cut; not an unstaked balance anywhere.
4. Buy on a schedule. Twelve equal monthly purchases. On our tape the twelve-month schedule from the October 2025 top finished up 7.8% where the lump sum finished down 45%.
5. Set up the tax file on day one. A crypto tax package connected to the exchange and the wallet, a transfer log for every hop, and, for stETH, an acceptance that the rewards are income daily.
6. No restaking, no points, no leverage, no yield above the staking rate in the first four years. Every ether-denominated disaster since 2016 began with a holder deciding that holding was not enough.
7. Write the exit rule now. The thresholds in the next section are ours; adopt them or write your own, but decide before the next 60% move what would make you sell, and what would make you add.

The case for ether is a set of numbers, each of them public; these are the readings that would change our view, with the levels.

The fee base. Coin Metrics’ FeeTotNtv, the daily ether paid or burned on the main chain, against daily issuance of about 2,800. Above 2,800 for a quarter: supply is shrinking again and the 2021 case reopens, upgrade the view. Below 200 for a quarter (December 2025 and January 2026 printed 123 and 152; February was back at 271): the layer 2s are winning outright, downgrade it.

The ETH/BTC ratio on our tape. 0.0316 on September 8, 2026; low 0.0258 on June 7. A close above 0.040, the September 2025 level, would be the first sign in four years that ether is winning the relative trade; a close below 0.0258 would mean the June low was not the low. The Coin Metrics daily series bottomed at 0.0181 on April 21, 2025, the lowest since January 2020, and that is the level at which we would treat the relative trade as broken for the cycle.

The staking rate and the queue. 35% of supply staked in September 2026. Above 40%, the base yield falls toward 2.4% and the ETF pass-throughs shrink with it; a sustained exit queue longer than the entry queue, which has not happened in 2026, would mean stakers are leaving and the structural bid is gone.

ETF terms and flows. Fidelity’s FETH staking amendment going effective, then the pass-through rates for a race to 90%. Monthly net flows above $1 billion for two consecutive months would match the bitcoin funds’ August 2026 turn signal, scaled to the ether funds’ size.

The treasury companies. BitMine’s weekly holdings release and its mNAV. A quarter of net sales, as Strategy showed for bitcoin in mid-2026, would put 5.9 million ether, 4.9% of supply, on the sell side of a market that absorbs a few hundred million dollars a day through the funds.

The roadmap. Glamsterdam on or near November 4, 2026; a slip of a quarter is normal. The signal would be any proposal that reprices blob space upward or routes layer-2 fees to the base chain, because that is the value-accrual fix and as of September 2026 none is scheduled.

Our own tape. Ether at $2,482 needs to clear $2,968, the December 31, 2025 close, before we would call the 2026 leg over; the crypto sub-index at 59.05 needs its mid-January 2026 level of 80.