Bitwise Staking Report: Revenue of Six Major Public Chains Generally Declines, What Supports the Value of Tokens?

Author: Bitwise "Staking Report Q3 2026"

Compiled by: Jiahua, ChainCatcher

In the second quarter of 2026, market sentiment and network fundamentals diverged in two directions. In the networks covered in this article, block space became cheaper, on-chain activity increased, and fee income decreased.

The decline in income was mainly caused by protocol design, as various chains actively made block space more abundant and cheaper. Some networks did experience weakening demand, but that was not the general case.

The divergence between income and activity was the main theme of this quarter. Fees are falling while usage is rising.

After increasing the gas limit, Ethereum processed 26 transactions per second, up from 15 transactions a year ago. Solana processed about 10 billion non-voting transactions, higher than any quarter in 2025. Avalanche's C chain processed 236 million transactions, up from 58 million a year ago. The decrease in block space costs does not indicate weaker demand.

Institutional adoption became quantifiable for the first time this quarter. In terms of staking, Ethereum's active staking volume reached a historical high of 40.2 million ETH, accounting for 33% of the supply, while Coinbase and Circle each staked 500,000 HYPE under Hyperliquid's AQAv2 framework.

In terms of tokenized real assets, the trading volume of tokenized stocks on Solana increased by 2,479 times year-on-year, rising from $1.34 million in June 2025 to $3.32 billion a year later, driven by the launch of SpaceX and Micron on Backpack Securities. The S&P Dow Jones Indices authorized the S&P 500 for products on Hyperliquid, and FIFA's World Cup ticketing system runs on a dedicated Avalanche L1.

In payments, Deel issued about $30 million to 7,200 contractors in its first full quarter on Tempo.

Ethereum

Staking Rewards and Validator Economics

Ethereum's staking rewards consist of two parts. Consensus rewards are new ETH issued by the protocol to validators and stakers, which is pure issuance and will slowly push up the total supply. Execution rewards come from user-paid priority fees and MEV, reflecting the real demand for block space.

The income structure is primarily based on issuance. In Q2 2026, 2.84% of total income came from consensus rewards (2.65 percentage points) and execution rewards (0.19 percentage points), with a ratio of about 93% to 7%.

Figure Caption: Annualized staking yield: 25Q1 3.35%, 25Q2 3.18%, 25Q3 2.94%, 25Q4 2.94%, 26Q1 2.91%, 26Q2 2.84%

Network Income

Ethereum's network income comes from four fee streams. Two of these are burned: base fees (the block space fees priced by EIP-1559, paid for each transaction) and blob fees (paid when L2 sends transaction data back to the mainnet for data availability).

The other two go to validators and stakers: priority fees (tips users pay to accelerate transactions) and MEV (the additional value captured by validators from transaction ordering, after deducting the share for block builders).

In Q2 2026, the entire chain collected about 31,166 ETH, of which 74% went to validators and stakers, and 26% was burned. Converted at the quarterly average price, this amounts to about $64 million, down from $131 million in Q2 2025, a year-on-year decline of 51%. After the upgrades of Pectra and Fusaka made blob space abundant, blob fees have approached zero.

Figure Caption: Quarterly network income in ETH: 25Q1 97,998, 25Q2 59,593, 25Q3 37,755, 25Q4 30,995, 26Q1 27,670, 26Q2 31,166

In ETH terms, income rebounded from 26Q1's 27,670 to 26Q2's 31,166, marking the first quarter-on-quarter increase in over a year. The decline in dollar figures was solely due to the drop in ETH prices during this quarter.

Active Staking and Capital Flow

Active staking steadily increased from 34.2 million ETH at the end of 25Q1 to 40.2 million ETH at the end of 26Q2, reaching a historical high and accounting for 33% of the total supply.

The inflow was concentrated in the first half of 2026. In Q1, an additional 2.81 million ETH was added, the largest single-quarter increase during this period, followed by another 1.73 million ETH in Q2. The more ETH staked, the slower and lower the consensus yield decreases.

Figure Caption: Total staking volume and its proportion of supply: 25Q1 34.2 million (28.4%), 25Q2 35.4 million (29.3%), 25Q3 35.7 million (29.5%), 25Q4 35.8 million (29.4%), 26Q1 38.5 million (31.7%), 26Q2 40.2 million (33.0%)

These inflows came from institutions. Bitmine Immersion held 5.62 million ETH as of June 14, 2026, of which 4.72 million had been staked. BlackRock's iShares Staked Ethereum Trust ETF (ticker ETHB) was listed on Nasdaq on March 12, 2026, with 70% to 95% of its staked holdings, distributing rewards to investors monthly.

Block Capacity and Usage

The mainnet's throughput nearly doubled within a year, increasing from 15 transactions per second to 26, as the block gas limit was raised from 30 million to 60 million. This increase was phased in by validators in 2025: to 36 million in February, 45 million in July, and 60 million by the end of November.

The gas limit is the computational budget available for each block. After raising it, each block can accommodate more transactions, thus increasing throughput without making blocks more congested.

Figure Caption: Average TPS of the mainnet: 25Q1 14, 25Q2 15, 25Q3 19, 25Q4 18, 26Q1 and 26Q2 both 26

The block fill rate has long been maintained at around half, which is by design. EIP-1559 set the target fill rate at 50% and automatically adjusts the base fee to maintain this level. If a block exceeds half full, the base fee for the next block can be raised by up to 12.5%. If it still exceeds, it continues to be raised until usage falls back. The reverse is also true.

The reason for targeting half-full rather than full is to leave a buffer for demand peaks. When activity surges, blocks can expand to the limit, while rising fees gradually push out lower-priority transactions.

Ranking contracts by gas consumption in Q2 2026, the largest single consumer was Tether (USDT), accounting for 6.27% of all mainnet gas, with the top ten contracts collectively accounting for nearly a quarter. The heaviest single use was ordinary stablecoin transfers, with USDT and USDC consuming about 9% of the gas.

Two results are noteworthy. The second-largest consumer, accounting for 3.39%, is a contract marked by Etherscan as address poisoning, which sends small fake transfers to numerous wallets, waiting for users to mistakenly copy the scammer's address. Aztec privacy rollup consumed 1.75% of the gas for verifying zero-knowledge proofs on L1, while ERC-4337 EntryPoint accounted for 1.54%. Named MEV bots are present, but none exceed 1.5% in scale.

Roadmap

Two upgrades shaped the current situation. Pectra (May 2025) introduced smart account functionality (EIP-7702) for regular wallets and raised the maximum balance for validators from 32 ETH to 2048 ETH (EIP-7251), allowing operators to combine staking and run the same amount with fewer validators.

Fusaka (December 2025) introduced PeerDAS, allowing nodes to sample rollup's blob data without needing to download and store it all. This expanded blob capacity, and the cost for L2 to send data to Ethereum is expected to decrease.

Next are two hard forks. Glamsterdam is targeted for the second half of 2026, with a specific date yet to be determined, mainly doing two things: incorporating block building into the protocol's own ePBS, eliminating the reliance on trusted relays that has existed since the merge; and introducing block-level access lists to inform the network in advance which states a transaction will touch, thus supporting parallel execution. This upgrade is also expected to further increase the gas limit, with long-term discussions reaching up to 200 million.

The second hard fork is Hegota, planned for the end of 2026, introducing FOCIL (EIP-7805), a change to the consensus layer aimed at anti-censorship. In each slot, a committee of 16 validators lists the valid transactions they see in the mempool, and block builders must include these transactions, with validators only voting for blocks if builders comply. This decentralizes the inclusion rights of transactions to many participants, making it structurally difficult for a single party to exclude a transaction.

Finally, regarding block time, a block is normally produced every 12 seconds, with final confirmation taking 12 to 15 minutes. EIP-7782 proposes to cut block time in half to 6 seconds by compressing three sub-tasks within each slot: block proposal from 4 seconds to 3 seconds, proof from 4 seconds to 1.5 seconds, and aggregation from 4 seconds to 1.5 seconds. This proposal is currently not attached to any scheduled forks; it was considered for Glamsterdam but did not make the cut.

Solana

Staking Reward Composition

Solana's staking rewards remain among the highest in mainstream public chains. Validator income has three sources: protocol inflation, block rewards, and MEV tips. In the first half of 2026, all three declined, with inflation decreasing according to Solana's reduction schedule, while block rewards and tip income weakened due to lower network activity.

Figure Caption: Network comprehensive annualized yield: 25Q1 11.50%, 25Q2 9.11%, 25Q3 8.17%, 25Q4 6.94%, 26Q1 6.89%, 26Q2 6.25%

Inflation rewards fell from 6.1% to 5.8% in the first half of the year, still the main source of income, accounting for over 90% of annualized staking returns. Block rewards and Jito tips contributed a total of 0.44% in Q2 2026, down from 0.79% in the previous quarter.

A notable change from 2025 into 2026 is the reversal of the internal structure of non-inflation rewards. In 25Q1, during the trading boom, Jito MEV tips accounted for about 70% of non-inflation income. By the first half of 2026, block rewards became the main source, accounting for about 75%.

Two factors contributed to this. First, reduced trading activity led to fewer arbitrage opportunities, decreasing the demand for competing for transaction ordering through Jito bundles. Second, new block-building clients like Jito's BAM and Harmonic have been increasingly adopted, with Harmonic using its own priority fee-based bundling scheme, thus cutting into the revenue share flowing through the traditional Jito ecosystem.

These trends accelerated after the market crash in October 2025 and were amplified by competition from new trading platforms like Hyperliquid.

REV and On-Chain Activity

Solana's Real Economic Value (REV, the total amount users pay to use this chain) has continuously declined during this period, dropping from a peak of $812 million in 25Q1 to $89 million a year later, and further down to $51 million in Q2 2026.

Figure Caption: REV in USD: 25Q1 $812 million, 25Q2 $272 million, 25Q3 $223 million, 25Q4 $90 million, 26Q1 $89 million, 26Q2 $51 million

In SOL terms, the decline is not as severe as the dollar figures suggest. In 25Q1, REV generated about 3.9 million SOL, while in 26Q1 it was 877,000 SOL and in 26Q2 it was 631,000 SOL. Compared to the activity peak in early 2025, it has significantly shrunk, but the decline is much milder than in dollar terms, as SOL prices fell by 62% during this period. There was positive momentum in 26Q1, indicating that demand had rebounded from the low point at the end of 2025, but it has not fully recovered.

Activity itself has shown resilience. In 26Q1, 10.1 billion non-voting transactions were processed, an 18% year-on-year increase, while in 26Q2, 9.8 billion transactions were processed, flat compared to the same period last year.

So the question arises: if transaction volume is not weak, why is REV declining?

Transactions on Solana consume computational resources, measured in computational units (CUs). Each block has a fixed CU capacity. This capacity was increased from 48 million to 60 million CUs in two updates in April and July 2025, and is expected to reach 100 million in 2026.

The average CU utilization for non-voting transactions dropped from 55.5% in 26Q1 to 47.5% in 26Q2. With reduced competition, the demand for priority fees and Jito tips also declined, leading to a 31% quarter-on-quarter drop in average rewards per block, from 0.041 SOL in 26Q1 to 0.0287 SOL in 26Q2.

The conclusion is straightforward: Solana's income decline is due to decreased congestion. Block space is no longer scarce, and users have little reason to bid aggressively for packaging, leading to a decline in REV.

Regarding slot time, Solana's target is about 400 milliseconds, with actual duration fluctuating based on validator behavior. Different schedulers have different trade-offs; some may slightly extend slots to capture more transactions and increase block income. After peaking in January 2025, slot times have been declining, reaching about 385 milliseconds in Q4 2025 and returning to nearly 400 milliseconds in Q2 2026. A reasonable explanation is that validators have economic incentives to extend slots to build higher-value blocks, at the cost of slowing down the network.

Staking Structure and Validators

Solana's staking rate is among the highest in public chains, with about 427 million SOL staked, accounting for 68% of the total supply of 628 million. This ratio has remained unusually stable over the past two years, supported by relatively high staking rewards.

Of the staked SOL, about 57 million (13.3%) is delegated through liquid staking tokens (LSTs), with the Solana Foundation Delegation Program (SFDP) accounting for 20 million (4.7%), and the remaining 350 million directly staked by market participants. SFDP was originally intended to distribute staking to smaller validators to support decentralization, but the foundation has increasingly only supported validators that genuinely contribute to the ecosystem, with the program's share of total staking dropping from about 10% to less than 5% over the year.

The total amount of liquid staking rose from 40 million SOL in January 2025 to a peak of 66 million a year later, before falling back to 56.5 million by June 2026.

JitoSOL saw the steepest decline, dropping from a peak of 18.5 million SOL in June 2025 to 9.94 million a year later. Marinade followed a similar trend, falling from 5.5 million to 2.4 million. Binance emerged as a clear winner, growing from 6.5 million at the beginning of 2025 to 10.1 million. DoubleZero briefly reached 13 million between November 2025 and February 2026 but has since fallen back to 5.2 million.

The key point is that the growth of liquid staking is no longer dominated by old players; JitoSOL and Marinade have lost market share, while BNSOL, dzSOL, dSOL, and a number of smaller LSTs have captured most of the incremental growth.

The number of active validators decreased from about 1,250 in June 2025 to about 830 a year later, a 34% reduction over 12 months. The reason is that light voting costs about 1 SOL per day, which, at recent prices, annualizes to about $25,000, not counting hardware and infrastructure expenses. Validators with insufficient delegation are increasingly finding it difficult to be profitable, and with the gradual contraction of SFDP staking, small operators are exiting the market.

Application Layer Revenue

Solana's application layer generated $227 million in revenue in Q2 2026, down 31% from $328 million in Q1 2026. This decline can largely be attributed to the SOL price, which fell similarly during the same period.

Transactions remain Solana's dominant use case, with seven of the top ten applications related to trading. Pump continues to lead by a wide margin, generating $90.1 million in Q2 2026, about three times that of the second-place Collector Crypt ($32.2 million), with the two combined accounting for over half of the total revenue of the top ten.

Collector Crypt is a collectibles platform focused on tokenized real assets, featuring cards similar to Pokémon. It gained traction after launching a gacha mechanic, where random card packs contain tokenized collectibles that users can hold, trade, or redeem for physical delivery. It is the fastest-growing among the top ten.

Pacifica is a newer entrant, generating $20 million in Q2 2026, slightly above $18.9 million in Q1 2026. It is a hybrid perpetual exchange, with the matching engine running off-chain for low-latency execution, while custody and settlement are on-chain.

The list misses a growing category of proprietary automated market makers (propAMMs). Unlike traditional AMMs, propAMMs typically do not charge explicit fees but earn through the bid-ask spread like traditional market makers, making profitability difficult to measure. Given their increasing importance in liquid markets, the true revenue of Solana's trading ecosystem should be higher than what the list indicates.

Tokenized Stocks

On June 12, the largest IPO in stock market history occurred. SpaceX went public with a valuation of $1.77 trillion, attracting significant attention from both institutions and retail investors, and Solana was well-prepared.

In 2025, the trading volume of tokenized assets was primarily driven by credit. In 2026, stock trading volume rose from $226 million in January to $871 million in May, and then skyrocketed to $3.3 billion in June due to SpaceX's IPO.

On Solana, tokenized SpaceX stock $SPCX can be purchased through three channels: Backpack, xStocks, and PreStocks. Since the IPO, the three have collectively traded $600 million, with 85% coming from Backpack.

Memory chip manufacturer Micron ($MU) began trading on Solana in early May, but only saw significant trading volume after launching on Backpack Securities on June 22. By the end of June, the stock had a total quarterly trading volume of $107.9 million, with $106.7 million occurring on Backpack.

Backpack dominates the tokenized stock space, likely because it allows users to redeem tokenized stocks on Solana through its exchange, Backpack Securities, for the underlying assets.

Block Building: BAM and Harmonic

The competition for block building began in the second half of 2025, with Jito's BAM launching in September and Harmonic in November, each envisioning a different future for Solana.

Traditionally, validators are responsible for both proposing blocks and deciding the trade-offs and ordering of transactions. As MEV value increased, concerns about centralization and market fairness arose. The new block building systems separate transaction ordering from block proposals. BAM emphasizes transaction privacy and predictable execution, while Harmonic focuses on competition among builders.

Both systems saw substantial adoption shortly after launch. Validators running BAM correspond to about 140.8 million staked SOL (33.0% of total staking), while Harmonic validators correspond to 99.4 million (23.3%), totaling about 56%, meaning that most blocks on Solana are now produced by these two systems. As of the end of June 2026, both ecosystems remain closed, with block building infrastructure privately operated by their respective teams.

BAM's architecture is built around a Trusted Execution Environment (TEE). TEE allows participants to verify that a node is indeed executing approved code without being altered to the extent that it harms transaction privacy or changes execution results, thus providing a more trustworthy environment for market participants submitting order flows.

In addition to privacy, BAM also brings a standardized building environment. Today, Solana's market makers face a fragmented ecosystem, with validators running different clients like Agave, Frankendancer, BAM, and Harmonic, often using different schedulers within the same client. Schedulers determine how transactions are ordered within blocks, resulting in inconsistent transaction landing behavior across different validators, forcing market makers to adjust strategies for different execution environments. BAM deploys a single scheduler across all BAM nodes to solve this issue.

BAM also connects to many of the fastest validators on the network. In the median slot time distribution from epochs 971 to 980, BAM validators are tightly clustered in the lower range, while non-BAM validators drag a long slow tail. BAM validators dominate in the 350 to 360 milliseconds range, with almost no presence in the slowest segment.

Harmonic's vision is closer to Ethereum's architecture, where independent builders compete to construct the highest-value blocks. To support this model, it offers several strategies aimed at different goals: first-come-first-served (FCFS) for low-latency execution, MREV for maximizing validator income, 50-millisecond frequent batch auctions for reordering transactions within short auction windows, and customizable batch timings, ordering rules, and bundle processing methods.

Harmonic also focuses on reducing toxic MEV, with its validators' sandwich attack rates consistently among the lowest in the network.

Figure Caption: As of June 15, 2026, the network's 30-day sandwich attack rate is 0.79%, 60-day 0.81%; Harmonic validators corresponding to 99.4 million staked SOL have 30-day rates of 0.69% and 60-day rates of 0.67%

Traditional finance relies on regulation and market monitoring to limit predatory trading; public chains can only address the same issues through protocol design and infrastructure.

Alpenglow and 200 Millisecond Slots

The most radical protocol upgrade currently under development is Alpenglow, a new consensus protocol intended to replace Tower BFT, planned for deployment in Q3 2026, marking the most significant change to Solana's core architecture since its launch.

The main benefit is faster finality. Solana's block time is already among the fastest in the industry, but achieving finality still takes several seconds; Alpenglow aims to reduce this to a few hundred milliseconds.

To achieve this, Alpenglow introduces a two-round voting consensus mechanism, replacing historical proof (PoH) with validators' local clocks, and shifting from the current gossip message propagation to direct communication between validators. These changes significantly simplify the consensus layer and reduce communication overhead.

Alpenglow also eliminates on-chain voting. With consensus messages moved off-chain, voting transactions disappear, lowering validator operating costs while freeing up block space for non-voting activities.

If finality time is shortened without a corresponding reduction in slot time, it would seem incomplete. To address this, the target slot time is expected to be reduced from about 400 milliseconds to 200 milliseconds with SIMD-0525, which is anticipated to be deployed before Alpenglow goes live.

Halving the slot time means doubling the block production frequency and reducing latency, while shortening the leader window from 1.6 seconds to 800 milliseconds, leaving validators less time to delay block production in search of more profitable transactions. For traders and market makers, this means faster execution and more predictable outcomes; for validators, it increases operational pressure, as the volume of voting transactions under the existing consensus model will roughly double before Alpenglow is fully implemented, raising costs accordingly.

Inflation and Burn

SOL inflation has long been a contentious topic in the community. When Solana launched in 2020, it adopted an 8% inflation rate to stimulate validator participation and ecosystem growth, decreasing by 15% annually according to the current schedule, reaching a terminal value of 1.5% in 2031.

As the network matures, more participants believe that the inflation rate has fulfilled its historical mission. Proponents of reducing inflation argue that the current 3.8% inflation rate makes staking relatively more attractive, thereby suppressing DeFi participation, while stakers sell rewards to pay taxes and cover operational expenses, creating ongoing selling pressure.

One reference point: when SOL first reached $250 in Q4 2021, its market cap was about $75 billion; when it returned to the same price in Q3 2025, its market cap had risen by 80% to $135 billion.

The first large-scale attempt was SIMD-0228 proposed by Multicoin in early 2025, which sought to replace the fixed reduction schedule with a market-based mechanism linked to staking participation rates. This proposal sparked one of the largest governance debates in Solana's history but ultimately failed to secure the required absolute majority.

Subsequent discussions shifted to more targeted proposals. Helius proposed SIMD-0550, raising the reduction rate from 15% to 30%, maintaining the issuance model, but the network would reach a terminal inflation value of 1.5% in 2028 instead of 2031, cumulatively issuing nearly 19 million fewer SOL.

Temporal proposed SIMD-0547 from another direction: not reducing issuance but increasing destruction. Today, users can reserve a large amount of computational resources without paying for the portion that is not actually consumed. According to this proposal, transactions will pay a small fee based on the requested resources, regardless of actual consumption, with all revenue being burned.

Figure Caption: Estimated daily burn amount under SIMD-0547: 0.1 lamport approximately 2,851 SOL (4.3% of issuance), 0.25 lamport approximately 7,128 SOL (10.7%), 1.0 lamport approximately 28,512 SOL (42.9%)

Based on the suggested starting parameter of 0.25 lamport per computational unit, this mechanism would burn about 7,128 SOL daily, equivalent to about 11% of the current daily issuance. While insufficient to offset inflation, it would significantly increase the burn rate of SOL.

Hyperliquid

Value Capture Mechanism

In Q2 2026, Hyperliquid had seven revenue streams, all of which ultimately lead to the exit of HYPE from circulation, either through the aid fund for buybacks or direct destruction. The aid fund is a system address built into the protocol, with no private key or owner; fees are automatically converted to HYPE and sent in, permanently unrecoverable. Following a governance vote in December 2025, all HYPE held in the fund was officially deemed burned and removed from circulating supply.

The seven revenue streams are:

Native listing market perpetual trading fees. Each perpetual trade is charged based on the quoted asset (USDC, USDT, USDe, etc.), which is the largest revenue stream, with fee flows going into the aid fund for buying and destroying HYPE.

HIP-3 perpetual fees (net). Any team staking 500,000 HYPE can deploy a perpetual market, with underlying assets being either crypto or real assets like commodities, stocks, or indices. Trading fees for HIP-3 markets are distributed between Hyperliquid and the deploying party, with a default set at twice the standard rate and split 50-50; the deploying party can take less but cannot exceed half. This line is currently significantly compressed, as most HIP-3 markets are in "growth mode," reducing total fees (including both parties' shares) by about 90% to help new markets attract liquidity.

Spot trading fees. Portions charged in USDC go into the aid fund for buybacks, while portions charged in HYPE are directly destroyed.

HyperEVM gas fees, which users pay when trading on HyperEVM, with both base fees and priority tips being destroyed.

Token listing and deployment auctions. To list a new spot token or deploy a new HIP-3 market, teams must win a Dutch auction, with the winning amount paid in HYPE and destroyed.

Order priority fees. Traders can pay to have their orders prioritized in the execution queue, paid in HYPE and directly destroyed. This line is zero until 2026.

Gossip priority fees. Participants bid in a cyclical auction to obtain market data faster than standard data sources, also paid in HYPE and destroyed, which is also zero until 2026.

Revenue Composition

Fees flowing to buybacks rose from $22 million in 25Q1 to a peak of $332 million in 25Q3, then fell back to $186 million in 26Q1 and $175 million in 26Q2. Native listing market perpetual fees accounted for 83% of total revenue in Q2 2026, with the remaining 17% spread across HIP-3 perpetual, spot, HyperEVM gas, priority fees, and auctions.

More notably, diversification is occurring. HIP-3 launched on October 13, 2025, with negligible fees in its first quarter, but by 2026 it had become a significant second revenue line, generating $18 million in Q1 2026 and $14 million in Q2 2026, accounting for about 8% of total revenue. As core crypto perpetual fees declined, the growth in fees from non-crypto markets like commodities, stocks, and indices partially offset the decline.

Two revenue lines have yet to be reflected in the Q2 2026 figures. The HIP-4 outcome market has launched and is growing, with trading volume rising from $149 million in May 2026 to $287 million in June, but trading fees remain off during the testing phase. AQAv2 is a framework where stablecoin issuers participating share about 90% of their reserve income with the protocol, contributing to the aid fund for HYPE buybacks. This framework has been approved by validator governance and is applicable to USDC on the Coinbase and Circle sides, but as of Q2 2026, the first payment has not yet begun.

Staking and Delegation

438 million HYPE is staked, accounting for 43.8% of the total supply. Staking is done through delegation, where holders assign HYPE to validators, who use it to participate in HyperBFT consensus and earn staking rewards. Rewards are automatically reinvested back into the staking balance daily, currently at an annualized rate of 2.2%. This portion of income does not come from trading fees or protocol revenue but is paid by a dedicated emission reserve that issues HYPE.

The validator set remains concentrated but is becoming decentralized. Initially, there were only five foundation validators, which gradually expanded to 16, 21, and 24, reaching 27 by the end of Q2 2026. The five validators operated by the Hyper Foundation currently hold about 49% of all staked HYPE, down from over 60% in Q2 2025.

Figure Caption: Net inflow of delegation: 25Q2 +2.87 million, 25Q3 +720,000, 25Q4 +426,000, 26Q1 +59,000, 26Q2 +59,000

This quarter's inflow primarily came from institutions. Coinbase and Circle each staked 500,000 HYPE as conditions of the AQAv2 framework. Hyperliquid Strategies, a corporate treasury that converted its balance sheet to HYPE, is one of the largest holders outside the foundation. Bitwise launched BHYP on the NYSE on May 15, 2026, staking through its infrastructure division Bitwise Onchain Solutions; 21Shares launched THYP on Nasdaq on May 12, and Grayscale launched HYPG on June 3, both also staking on-chain.

HyperCore and HyperEVM

Hyperliquid has two execution environments, HyperCore and HyperEVM, running on the same chain and sharing the same set of validators. Both share a single state, allowing applications on HyperEVM to directly read prices and place orders on HyperCore without needing cross-chain bridges.

Two mechanisms support this. Read precompiles allow HyperEVM contracts to access real-time data from HyperCore, including oracle prices, positions, and balances. The CoreWriter system contract allows HyperEVM contracts to send actions back to HyperCore, including placing and canceling orders, transfers, and staking.

HyperCore is a native order book exchange, with block times under 100 milliseconds and an average throughput of about 24,000 transactions per second, theoretically capable of handling up to 200,000 transactions per second, recording billions of transactions and operations daily.

Figure Caption: Average TPS of HyperCore: 25Q2 5,941, 25Q3 18,161, 25Q4 25,735, 26Q1 22,707, 26Q2 23,724

HyperEVM is a general-purpose permissionless smart contract environment, secured by the same consensus as HyperCore. This chain processes about four transactions per second, with a block fill rate of about 20%. In terms of unique users, the most used transactions are stablecoin transfers, primarily using the native Circle USDC, followed by a few decentralized exchanges and aggregators.

Perpetual Volume and Non-Crypto Assets

Hyperliquid's perpetual trading volume in Q2 2026 was $652 billion, roughly flat compared to $633 billion in Q1 2026, remaining among the highest in on-chain venues. More interestingly, the structural change is notable. By Q4 2025, the first full quarter after HIP-3 launched, non-crypto markets accounted for 2% of total trading volume. By Q2 2026, this portion reached $212 billion, accounting for 32%.

Trade.xyz is the leading team in the HIP-3 market, accounting for 95.1% of HIP-3 trading volume in Q2 2026 ($202.6 billion). It is the first deployer under HIP-3, building products around non-crypto perpetuals, starting with US stocks and expanding to commodities and indices. On March 18, 2026, S&P Dow Jones Indices authorized the S&P 500 for Trade.xyz, making it the first and only officially authorized perpetual derivative based on that index on any decentralized platform. Qualified non-U.S. investors can gain leveraged exposure to the S&P 500, available for trading 24/7, with index data provided by S&P Dow Jones. This represents a substantial difference from previous synthetic stock products in DeFi that relied on unofficial price sources.

Open interest is also worth noting. Open interest only grows when capital enters the market and stays, making it a better reflection of real market activity than trading volume. The average open interest in Q2 2026 was $8.7 billion, up 32% from $6.6 billion in Q1 2026, indicating that traders held positions longer in the second quarter.

Three deployers exited during the quarter. Felix was the first to announce on June 8, followed by Ventuals, which provided perpetual contracts for unlisted companies like OpenAI and Anthropic, announcing market closure on June 15. A few days later, Dreamcash, supported by Tether and operating USDT-collateralized perpetuals, exited, citing that Hyperliquid's native integration of USDC made its USDT-based model uncompetitive.

On the front end, builder code shows which applications are sending order flows to exchanges. These fees are layered on top of Hyperliquid's own trading fees and belong to the applications rather than the protocol.

Figure Caption: Q2 2026 builder code fees: Phantom $4.13 million, MetaMask $2.25 million, Based $760,000, Rabby $690,000, Tread.fi $610,000

General wallets have surpassed specialized trading terminals, indicating that Hyperliquid's user base is expanding beyond its initial scope.

Outlook

Hyperliquid's value capture primarily relies on trading fees, which fluctuate with trading volume, and trading volume itself is cyclical. Two changes are altering the composition of this revenue.

The first is the widening of trading targets. In the three quarters since HIP-3 launched, the non-crypto market has reached 32% of trading volume. HIP-4 extends the model further to outcome and prediction markets, with a cumulative nominal trading volume of about $436 million as of Q2 2026, including about $149 million in the first 30 days, but trading fees remain off during the testing phase, temporarily not contributing to protocol revenue.

The second is AQAv2, a revenue line different from trading fees. It directs about 90% of the reserve income from USDC on Hyperliquid to the aid fund, with the first payment set for October 3, 2026. The platform has about $5 billion to $6 billion in USDC, with Coinbase serving as the capital deployment party, expected to contribute $135 million to $160 million annually. This figure depends on short-term interest rates and whether USDC remains on the platform, but it does not fluctuate with the crypto market cycle.

Avalanche

Staking and Yield

By the end of Q2 2026, Avalanche had 196 million AVAX staked across 591 validators, with a staking rate of 41.5%. The staking rewards for this chain are entirely provided by inflation, unrelated to transaction fees. The protocol issues new AVAX according to a fixed schedule, with a permanent cap of 720 million AVAX, while all transaction fees are burned, partially offsetting inflation.

The reward rate depends on two variables: the difference between the current supply and the 720 million cap, and the staking duration. The longer the lock-up, the higher the fee rate, with the maximum consumption rate for a one-year commitment being 12%.

Reward Rate = (720 million - Current Supply) ÷ Current Supply × Consumption Rate

As of June 2026, the circulating supply was about 472 million, leading to: (720 million - 472 million) ÷ 472 million × 12% ≈ 6.3%.

Figure Caption: Annualized yield for maximum one-year staking: 25Q1 7.0%, 25Q2 6.8%, 25Q3 6.7%, 25Q4 6.5%, 26Q1 6.4%, 26Q2 6.3%

On-Chain Activity and Network Income

Avalanche's network income encompasses all transaction fees from the C chain, P chain, and X chain, all of which are burned, with the C chain accounting for about 97%.

In addition to C chain fees, there is a second stream: L1 validator fees. Since Avalanche9000 removed the requirement to lock 2,000 AVAX to launch a sovereign L1, each L1 validator now pays about 1.33 AVAX per month to the P chain, which is burned. This stream is currently small, amounting to a few thousand AVAX per quarter, compared to tens of thousands from the C chain, but it grows slightly with each new L1 launched. The contribution from the X chain can be ignored.

Transaction costs have significantly decreased over the past year due to three upgrades. The most radical, Etna in December 2024, cut the minimum base fee for the C chain by 96%, from 25 nAVAX per unit of gas to 1 nAVAX. Octane in April 2025 improved the responsiveness of fees to demand; previously, surges in activity would quickly push fees up and slowly bring them back down, but Octane smoothed this adjustment process, bringing fees closer to actual usage levels, with average transaction fees dropping by another 43%. Granite in November 2025 adjusted block production speed according to demand, making it faster during busy periods and slower during quiet times, without affecting sub-second finality.

Each upgrade has reduced the cost per transaction, leading to increased activity. C chain transactions rose from 58 million in Q2 2025 to 236 million in Q2 2026, nearly quadrupling in a year.

Network income in AVAX terms rose from 58,000 AVAX in Q1 2025 to a peak of 134,000 AVAX in Q4 2025, then fell by about 70% in 2026 due to softening demand and a smoother fee mechanism lowering the cost per transaction. The decline in dollar terms was even greater, as the price of AVAX dropped from about $26 in Q3 2025 to about $8 in Q2 2026.

Overall, the decline in network income is not due to fewer users, but primarily because the cost of block space has become cheaper.

L1 Ecosystem

Currently, 33 sovereign L1 networks are operating on the Avalanche mainnet. L1s on Avalanche are application-specific chains with their own sets of validators, offering benefits such as: validator sets can be permissioned or permissionless; transaction fees can be paid in the project's own tokens rather than AVAX; validators can be required to undergo identity verification, operate in specific jurisdictions, or hold licenses; each network has dedicated block space, avoiding competition for throughput with other chains while maintaining connectivity through cross-chain messaging.

AvaCloud is Ava Labs' managed deployment platform, allowing teams to avoid building and maintaining validator infrastructure, RPC nodes, indexers, and block explorers, all configurable through a web console.

FIFA has partnered with Modex to build a dedicated Avalanche L1 for the 2026 World Cup, bringing the secondary ticket market into its ecosystem, eliminating the need for third-party resale platforms like StubHub.

Fans first receive a "purchase right" token, which allows them to prioritize purchasing tickets for designated matches before the public sale. This token can be traded on FIFA's own market, with both buyers and sellers paying fees to FIFA. When fans actually want to buy tickets, they exchange the purchase right for a "ticket collection right," completing the purchase through FIFA's regular ticketing system. Each token is verifiable and traceable on-chain, eliminating fraud from the source. Users operate on FIFA's standard platform without needing to know they are using blockchain.

More than 100,000 purchase right tokens have been issued, with a total trading volume exceeding $25 million. In the days leading up to the event, ticketing activity on Avalanche generated over 60,000 transactions, with a single-day peak of 24,110 transactions, several times the usual volume.

Another example is Progmat, Japan's largest tokenized securities platform, initiated by Mitsubishi UFJ Financial Group, with major Japanese banks, exchanges, and technology companies as shareholders, accounting for 63% of Japan's cumulative issuance of security tokens, covering real estate, corporate bonds, and other institutional assets. The platform's benefits are operational: settlements are completed in real-time rather than taking several days, assets can be traded around the clock, and ownership is automatically verified, eliminating the need for manual reconciliation between intermediaries.

Progmat previously operated on Corda, a private ledger used by financial institutions but closed to the outside. In February 2026, it announced a migration to a dedicated Avalanche L1, making its tokens EVM-compatible and allowing interaction with a broader public chain ecosystem while obtaining a compliant environment with its own set of validators, which is necessary for regulated securities issuance. The migration involves over $2 billion in tokenized assets and is still ongoing.

Upcoming Upgrades

Four developments define Avalanche's recent direction, one already in effect, two in the proposal process, and one in development with no set timeline.

ACP-267 regarding validator online rates took effect on April 1, 2026, raising the minimum online rate requirement for all staking periods starting on that date from 80% to 90%, with validators below the threshold receiving no rewards for that period. This change will gradually take effect as existing staking periods expire and validators rebind.

ACP-236 regarding continuous staking has not yet launched. Currently, stakers must set an end date when delegating, and after expiration, they must manually initiate a new staking transaction to rebind; this proposal eliminates that requirement, allowing staking to automatically renew at the end of each period.

ACP-194 regarding streaming asynchronous execution has not yet launched. Currently, a block must be fully executed before it can be accepted by consensus; this proposal decouples the two, allowing consensus and execution to run in parallel, with the network accepting blocks as valid once recognized, while execution completes separately. The effect is higher throughput and lower latency under heavy load, without changing how validators and users interact with the network.

FIREWOOD is a new state database written in Rust by Ava Labs, providing more efficient access to blockchain state than existing solutions, with the mainnet timeline yet to be determined, primarily aimed at high-throughput L1s where state access has become a bottleneck.

After the past year's fee reductions, the C chain's burn relative to inflation has become quite small. Moving forward, the more important driver will be the expansion of L1 numbers, as each sovereign L1 pays about 1.33 AVAX per validator per month to the P chain, which is burned; the more networks there are, the larger this fee stream becomes. The actual yield is currently around 5.7% and is gradually decreasing as the circulating supply approaches the 720 million cap. The optimistic scenario depends on whether the scale of L1 fees can be substantial enough to materially offset this decline.

NEAR

Staking Rewards

NEAR's staking rewards are provided by protocol inflation. In Q4 2025, NEAR reduced its target inflation rate from 5% to 2.5%, with 10% allocated to the protocol treasury operated by the NEAR Foundation, and the remaining 90% distributed to validators and delegators. This 2.5% rate remains unchanged unless modified again through a protocol upgrade.

The total supply is about 1.3 billion, with approximately 588 million staked, resulting in a staking rate of about 45%. Since rewards are only distributed to staked NEAR, the current comprehensive annualized staking yield is about 4.9%.

Figure Caption: Comprehensive annualized staking yield: 25Q1 9.72%, 25Q2 9.83%, 25Q3 9.81%, 25Q4 6.34%, 26Q1 and 26Q2 both 4.88%

The reduction in issuance has not led to a decline in staking participation. Since the change, the staking rate has remained stable at around 45%, and the nominal yield decrease has not altered staking behavior.

A key point for validators and delegators is that NEAR's staking rewards are not driven by transaction fees. Users pay transaction fees in NEAR, determined by the consumed gas and the network gas price. During smart contract execution, about 30% of the execution fee is returned to the contract account as developer compensation, while the remaining 70% is burned. Increased activity thus raises the burn amount and contract income but does not directly increase rewards for validators or delegators.

On-Chain Activity

Total gas consumption dropped from 419,000 in Q1 2025 to 308,000 in Q4 2025, with a faster decline in 2026, estimated at 246,000 in Q1 2026 and 82,000 in Q2 2026, an 81% decrease compared to Q1 2025.

The major reason for the sharp decline in Q2 was the collapse of Kai-Ching activity. Kai-Ching is the rewards and payment layer behind KaiKai, a shopping and loyalty application developed by Cosmose AI in Singapore. Until March 2026, Kai-Ching was one of the largest sources of activity on NEAR, often accounting for about half of total gas consumption.

Public information does not clearly disclose the specific commercial reasons for the activity collapse. Public materials show that KaiKai and Kai-Ching are built around rewards, cashback, payments, and in-app user interactions, with transaction costs hidden from end users.

Figure Caption: Total transaction count and Kai-Ching transaction count: 25Q1 619 million (of which 233 million), 26Q1 310.8 million (of which 202 million), 26Q2 77.7 million (of which 1.4 million)

Transaction numbers show the same pattern. Total transactions fell from 619 million in Q1 2025 to 310.8 million in Q1 2026, and further down to 77.7 million in Q2 2026, a 75% quarter-on-quarter decline. Kai-Ching transactions dropped from about 202 million in Q1 2026 to only 1.4 million in Q2 2026, explaining most of the decline.

Since NEAR's fees are a function of gas consumption, on-chain revenue has declined alongside activity. Comprehensive fee income fell from about $1.4 million in Q1 2025 to $306,000 in Q1 2026 and $137,000 in Q2 2026. The decrease in gas usage remains the primary reason for the fee decline in Q2, although NEAR's price rebounded from about $1 to about $2.8, partially buffering the impact in dollar terms.

Intents

Intents have become a key new growth point for NEAR. It is an execution layer where users specify the desired outcome, such as exchanging or transferring an asset across chains, and specialized entities called solvers compete to execute and provide the best price.

After going live, transaction volume grew rapidly, rising from $332 million in August 2025 to a peak of $3.7 billion in November. Monthly transaction volume stabilized between $1.9 billion and $2.7 billion in the first half of 2026. Ethereum is the largest source of transaction volume. In June 2026, it rebounded to $2.4 billion, with the chain's composition becoming more diversified, with significant contributions from Zcash, NEAR, TRON, Solana, and BSC.

Fees generally follow transaction volume but are not perfectly synchronized. The most notable divergence occurred in December 2025, when transaction volume dropped from $3.65 billion to $2.01 billion, yet fees remained high at $4.12 million. A smaller divergence occurred in June 2026, when transaction volume rose to $2.35 billion while fees remained roughly flat at $3.11 million, still below early-year levels. The differences mainly stem from the combination of routing and fee mechanisms; different Intents routes will touch different fee structures, so total fees will not correlate one-to-one with transaction volume.

In February 2026, NEAR introduced a buyback mechanism for Intents, using the protocol's retained Intents income to buy NEAR after deducting shares for partners and integrators. From the mechanism's launch until June 2026, Intents generated about $2.26 million in buyback funds.

Figure Caption: Distribution of assets in Intents transactions: USDT $1.07 billion (45.5%), USDC $516 million (21.9%), BTC $248 million (10.5%), ETH $154 million (6.5%), ZEC $142 million (6.0%)

Transaction volume is highly concentrated in stablecoins, with USDT and USDC together exceeding 67%. Next are BTC, ETH, and ZEC, followed by longer tails like wNEAR and SOL.

The comparison of Intents with the underlying chain in terms of REV is now quite stark. In Q2 2026, Intents generated comprehensive fees of $9.3 million, while the NEAR chain itself generated comprehensive fees of $137,000. In this regard, the fees generated by Intents are approximately 68 times that of the underlying chain.

Confidential Execution and AI Infrastructure

Privacy is one of the main pillars of NEAR's current roadmap. Confidential Intents (CIs) add a layer of privacy to Intents, with a model similar to regular Intents, but execution is routed to a restricted environment, where details such as assets, amounts, routes, or counterparties are not fully exposed before settlement.

It should be noted that this is not complete cryptographic privacy. The current design relies on a private NEAR shard, permissioned validators, and TEE-based infrastructure, so a more accurate term would be confidential execution rather than complete privacy.

By the end of June, the total TVL of Intents rose to $78.7 million, with confidential Intents reaching $28.5 million, accounting for about 36%. Confidential execution is becoming a significant part of the Intents stack.

At NEARCON 2026, NEAR AI launched IronClaw, a secure runtime for autonomous agents. It is an open-source Rust implementation inspired by OpenClaw, designed around privacy, tool isolation, and credential protection.

The main feature is sandbox execution. IronClaw runs untrusted tools in isolated containers, limiting the scope of access for agent tools, and reducing the impact when tool behavior is abnormal or manipulated. Its design also ensures that credentials are not exposed to large models.

Another aspect is infrastructure integration. NEAR AI positions IronClaw as part of a broader confidential computing stack, which includes NEAR AI Cloud, cryptographic enclaves, private inference, and a confidential GPU market. In this setup, IronClaw provides the agent runtime, the market provides private computing power, and Intents provide the transaction and settlement layer.

The AI product roadmap is still in its early stages. IronClaw's public repository remains active, but currently, there is limited visible evidence of adoption.

Tempo

Transaction Volume, Annualized Scale, and Fees

Tempo is an EVM-compatible L1 incubated by Stripe and Paradigm, built for one purpose: fast, low-cost stablecoin payments. Transaction fees are paid directly in stablecoins, and the protocol automatically converts them into the validators' preferred tokens. The mainnet launched on March 18, 2026.

In Q2 2026, Tempo processed approximately $386 million in stablecoin transfers (excluding minting, treasury, and fee system addresses). Transaction volume grew from $23 million in March (the first incomplete month of the mainnet) to $63 million in April, $118 million in May, and $205 million in June. At June's pace, the chain's annualized stablecoin payment volume is about $2.5 billion.

Transaction fees have been intentionally kept very low. Total fees in Q2 were about $39,800, with a median fee of $0.0012, almost all paid in pathUSD and USDC.e.

April was particularly notable, accounting for $28,900 of the total. Throughout the quarter, gas prices remained stable at 20 to 21 gwei, so the difference was not due to price changes but rather the types of transactions. April saw a concentration of computationally intensive activities: cross-chain bridging, NFT market transactions, token authorizations, and new contract deployments. These operations consumed about 276,000 gas, while standard stablecoin transfers only required 52,000, directly reflecting higher fees under unchanged gas prices. In May and June, the chain returned to lightweight stablecoin transfers, with median fees dropping to about $0.001.

Stablecoin Structure

Stablecoin TVL at the end of Q2 2026 was about $40.2 million, growing about sevenfold from $6.3 million at the beginning of the quarter. The gap between TVL and transaction volume reflects the nature of the payment chain: the same batch of stablecoins is repeatedly transferred rather than sitting idle on the chain.

The two largest stablecoins on Tempo, USDT0 and USDC.e, both came in through cross-chain bridges, rather than being natively issued on this network. USDT0 uses LayerZero's burn-mint standard, where USDT is locked on Ethereum, and when funds arrive, USDT0 is minted on Tempo, and destroyed when leaving. USDC.e comes in through Stargate, which is also a pool-based bridge based on LayerZero. Since both enter via bridges, most of their transaction volume flows through a few bridging contracts and routers, rather than directly between end users.

Figure Caption: Q2 2026 transaction volume: USDT0 $177.5 million, pathUSD $88.7 million, USDC.e $71.3 million, DLUSD $43.2 million, USDB $5.2 million

The difference lies in the breadth of usage. USDC.e was completed through about 146,000 different senders via 2.3 million transactions, with identifiable payers including Deel. USDT0 had only about 2,100 senders and 90,000 transactions, with the largest senders being automated LayerZero bridges and routing contracts.

PathUSD is another case. Issued by Stripe's Bridge, it is minted directly on Tempo, recording 800,000 transactions and about 49,000 senders, without bridging intermediaries.

Thus, pathUSD and USDC.e reflect real multi-user payment activity, while USDT0's volume is primarily due to bridging and operational flows. This portion is still counted, as it flows through bridges and routing contracts rather than system addresses.

Deel's Stablecoin Payroll

The most explicit corporate use case on the network is the global payroll platform Deel. Its payment wallet issued about $30 million to approximately 7,200 contractors in Q2 2026.

Global contractor salaries are settled in stablecoins at nearly zero fees, which is precisely what Tempo was created to do. In the first 100 days since the mainnet launch, the annualized transfer volume of this network has already exceeded $2.5 billion, with confirmed corporate usage like Deel, making pathUSD the leading natively issued stablecoin on-chain.