Overview of Recent Key Events: ETH Investment Logic Shifts, AI Valuation Raises Red Flags, Multicoin Bets on ZEC and HYPE
TechFlow Selected TechFlow Selected
Overview of Recent Key Events: ETH Investment Logic Shifts, AI Valuation Raises Red Flags, Multicoin Bets on ZEC and HYPE
Summary of the four most important investment research observations in the recent crypto market.
Navigating Web3 tides with focused insightsAuthor: ChainFeeds
Compiled by: TechFlow
TechFlow Editor's Note: This article summarizes four of the most important investment research observations in the crypto market recently: Ethereum's value capture failure, BlackRock warning that AI valuations are at the "halfway point," Multicoin heavily positioning in Zcash and HYPE, and why Wall Street is rejecting ChatGPT. For investors wanting to understand the allocation logic of the current cycle, these four judgments directly relate to whether you should still hold ETH, how long AI stocks can continue to rise, and whether the privacy AI sector is a real demand.
Ethereum Ecosystem Panorama Analysis: What Changes Have Occurred in Investment Logic?
Nick Researcher re-examines Ethereum from a macro and financial perspective. Data from Q2 2026 reveals a complex signal: Ethereum revenue improved slightly compared to the previous quarter, but L1 (Mainnet) fee capture capability remains far lower than last year; on-chain yields are near historic lows; DeFi activity has declined; although the L2 (Layer 2) ecosystem continues to expand, and projects like Robinhood have begun developing based on Ethereum-related infrastructure, these activities have not contributed sufficient value to L1; meanwhile, ETH's dilution rate remains maintained at a level close to Bitcoin.
The biggest controversy surrounding current ETH is that Ethereum has not lost competitiveness, but the past investment logic is changing. Previously, the bullish logic for ETH in the market was very simple: more users enter Ethereum, driving increased L1 activity, generating more fees, bringing more ETH burns, thereby enhancing ETH's value capture capability. But this model is weakening. Users are gradually migrating to L2, while some users are leaving the ecosystem because the L2 experience did not meet expectations; transaction fees are declining, while Blob supply growth speed exceeds demand; although L2 processes a large volume of transaction activity, the fees paid to the Ethereum Mainnet remain limited.
The most critical data is that in Q2 2026, the Real Economic Value generated by Ethereum L1 was $88.4 million, a 7% quarter-over-quarter increase, but a 68% year-over-year decrease. At the same time, the application layer on Ethereum L1 generated approximately $1.79 billion in fee revenue. This indicates that applications within the Ethereum ecosystem still possess strong economic value, but as the base layer, the Ethereum Mainnet captures only a small portion of it. This is also the core contradiction in the current ETH investment logic.
Ethereum still carries a large volume of important financial activities, including major protocols such as Tether, Circle, Lido, Aave, and Uniswap, which are all important participants in the Ethereum ecosystem. Stablecoins remain one of Ethereum's strongest advantages; in Q2 2026, the Ethereum L1 stablecoin supply reached $172.9 billion, and despite a quarter-over-quarter decrease of about 4%, it still maintains a huge scale. But scale is not the only key factor; capital flow velocity is equally important. If stablecoins just stay on-chain without undergoing transactions, settlements, collateralization, and other financial activities, then they will not create sufficient economic value. The current Ethereum possesses a massive asset scale but lacks sufficient capital turnover efficiency.
Real World Assets (RWA) may become an important growth driver for ETH in the next stage. Currently, the on-chain RWA scale on Ethereum L1 has exceeded $15.7 billion, a year-over-year increase of about 90%, including tokenized treasury bonds, commodities, and stock assets. But simply having higher TVL is not enough to prove value capture capability. In Q2 2026, Solana's RWA daily trading volume exceeded Ethereum, despite its lower RWA TVL, indicating that Ethereum's advantage lies more in institutional depth, while Solana's advantage lies in capital flow velocity.
For ETH, the future bullish logic requires three conditions to hold simultaneously: First, more institutional assets enter the Ethereum ecosystem; Second, more financial settlement activities occur on the Ethereum network; Third, on-chain assets need to generate higher real transaction frequency. ETH's token economic model currently still has advantages, with an annualized net dilution rate of about 0.85% in the second quarter, close to BTC levels. But risks also exist; the total on-chain yield has dropped to 2.68%, hitting a historic low, where 94% of the yield comes from ETH issuance rather than real user fees. This means whether ETH can obtain revaluation in the future depends critically on whether it can become the settlement layer within the institutional financial system.
BlackRock Report: This AI Rally Has Reached the "Halfway Point" of the 2000 Internet Bubble, One Indicator Has Flashed a Red Light
According to BlackRock citing Morningstar data, over the 7 years from 1993 to 1999, U.S. tech stocks cumulatively rose 1097%, while the overall U.S. stock market rose 292% during the same period. Tech stocks had annualized returns of no less than 19.9% for 7 consecutive years, with 1998 and 1999 being particularly astonishing, reaching 78.1% and 78.7% respectively.
In comparison, during the AI rally cycle from 2019 to June 30, 2026, tech stocks cumulatively rose 569%, while the overall U.S. stock market rose 237% during the same period. Although this round of rally also performed strongly, the rhythm was significantly different. During this period, tech stocks experienced a significant correction in 2022, falling 28.2% for the full year, subsequently rebounding 57.8% in 2023, rising 36.6% and 24.0% in 2024 and 2025 respectively, and continuing to rise 19.8% in the first half of 2026.
The biggest difference between the two rallies appeared in the latter half: during the Internet bubble period, the market accelerated rapidly in the last two years, with cumulative gains in 1998 and 1999 approaching 200%; while although the AI rally saw significant acceleration in 2023, the magnitude of subsequent gains gradually narrowed. In other words, the current AI cycle trend is smoother compared to the Internet bubble, but there is still significant divergence in the market regarding whether it will enter a final crazy surge phase.
Current market valuations have become the core of the AI rally controversy. The S&P 500 Shiller P/E Ratio (Shiller CAPE) has risen to about 40 times, returning to the high levels of the Internet bubble period. This indicator calculates valuation levels based on the average earnings over the past 10 years adjusted for inflation; 40 times means investors are willing to pay $40 for every $1 of long-term average profit, a level similar to this has only appeared historically around 2000.
However, BlackRock believes that focusing solely on long-term valuation indicators is not comprehensive; the 12-month forward P/E ratio provides another perspective. Currently, the S&P 500 forward P/E ratio is about 21 times, mainly because corporate earnings expectations have also risen synchronously with stock prices. Data shows that S&P 500 second-quarter earnings are expected to grow 23% year-over-year, maintaining double-digit growth for the seventh consecutive quarter. BlackRock believes this earnings growth is relatively rare in history. At the same time, the Mag 7 tech giants currently have a P/E ratio of about 26 times, while earnings growth is expected to exceed 30%, with comprehensive earnings growth at about 27.6%.
Therefore, the biggest contradiction in the current market is: long-term valuation indicators have released overvaluation risk signals, but corporate earnings growth still provides support for high valuations.
As of May 31, 2026, according to Morningstar data, the proportion of tech stocks in the total U.S. stock market value has reached 37.5%, exceeding the level during the Internet bubble period in the late 1990s. If further considering companies such as Alphabet, Meta, and Amazon, which although classified outside the tech sector, are deeply involved in the AI industry, the actual AI-related asset concentration may be higher.
The current market leadership is also spreading from the traditional Mag 7 to a broader range of AI beneficiary enterprises; a new market concept MANGOS is forming, representing Meta, Anthropic, Nvidia, Google, OpenAI, and SpaceX. The Morningstar Global Next Generation Artificial Intelligence Index cumulatively rose about 45% in April and May 2026, subsequently experiencing a pullback in June.
Market concentration is one of the most similar places between the current AI cycle and the Internet bubble. At the end of 1999, a few tech companies such as Cisco, Intel, Microsoft, and Oracle drove the last round of Nasdaq gains. Nowadays, although AI leading enterprises possess stronger profitability, if future earnings growth cannot meet market expectations, highly concentrated portfolios may still face rapid adjustment risks.
BlackRock believes that judging that AI has already formed a bubble is itself a major judgment, because this means the market assumes AI cannot bring long-term productivity improvements. The question current investors truly need to focus on has shifted from "how much more can AI rise" to "how long can AI earnings growth continue."
Dialogue with Multicoin Partner: Crypto Market Has Bottomed, Bullish on Three Cryptocurrencies in This Cycle
Multicoin Capital Managing Partner Tushar Jain shared his views on the current crypto market and elaborated in detail on the investment logic for Solana, Hyperliquid, and Zcash.
Tushar Jain stated that he still believes Solana is the correct technical architecture for the internet capital market, needing a permissionless open-source chain to integrate everything into one platform. He remains bullish on Solana's performance and architecture. But at the same time, derivatives trading volume is shifting towards Hyperliquid. He currently has large positions in both these assets and is bullish on both. Solana is the leader in spot trading, will carry spot trading of tokenized securities, but Hyperliquid is obviously leading in derivatives. Rather than being an extremist, it is better to think from a probability perspective and hold both simultaneously. He is not a maximalist for any asset and will not fight to the death with a certain position or view.
Looking ahead to 2026, a very obvious choice for him is Zcash (ZEC), although due to liquidity and market cap limitations, his position is relatively small, but Multicoin has accumulated a fairly large proportion of the total supply. He likes Zcash's momentum, use cases, and community, reminding him of early Bitcoin. When he saw it rise last year, he communicated with many early bulls and found that even when prices fell back, they still stuck to their beliefs; this is not a short-term hot money game. Additionally, Zcash has no fundamentals (no cash flow and revenue), meaning its value depends entirely on people's consensus, which instead gives it greater upside space; as a store of value, the larger its scale the better.
Multicoin does hold HYPE positions, but Tushar Jain suggests investors look at their derivation logic and draw their own conclusions. The assumptions they set are not aggressive: first, crypto derivatives compound annual growth rate of 35% (past 5 years was 45%, already cut by a quarter of growth speed); second, DEX occupies 32% of derivatives market share (from almost zero in 2022 to 16% now, doubling to 32% within two years fits the trend); third, Hyperliquid maintains 30% of decentralized derivatives share (this is also conservative, because trading volume data is easy to brush, but currently Hyperliquid occupies 59% of real open interest across the network, this data is hard to fake); fourth, USDC collateral grows linearly with trading volume (as long as traders' leverage preferences remain unchanged, stablecoins as collateral will naturally grow proportionally with trading volume and open interest).
AI at the Crossroads: Why Are Wall Street Firms Saying "No" to ChatGPT and Claude?
Privacy AI is not a single technical route, but unfolds around the same core question: during the process where a prompt leaves the user device, passes through network transmission, enters the server running the model, and returns results, where exactly does plaintext exist, who can read it, and how can users verify whether their data is truly protected. Currently, privacy mechanisms on the market are essentially solving the same event but adopting different trust models.
Protocol-level privacy relies on service provider commitments, for example, in enterprise version zero-retention solutions, the service provider can know user identity and can process user prompts, but promises not to save data; execution relies mainly on contracts and brand reputation. Anonymous proxies hide user identity but do not hide the content of user input; downstream model service providers can still see plaintext. TLS can only protect data security during transmission between machines, but the receiver can ultimately still read all content.
Oblivious HTTP (OHTTP) further splits the right to know identity and content, letting the relay know the request source but unable to read content; the receiver can process the request but does not know who sent it. OHTTP has become an IETF standard and has begun to be used by some enterprises in production environments. However, for closed-source flagship models, such solutions have approached the limit of privacy protection, because model weights themselves are the core assets of AI companies. A top-tier model training cost reaches billions of dollars; labs rely on model capability gaps to maintain valuations, therefore will not easily open model weights or complete service code.
Structural-level privacy solutions attempt to replace traditional trust commitments through hardware, cryptography, or physical isolation mechanisms. Among them, Trusted Execution Environment (TEE) confidential computing is currently the path closest to commercial landing. TEE puts the model inference process into a hardware enclave to run; this area is similar to a sealed space inside the chip; even the server operator cannot directly read data within it. The chip will generate attestation (remote attestation) to prove to the user that the specified model and code are running.
But TEE still has limitations: prompts are only protected after entering the enclave; there may still be reading risks in the proxy and relay stages before entering. End-to-End Encryption (E2EE) further closes intermediate links; user devices directly use enclave keys to encrypt prompts; intermediate nodes can only pass ciphertext. However, the cost of E2EE is increased engineering complexity, because all functions relying on plaintext data operation need to be redesigned.
Fully Homomorphic Encryption (FHE) and Multi-Party Computation (MPC) attempt to completely eliminate trusting parties, letting servers compute directly in ciphertext state. But because Transformer models involve a large volume of complex operations, FHE inference costs remain far higher than ordinary inference; ciphertext computation costs may reach tens of thousands of times that of plaintext. Currently, encryption chips are developing, but distance from large-scale commercial application still requires time. In comparison, local inference is the most thorough privacy method, because the model runs on the user's own device; there are no server, relay, and data leakage issues, but the cost is model capability and hardware costs.
The future competition point of Privacy AI may not just be chat scenarios, but more complex Agent workflows. Currently, all privacy inference mechanisms mainly solve data protection between prompt and model, but when AI Agents execute tasks, they also need to call external tools, such as calendars, databases, search engines, and enterprise internal systems, and these tools will all become new plaintext exposure points. An Agent running completely locally, if wanting to obtain information outside the training set, still needs to send queries to external services, and if service providers cannot read plaintext they cannot complete tasks.
Currently mainstream solutions still stay at the protocol layer, for example, managing tool calls through central gateways, hiding personal identity information before sending requests, controlling access permissions, and recording call behavior. But this method still relies on service provider trust, because tool servers still need to read plaintext queries. Structural-level solutions attempt to run tools like MCP Server directly in TEE, letting users verify privacy commitments through attestation. However, TEE can only protect the transmission process, cannot guarantee the final service provider does not read query content. The truly difficult part is open search and complex Agent scenarios, because encrypted search currently still faces performance and cost issues.
The future value capture point of Privacy AI may concentrate on unsolved problems: running training loops in enclaves, end-to-end protecting tool calls, search systems without exposing query content. Whoever can solve one of these core links may establish truly difficult-to-commoditize infrastructure advantages.
After Gold Tokenization: How On-Chain RWA Creates Real Yield?
Most on-chain Real World Assets (RWA) currently still concentrate on low-risk assets, such as U.S. Treasury Bills (T-bills), and are gradually expanding to other asset categories such as stocks. Among them, gold is currently the largest commodity asset on-chain and is also an important case driving asset tokenization development. Currently, on-chain gold scale has exceeded $4.9 billion; its unique value storage attribute makes it one of the earliest traditional assets to be tokenized.
However, currently most on-chain gold products remain relatively limited; main functions are just letting users buy spot gold, lacking mechanisms to further utilize these assets to create yield. This leads to an efficiency gap between on-chain RWA and Traditional Finance (TradFi) products, also limiting the actual value and application scenarios of on-chain assets.
The next stage of RWA development focus may no longer be just expanding asset on-chain scale, but letting these assets possess productivity and yield capability. Taking gold as an example, traditional financial markets have already through covered call ETF and other products, let investors be able to use options to obtain yield or hedge risks. But traditional products usually exist thresholds high, fees high, need KYC, custody, and broker participation etc. limitations. For example, currently the more mature gold covered call ETF GLDI charges about 0.65% management fee, and will deduct directly from investor yields.
In comparison, on-chain gold products can through smart contracts and structured strategies, lower participation thresholds, and attempt to transform originally non-cash flow generating gold assets into yield-generating assets. Letting gold assets generate yield is an important direction for RWA next stage development. Gold itself is about $30 trillion scale asset category, and is also one of the earliest commodities to achieve on-chain tokenization. Although currently on-chain already exists over $4.9 billion gold assets, but vast majority funds still remain in idle state, will not generate yield.
With traditional financial market covered call strategy development, investors have already been able to through options obtain gold holdings outside extra yield, simultaneously reduce part price volatility risk. And Enhanced etc. on-chain protocols attempt to introduce this model into blockchain, through structured strategies boost RWA capital efficiency.
Gold why suitable as first case, is because it possesses several characteristics: first, gold long term regarded as value storage asset, recent price continuously hit new highs, attract more investors allocate; second, global geopolitical and macroeconomic uncertainty increase, further strengthen gold demand; last, gold price usually not like high volatility assets that剧烈 change, more suitable through covered call options obtain stable premium income.
Covered call strategy logic is, investors hold gold spot, simultaneously sell call options, thereby obtain option premium income. If gold price not exceed strike price, investors retain gold and obtain yield; if price rise exceed strike price, then need give up part upside space. Therefore, this strategy more suitable for gold long term bullish, but expect price will not large single side rise investors.
Enhanced launched PAXG Volatility Income Vault, is its first Thesis Vault product, goal is utilize gold volatility for users create yield. This product based on PAXG (on-chain gold token), through covered call option strategy, let users in hold gold assets simultaneously obtain option yield. Its operation mechanism based on RFQ (Request for Quotes). In backend, user deposited assets will through batch auction method, by market makers provide quotes, subsequently on-chain execute option trades, users early obtain option premium income.
Participants can also directly against own assets sell covered call options, and customize execution parameters, such as strike price, term, and direction. Future, this mechanism can also expand to gold outside other ERC-20 assets.
PAXG Vault adopts European-style options, can only at expiry date execute, funds will in each cycle within lock. Users can deposit PAXG or USDC, system will automatically convert USDC to PAXG.
Option cycle set to two weeks once, annually about 26 cycles, strike price expected set at current gold price above 3%-7% interval.
Users can choose two yield modes: compound mode will automatically obtain USDC premiums exchange into PAXG, and join next cycle continue generate yield, more suitable long term hold gold investors; income mode will will yield separately store, users can anytime extract USDC, more suitable hope from idle gold assets obtain cash flow large fund holders.
This mode attempts solve traditional RWA core problem: not just let assets on-chain, also let assets truly generate economic value.
Most on-chain Real World Assets (RWA) currently still concentrate on low-risk assets, such as U.S. Treasury Bills (T-bills), and are gradually expanding to other asset categories such as stocks. Among them, gold is currently the largest commodity asset on-chain and is also an important case driving asset tokenization development. Currently, on-chain gold scale has exceeded $4.9 billion; its unique value storage attribute makes it one of the earliest traditional assets to be tokenized.
However, currently most on-chain gold products remain relatively limited; main functions are just letting users buy spot gold, lacking mechanisms to further utilize these assets to create yield. This leads to an efficiency gap between on-chain RWA and Traditional Finance (TradFi) products, also limiting the actual value and application scenarios of on-chain assets.
The next stage of RWA development focus may no longer be just expanding asset on-chain scale, but letting these assets possess productivity and yield capability. Taking gold as an example, traditional financial markets have already through covered call ETF and other products, let investors be able to use options to obtain yield or hedge risks.
But traditional products usually exist thresholds high, fees high, need KYC, custody, and broker participation etc. limitations. For example, currently the more mature gold covered call ETF GLDI charges about 0.65% management fee, and will deduct directly from investor yields.
In comparison, on-chain gold products can through smart contracts and structured strategies, lower participation thresholds, and attempt to transform originally non-cash flow generating gold assets into yield-generating assets.
Letting gold assets generate yield is an important direction for RWA next stage development. Gold itself is about $30 trillion scale asset category, and is also one of the earliest commodities to achieve on-chain tokenization.
Although currently on-chain already exists over $4.9 billion gold assets, but vast majority funds still remain in idle state, will not generate yield. With traditional financial market covered call strategy development, investors have already been able to through options obtain gold holdings outside extra yield, simultaneously reduce part price volatility risk.
And Enhanced etc. on-chain protocols attempt to introduce this model into blockchain, through structured strategies boost RWA capital efficiency.
Gold why suitable as first case, is because it possesses several characteristics: first, gold long term regarded as value storage asset, recent price continuously hit new highs, attract more investors allocate; second, global geopolitical and macroeconomic uncertainty increase, further strengthen gold demand; last, gold price usually not like high volatility assets that剧烈 change, more suitable through covered call options obtain stable premium income.
Covered call strategy logic is, investors hold gold spot, simultaneously sell call options, thereby obtain option premium income. If gold price not exceed strike price, investors retain gold and obtain yield; if price rise exceed strike price, then need give up part upside space.
Therefore, this strategy more suitable for gold long term bullish, but expect price will not large single side rise investors.
Enhanced launched PAXG Volatility Income Vault, is its first Thesis Vault product, goal is utilize gold volatility for users create yield. This product based on PAXG (on-chain gold token), through covered call option strategy, let users in hold gold assets simultaneously obtain option yield.
Its operation mechanism based on RFQ (Request for Quotes). In backend, user deposited assets will through batch auction method, by market makers provide quotes, subsequently on-chain execute option trades, users early obtain option premium income.
Participants can also directly against own assets sell covered call options, and customize execution parameters, such as strike price, term, and direction. Future, this mechanism can also expand to gold outside other ERC-20 assets.
PAXG Vault adopts European-style options, can only at expiry date execute, funds will in each cycle within lock. Users can deposit PAXG or USDC, system will automatically convert USDC to PAXG.
Option cycle set to two weeks once, annually about 26 cycles, strike price expected set at current gold price above 3%-7% interval.
Users can choose two yield modes: compound mode will automatically obtain USDC premiums exchange into PAXG, and join next cycle continue generate yield, more suitable long term hold gold investors; income mode will will yield separately store, users can anytime extract USDC, more suitable hope from idle gold assets obtain cash flow large fund holders.
This mode attempts solve traditional RWA core problem: not just let assets on-chain, also let assets truly generate economic value.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@ChainFeedsxyz
