AI x Crypto Has Data Now. Ethereum Still Needs Proof.
AI x crypto is getting real data, but investors need to separate the outside benchmark from the EVM watchlist.
In this edition of The Weekly Alpha:
🧑🌾 AI x Crypto on Ethereum
🧐 Onchain Analytics
📚 This Week’s Intel
🎧 Podcast Picks
Hello everyone,
AI x crypto is back in the market’s attention.
The useful read is narrower than the headline.
Four different things are being priced under one label: AI networks, access tokens, payment rails, and early onchain AI markets. Those are not the same trade.
The question for investors is simple: where is there real usage, where does crypto improve the product, and who captures value if that usage grows?
AI x Crypto on Ethereum
Start by separating the benchmark from the Ethereum trade.
Bittensor is the outside benchmark. It has an active subnet market and a real cost to compete for capacity, but its EVM runs on Bittensor/Subtensor, not Ethereum. For ETH and EVM investors, TAO is a comparison point, not direct Ethereum-native AI exposure.
The EVM watchlist is much narrower.
Investor map:
Bittensor: outside benchmark. What has to be proven: useful subnet demand. beyond TAO speculation.
Venice / VVV: Base-nativeprivate AI access token. What has to be proven: users stake or hold VVV for inference, not. just narrative exposure.
x402: stablecoin payment rail fo. APIs and agents. What has to be proven:API providers, agents, wallets, and facilitators actually adopt it.
Gensyn: liveAI-settled information-market product, with optional verifiable settlement. What has to be proven: real users, workloads, and settlement volume. has to be proven: users stake or hold VVV for
That is the map for this issue.
Bittensor: The Outside Benchmark
Bittensor should not be treated as an Ethereum AI project.
Its docs describe a subnet model where miners produce digital commodities and validators evaluate the work. Those commodities can include AI inference, training, compute, storage, prediction markets, and more.
Subnet registration cost is one useful demand signal. At 1,500 TAO, the slot price is roughly $326K using the June 24 CoinGecko price of about $217 per TAO.
TAO itself is weak, down about 14.8% over seven days at the time I checked. That does not kill the subnet-demand story, but it does mean price and network demand are moving in different directions.
For Ethereum investors, the point is simple: Bittensor sets the bar. If EVM AI projects want serious capital, they need to show comparable demand instead of borrowing the AI label.
VVV: The Cleanest Base Token to Watch
Venice Token is an ERC-20 on Base tied to Venice AI, a privacy-focused generative AI product. CoinGecko describes the token as a way to access Venice Pro and API inference capacity through staking.
This is not mainly an “AI verification” protocol. It is closer to an access and capacity token for private AI inference.
VVV had a market cap of about $609M when I checked on June 24, down about 21.2% over seven days.
The underwriting question is whether VVV becomes a service-access asset or just another AI narrative coin. I would watch staked supply, API usage, paying users, contract/admin risk, and whether demand holds up when price is weak.
x402: Important, but Not a Token Trade
x402 is not a GPU compute tokenization project.
It is an open payment protocol developed by Coinbase that revives HTTP 402 Payment Required. The idea is simple: an API or digital service can request payment over HTTP, and a human developer or AI agent can pay programmatically. Stablecoins are the cleanest current use case.
That could matter if AI agents start buying APIs, data, tools, or content per request.
The bad trade is buying random x402-labeled tokens as if they own the standard. For now, the cleaner expression is indirect: Base, USDC usage, Coinbase distribution, and real API providers adding support.
Gensyn: Too Early for Conviction
Gensyn is building decentralized infrastructure for AI to train, verify, trade, and evolve. Its Delphi product is live as a Gensyn-built information market where users trade outcomes and markets use AI settlement prompts. Some markets can use verifiable settlement through Gensyn’s REE, but that is still early and not broad adoption yet.
That is better than a whitepaper, but still early.
Gensyn’s AI token had a market cap around $31M when I checked. I would want to see recurring users, real settlement volume, and workloads that clearly need crypto coordination before treating it as more than optionality.
My Ranking
If I had to rank the current setup:
Outside benchmark: Bittensor
Cleanest EVM/Base AI token to watch: VVV
Most important EVM-adjacent infrastructure idea: x402
Too early for conviction: Gensyn
The main mistake to avoid is paying conviction prices for projects that only have narrative evidence.
Onchain Analytics
This is what I checked on June 24.
Hyperliquid Is Still a Fee Machine, but Ecosystem TVL Cooled
Hyperliquid Perps did about $2.49M in fees over 24 hours and $14.5M over seven days in the DefiLlama fee dataset I checked on June 24.
That is still serious activity.
But the broader Hyperliquid ecosystem cooled over the week. Hyperliquid Bridge was around $5.75B, down about 2.8% over seven days. Kinetiq kHYPE was around $931M, down 13.5%. HyperLend Pooled was around $419M, down 13.1%. stHYPE was around $174M, down 13.3%.
The read: Hyperliquid remains one of the strongest onchain trading venues, but HYPE-denominated ecosystem TVL can move hard when the token moves.
Pendle Is Weak, but Not Below $1B
DefiLlama had Pendle around $1.03B, down about 13.0% over seven days.
So Pendle is back above $1B, but the seven-day contraction still matters. When yields compress, demand for splitting principal and yield gets weaker.
Spark Savings Is Drifting Lower
Spark Savings, the yield sub-pool rather than total Spark protocol TVL, was around $1.56B, down about 3.9% over seven days.
That is not a collapse. It is capital slowly repricing lower passive stablecoin yield.
Aave V4 Is Growing, but Still Tiny
Aave V4 was around $142M TVL, up about 10.2% over seven days.
That is worth tracking, but it is still tiny compared with Aave V3. Treat it as early migration data, not a full protocol shift.
This Week’s Intel
Three DeFi stories worth watching after this issue.
1. THORChain Comes Back Online After Its Longest Shutdown
THORChain resumed swaps, trading, and LP operations on June 23 after being offline for more than five weeks.
The halt followed a May 15 exploit that drained about $10.7M from one Asgard vault across Bitcoin, Ethereum, BNB Chain, and Base. The team shipped an emergency patch on May 20, a larger vulnerability fix on June 9, and a KeyVerify update on June 11 before reopening the network.
The investor read: the restart matters less than the market reaction. RUNE barely moved after trading resumed, which suggests the recovery was already priced in. The real test is whether liquidity returns to native cross-chain swaps after five weeks of downtime.
Watch: swap volume, LP deposits, RUNE liquidity depth, and whether Monero/Zcash integrations bring users back or just create roadmap noise.
Source/image: BitcoinFoundation report
2. Coinbase Opens a Door to DeFi Yield
SteakhouseFi launched a High Yield Vault inside the Coinbase app, powered by Ethena’s USDe on Morpho and running on Base.
That matters because Coinbase is not sending users to a DeFi front end. It is packaging DeFi yield inside a familiar consumer app. Portals says the vault is available to Coinbase’s 100M-plus users, while Ethena TVL rose 7.47% over the week to $4.94B.
This landed alongside Morpho’s $175M raise at a $2B valuation, co-led by Paradigm, a16z Crypto, and Ribbit Capital, with Apollo, VanEck, and Circle also participating. The pitch is bigger than lending TVL: Morpho wants to become an “Open Credit Network” for institutional credit markets on-chain.
The investor read: this is one of the cleaner DeFi adoption stories of the week. The question is whether Coinbase becomes a real distribution channel for DeFi yield, or just a short-term inflow event.
Watch: Ethena inflows, Morpho deposits on Base, vault retention after the first yield cycle, and whether Coinbase adds more DeFi-native products.
Source/chart: Portals Week 2, Portals Week 1
3. Aave Turns a Stress Event Into a Risk Framework
After April’s $292M KelpDAO exploit triggered an $8.45B, 48-hour deposit run on Aave, LlamaRisk proposed a new four-layer risk framework for the protocol.
The framework would require stronger bridge verification, a minimum bug bounty, automated Freeze Guardians, and better oracle circuit breakers. The important part is not that Aave avoided collapse. It is that the largest DeFi lender is trying to turn a live stress test into stricter collateral and bridge standards.
Separately, Aave’s “Aave Will Win” proposal shows how the DAO is thinking about monetization: revenue from Aave-branded products would flow to the treasury, with a one-year development budget to support growth. The proposal says Aave has roughly 60% lending market share.
The investor read: Aave is still the benchmark for DeFi lending. If it tightens risk without killing growth, it strengthens the moat. If new rules reduce supported collateral too aggressively, it could slow some usage in the short term.
Watch: whether the risk framework passes, which collateral assets are affected, and whether Aave keeps deposit share after stricter standards.
Source: Portals Week 2, Aave governance proposal
Podcast Picks
Podcasts worth your time this week.
Bankless: Venice is Here to Win: How a Private AI Company Plans to Take On OpenAI and Anthropic
Good companion to the Venice / VVV section. The useful question is whether Venice can turn private AI access and inference demand into a real token utility loop, not just another AI narrative trade.
Odyssey: #088 - x402 vs MPP and All About Agents w/David
Best fit for the x402 section. It gets into agentic commerce, pay-per-request services, and why payments may be the cleaner crypto use case than decentralized GPU compute.
Crypto Prime: Hyperliquid Explained with Bitwise CIO Matt Hougan
Useful after the onchain analytics section. Hyperliquid still has real fees and institutional attention, but HYPE and ecosystem TVL can move hard. The distinction matters.
That is it for this week.
My read is simple: AI x crypto is getting more interesting, but only when you separate usage from branding. Bittensor has the strongest network signal. VVV has the cleanest Base token surface. x402 is infrastructure, not a token thesis. Gensyn is live, but still needs usage.
Share this with someone trying to separate AI crypto infrastructure from the noise.
Sources checked: CoinGecko TAO, CoinGecko VVV, CoinGecko Gensyn, Bittensor docs, Bittensor EVM docs, Coinbase x402 docs, Venice AI, Gensyn, and DefiLlama.
Nothing in this newsletter is financial advice. I personally use several of the protocols mentioned here, but that does not make them safe. Crypto is risky, DeFi is risky, and things can break quickly. Do your own research, size positions carefully, and never invest more than you can afford to lose.