# AI infrastructure stocks — X 热门讨论 (2026-09-21 14:08 UTC)
## @Tanaka_L2 (Tanaka) · 09-21 12:05 · ♥80 ↻3 💬24 AI-aligned tokens are entering a different stage on Robinhood.
Frens asked me how I think about $AI in specific, and AI tokens onchain in general.
So I share my genuine thoughts here, and I’m open to your opinions as well.
Giving an agent real capital, enforceable permissions and access to tokenized financial assets is hard.
The protocols I currently see building toward that second model:
[1] @virtuals_io
Virtuals is building the main agent-launch and coordination layer on @RobinhoodCrypto.
Its agents can raise capital, launch tokens and manage tokenized-stock positions.
The catalyst I'm watching is whether these agents begin generating measurable returns from real financial assets instead of relying mainly on token speculation.
[2] @sherwoodagent
Sherwood is experimenting with agent-managed ERC-4626 vaults.
Agents propose strategies, depositors vote and guardians verify execution.
The available assets include Stock Tokens, USDG and perps.
This is closer to an onchain asset-management model.
The catalyst is live vault AUM and a transparent performance history.
[3] @archeragentAI
Archer is building controlled agent execution:
Intent → policy check → human approval → transaction.
I find this important because financial agents need defined permissions before serious capital can use them.
The catalyst is integration with more Robinhood markets and recurring execution volume.
[4] @HoodAI0x
Hood AI is working on 2 less visible problems:
– correctly reading Stock Token balances after corporate actions.
– enabling agents to provide services through escrow.
This is infra rather than a narrative trade.
The catalyst is adoption by wallets, explorers and other agent protocols.
[5] @longdotxyz
LONG is not purely an AI protocol, but it provides important market infrastructure for AI-aligned tokens.
Tokens can trade directly against Stock Tokens such as $NVDA, while LongX adds leveraged stock exposure.
$AI paired with $NVDA is one early example, but the wider opportunity is allowing tokens, agents and communities to hold or earn productive financial assets.
Around these projects, @arcus_xyz, @Lighter_xyz and @Morpho provide the execution, leverage and lending rails that agents can eventually use.
This is where the sector becomes more interesting to me.
An AI token alone has limited utility.
An agent that can trade tokenized equities, borrow against them, hedge through perps and report its performance has a financial product.
The next immediate catalyst is Arc public mainnet on September 16.
Arc launches with USDC as gas, tokenized-asset infrastructure and Circle’s agent stack.
Native USDC should also remove much of the current friction around liquidity and settlement.
I do not see a clear AI-token leader on Arc yet.
Most projects are still launchpads, terminals or pre-mainnet experiments.
I want to see which teams attract real capital after launch rather than only temporary volume.
My current framework is simple:
– Agent tokens need recurring activity
– Agent vaults need AUM and verifiable returns
– Execution protocols need real transaction volume
– AI infrastructure needs integrations
– Stock-aligned tokens need transparent reserves and clear holder rights
At the macro level, the setup is strong.
AI spending continues to support assets such as $NVDA, while tokenized stocks are becoming usable across spot markets, lending, perps and automated portfolios. https://x.com/Tanaka_L2/status/2102006344245260734