IOSG: A full breakdown of Robinhood’s self-built L2—from meme cold starts to real-world asset (RWA) implementation

Author | Mario @ IOSG

Key takeaways

Robinhood no longer rents other people’s blockspace. It built its own L2, taking full control of trading, settlement, collateral, yield, and asset transfers. This is a positive response to Coinbase Base: it changes from being a tenant on other chains to becoming the “landlord” of its own settlement layer. The entire suite of tokenization products (24/7 stock tokens, USDG lending, and perps) is designed for one purpose only: keeping both users and economic value within Robinhood’s own arena.

This launch also unexpectedly gained a marketing machine no one saw coming: meme coins. Within one week of mainnet上线, Tenev shifted from publicly downplaying memes to following the CASHCAT account on X. That stance sparked a wave of speculation, and Robinhood Chain became one of the busiest chains in crypto-world within its first month. No matter how you evaluate the quality of that traffic, it solves most of the cold-start problems that kill new L2s (see Section 2 for details).

The first three weeks: memes arrived before stocks

Robinhood built this chain for tokenized stocks, but the first thing to move in was a meme casino. After three weeks on mainnet, the casino still accounts for most active usage, yet the first batch of genuinely interesting RWA-native projects also grew out of this place.

As of 20 July 2026 data:

What’s being traded for real? Memes. The leader is $CASHCAT, a cat coin named after Robinhood’s former mascot before its redesign. In its first week it surged more than 2000%, and market cap climbed to about $156 million—about an order of magnitude larger than the RWA assets on the entire chain. A whole batch of memes (Cash Dog in Hood, Little John, Hoodrat) and the supporting launch infrastructure (NOXA.fun launchpads, basedbot) were all set up within days. The total market cap of the meme sector is roughly between $60k and $200 million.

The second flywheel: AI agents. Speculative traffic isn’t only memes. From day one, Robinhood integrated Virtuals Protocol’s agent infrastructure. This isn’t a supporting role—“Agentic Trading” is literally written in the title of Robinhood’s official release. Tenev was blunt about the direction: in May this year, Robinhood had already launched Agentic Trading and an Agentic credit card inside the brokerage app. He told CNBC, “Every kind of operation a human can do, AI agents will be able to do.” The end goal is for ordinary people to get, for the first time, the same “tools, same compute, same capabilities” that high-frequency trading institutions have enjoyed for decades. This chain is an open sandbox for that argument: through Virtuals’ Agent Commerce Protocol, anyone can launch, fund, hold, and use agents in tokenized markets—each agent comes with an on-chain identity, a non-custodial wallet, a payment card, and an inbox (Virtuals calls it EconomyOS).

The growth curve for agents is even steeper than memes. In the first week: 2100+ agents, about $77 million in trading volume, and developers earned $1.3 million. Agent trading volume went from 0 to $100 million in two weeks, and from $100 million to $150 million in three days. By 17 July, agent count was 4500+ with trading volume of $150 million+, and cumulative developer fundraising was $2.3 million. That same week also shipped the largest on-chain agent and robot project so far. Distribution channels are widening too: starting 18 July, all Virtuals agents on Robinhood Chain can be discovered in Binance Wallet’s Meme Rush. At this stage, no single “dominant” agent token has emerged—the real big player is Virtuals itself as the infrastructure layer, and $VIRTUAL rose about 20% on the partnership news. To be frank: most agent-token trading today is essentially memes wrapped in an AI shell. Until agents show sustained revenue, this trading volume should still be viewed as speculative flows.

What these agents look like specifically (Virtuals examples on Robinhood Chain):

2. Monvera ($MONVERA) is the most typical RWA-native case: an AI brokerage launched on 14 July. It directly connects tokenized stocks on-chain—wrapping about 95 Robinhood on-chain stock tokens into a single agent. The agent does research, quotes, and routes trades on behalf of users. This is the combination of agents and stock tokens—not agents and memes.

3. Quiver Protocol ($QUIV) claims to be the first AI-driven yield aggregator on-chain: in LP vaults, agents rebalance positions, reinvest, and stop losses on-chain, but at the architecture level they are forbidden from withdrawing users’ funds.

4. Grid Arena turns price charts into a prediction arena: lock grids inside Nvidia, Tesla, or Apple price grids, and each grid has its own real-time odds multiplier.

5. Hyperium ($HYP) is a multi-terminal trading / development environment aimed at traders who are tired of constantly switching tabs back and forth.

6. Root Edge is an autonomous perpetual trading agent (Hyperliquid). After about 8 months of R&D it entered beta, distributing rootAI “Skill” NFTs to early users.

Read through this list and it’s obvious at a glance: the two projects that actually “ran out” both connect to RWA (Monvera connects stock tokens, Quiver connects on-chain yields). That’s the kind of agent an RWA chain is trying to grow. The rest still look like memes wrapped in an AI shell—same pattern as the earlier tokens.

Then the meme faucet got shut. NOXA deployed 60k+ tokens in under two weeks (about 75% of the chain’s total token issuance), collected nearly $12 million in fees, and suddenly stopped new token launches on 11 July. The team said the robots were constantly刷仿盘. Two days later it completely disappeared—domain lost, leaving only an IPFS interface, with no mention of a restart time. No matter what the original intent was, the objective effect was to forcibly cool meme issuance; the liquidity and attention that had been chasing new launches started shifting toward RWA-related tokens.

This is the second-week turning point that’s actually more interesting: the tokens that emerged were no longer pure memes—they started combining with stock tokens:

2. Arrow Finance ($ARROW) is a CDP (collateralized debt position—locking collateral to borrow funds) protocol, the first project to accept tokenized stocks and ETFs as collateral to mint its stablecoin aUSD. Plainly: deposit your AAPL token, don’t have to sell it, and you can borrow dollars. It also operates a launchpad (Arrow Pad). $ARROW rose from about $0.15 at its 7 July launch to about $1.79 (market cap about $16 million)—a 10x in less than two weeks.

3. $INDEX routes trading fees to buying on-chain stock tokens and distributing them to holders—essentially putting a rough dividend-like mechanism on top of the stock-token ecosystem. After Tenev publicly encouraged developers to build apps integrating tokenized stocks and RWA, it jumped about 150% in a single day, and market cap reached the ten-million-dollar level.

Tenev’s personal stance is worth reading closely, because he changed fast. On 2 July, the day after mainnet上线, he told CNBC that meme coins basically pushed the market into a dead end: creating assets without utility can’t produce persistent value. Issuing hundreds of such tokens is meaningless; tokenized RWA is the persistent direction. Six days later, with CASHCAT’s market cap nearing the nine-figure range, he posted on X: “We’re turning Robinhood Chain into the best RWA chain… but running memes on it is also very useful,” and he followed the CASHCAT account. By 14 July, he again publicly pushed developers to build apps integrating stock tokens and RWA—exactly the post that made INDEX jump 150% in a single day. Put together, this looks less like vacillation and more like a playbook: maintain RWA legitimacy with regulators and institutions while catching the meme traffic that’s paying the bills right now.

Our view: this is a replay of the Base script. Memes are liquidity on-ramp and user acquisition channels. They stress-test infrastructure, deepen DEX order-book depth, and give the chain a heartbeat in its first month that pure RWA traffic can’t provide. The signal worth tracking isn’t meme market cap, but whether the first batch of useful projects are all connecting stock tokens into DeFi primitives (Arrow uses them as collateral, INDEX does yield distribution). That’s precisely the behavior an RWA chain needs to grow—and Robinhood’s team is clearly fanning the flames. The open question: RWA assets still make up only about 4% of TVL. If the scale of stock tokens can’t keep up with the user volume memes bring, then this chain is just a casino with a brokerage signboard. Base at the time also didn’t truly solve that conversion.

How the chain is built, and who built it together

First, the plain-English version: Robinhood Chain is a rollup. It produces blocks itself—fast and cheap—then sends transaction data back to Ethereum, where Ethereum acts as the final record court. Robinhood controls the sequencer (the machine that queues transactions). That’s why this chain is “Robinhood-branded.” Details are in the table below.

There’s also an economic detail worth knowing: as an Arbitrum Orbit chain that does not settle to Arbitrum One, Robinhood Chain applies to the Arbitrum Expansion Program, which requires returning 10% of net protocol (sequencer) revenue back to the Arbitrum ecosystem: 8% into the ArbitrumDAO treasury, and 2% to the Developer Guild. This isn’t trivia: on 9 July, this chain’s daily transaction volume was $568 million, and ARB jumped 19% that day purely on the basis of that revenue-sharing logic. The remaining 90% of revenue, and control of the entire tech stack, belongs to Robinhood.

▲ Robinhood Chain architecture

This chain isn’t built by one company. Key partners and their roles:

Two kinds of dollars: USDG and USDe

Two types of “dollars” run on this chain with different functions—you must not mix them up.

USDG is this chain’s in-house dollar. It’s a fiat-collateralized stablecoin issued by Paxos, launched in late 2024. It is 1:1 backed by dollars held at DBS Bank and short-term US Treasuries. On Robinhood Chain, it’s settlement and pricing collateral: the deposit unit for savings, the margin and pricing asset for Lighter perps, and the “dollars” that flow between Wallet and chain. Gas is still paid in ETH, so USDG is money—not fuel—and it isn’t exclusive to this chain (native issuance on Ethereum, Solana, Ink, and X Layer, interoperable via the LayerZero standard).

Why Robinhood pushes it: Robinhood is a founding member of the Global Dollar Network, which returns about 97% of reserve yield to partners who drive adoption. Set USDG as the default dollar of its own chain, and Robinhood earns not only trading fees but also the yield from idle float. Economically and in default usage, USDG is the closest thing to a native stablecoin this chain has—despite being multi-chain technically.

USDe is a yield and collateral dollar, not a settlement dollar. It’s Ethena’s synthetic dollar, supported by crypto collateral hedged against short positions (a delta-neutral basis position). It’s not fiat in a bank; by design it comes with yield. It’s the largest token by on-chain value, but that number is mostly driven by cooperation and collateral, not by natural retail money flows. Ethena is the partner; USDe is bridged onto the chain and deposited into Robinhood’s savings vault as one of the collateral markets producing about 7% returns. So USDe’s large figure reflects it being introduced to support savings, not people using it as a daily currency. One sentence: USDe is the yield engine; USDG is the current account.

Three product layers: App, Chain, Wallet

Now that the chain and money are covered, let’s look at what’s different about the three user entry points. They’re often confused, but actually they are three different layers.

How they connect: Wallet is the user layer, Chain is the settlement and infrastructure layer, and the brokerage App is a separate custodial world (mainly acting as a fiat on-ramp). USDG flows between them.

Who can use what:

Perps: two venues, two different machines

There is no single “Robinhood perps.” Two on-chain venues do two jobs: Lighter does crypto perps, and Arcus does stock and RWA perps—easy to mix up. This section clarifies the two venues, Lighter’s operating mechanism, and how they differ. (Robinhood also has a custodial, compliant perp product inside its EU brokerage app—off-chain, not on the chain, and outside the scope of this article.)

Two venues

How Robinhood and Lighter collaborate across two chains

This is the easiest part to get wrong. Lighter isn’t a pool on Robinhood Chain; it’s another chain. The two collaborate via cross-chain collateral. You can think of two banks that signed wire protocols: your money is custodied with one (Robinhood Chain), while trading happens at the other (Lighter). Both sides keep the ledger synchronized via messages.

▲ Robinhood and Lighter dual-chain collaboration

How to read this figure:

2. Lighter is a CLOB perpetual DEX, not an AMM—no swap pools. Your counterparty is either a maker order or a taker order, or a LLP (Lighter Liquidity Provider) vault, which provides two-sided quotes and backs up settlement.

3. Users deposit USDG from Wallet as margin. According to Robinhood docs, USDG is transferred on Robinhood Chain and locked into the Lighter Relayer smart contract; Lighter then records an equivalent amount as margin on the trading interface. Wallet is self-custodial; Robinhood is just the entry point, not the custodian.

4. Matching and settlement run on Lighter’s own zk rollup—an independent execution layer: an off-chain sequencer plus a zk prover that do market-making real-time quotes.

5. LayerZero is the cross-chain messaging layer that keeps the two environments in sync.

6. Lighter sends the final state root and zk validity proofs back to Ethereum L1, and only after proof verification passes is the state finalized.

On the key liquidity details, Lighter has confirmed it directly. In a 2 July 2026 X post, Lighter explains that Robinhood integration is a Lighter Domain—an independent Lighter instance, with execution, sequencing, block space, and liquidity all separated. This isolation is intentional, so different markets can serve different ecosystems, partners, and regulatory requirements.

So Robinhood’s USDG order book is a truly independent instance and independent liquidity pool, not Lighter’s USDC main order book. Its depth must be built from zero by market makers on that instance (zero fees, 90-day gas subsidies, 2x points, and $11 million $LIT —that’s exactly what they’re doing), rather than inheriting Lighter’s roughly $39 billion main-pool depth. Robinhood users can’t access the main-pool depth. DefiLlama data also supports this: after the news was released, trading volume on Lighter main pool barely moved, while the token price rose.

Trade path and counterparties. Perps in Wallet can only place market orders, so Robinhood users are always taking. Your market order enters the Lighter Domain’s matching engine and, using price-time priority, takes the best standing limit orders. The maker side is professional market-making institutions and Lighter’s own liquidity vault: LLPs that provide two-sided quotes and backstop settlement, and XLPs (Experimental Liquidity Provider) used for pre-market and RWA. Because liquidity within a Lighter Domain is isolated from each other, these market makers are configured specifically for the USDG instance and not shared from the USDC main pool. Note: Robinhood’s in-house market maker Pleiades serves the spot stock-token AMM, not the Lighter perpetual book. So your counterparty is always a market maker or an LLP—never another Robinhood retail user—and Lighter doesn’t act as a market maker either. Your USDG stays locked in the Lighter Relayer contract on Robinhood Chain, while the positions live on the Lighter instance.

Lighter vs Arcus

Both are Robinhood-family perpetual venues, but the structures are completely different.

What’s under stock tokens

Plain-English version: stock tokens are an IOU issued by the Robinhood Jersey entity (a promissory note / payment commitment, not the asset itself). The price tracks the corresponding stock. What you get is price exposure, not shares. Details and caveats are as follows.

Robinhood Stock Token is a tokenized debt security issued by Robinhood Assets (Jersey) Limited (RHJ). Legally, it’s a linked debt instrument, similar to traditional ETNs (exchange-traded notes). Token holders only receive the economic exposure to the corresponding stocks, including price fluctuations and related economic benefits. They have no legal or beneficial ownership of actual shares and no shareholder rights such as voting. In simple terms: when you buy an AAPL token, you’re actually holding a debt note issued by RHJ, the Jersey entity. You are a creditor of RHJ, not a shareholder of Apple.

Robinhood’s design goal is that each stock token is roughly 1:1 hedged with US-listed stocks or ETFs held by the associated entity in custody. But the token itself is still a claim against RHJ—not a direct representation of the underlying shares or a trust beneficial interest. Official documentation does claim stock tokens are “1:1 supported” (underlying shares held by a U.S.-licensed broker-dealer/custodian, with custody and brokerage by Alpaca under a series), but that’s only a one-sided statement by the issuer: there is no publicly available proof-of-reserves (public verification that the underlying assets actually exist), nor regular third-party attestations to confirm it. Third parties generally describe it as “nominally 1:1 backed.” In addition, for private company tokens, the official documentation explicitly states they are not 1:1, and they are not redeemable. This is the key ownership-layer caveat: whether the token can ultimately be redeemed depends largely on RHJ’s credit and risk controls as the issuer.

Handling of dividends and corporate actions is also different from traditional stocks. It does not distribute cash dividends directly; instead, it adjusts on-chain multipliers under the ERC-8056 standard. When the underlying stock pays dividends or undergoes a stock split, the system increases the economic share ratio per token, so the intrinsic value of the tokens updates automatically. A user’s token balance remains unchanged until redemption. Keep the chain clean; keep the economics continuous.

Overall, the underlying structure of stock tokens is a hybrid: “RHJ debt obligations + Robinhood entities’ custody of the underlying stocks hedged.” This design delivers all ERC-20 features (free transferability, wallet interoperability, DeFi composability) and enables efficient issuance and global distribution within the regulatory framework. The cost is that users bear not pure stock risk, but a composite exposure layered with the issuer’s credit risk. Compared with RWA products that simply represent direct holdings or fully isolated custodian models, this approach has clear advantages in liquidity and innovation, but it embeds credit and operational risks that users must evaluate themselves.

How Robinhood compares to other major stock tokens

Robinhood is a late entrant—this market already exists. On-chain tokenized stocks are about $1.2 billion. Two issuers dominate: Ondo Global Markets (about half the market; first to surpass $1 billion TVL, with 260+ stocks) and xStocks by Backed Finance (most holders, about 162k versus Ondo’s about 70k; cumulative trading volume $25 billion+; routes via Kraken, Bybit, and Solana DeFi).

When Robinhood entered, its share was nearly zero (on-chain stock TVL about $10.7 million). But it had distribution weapons the others didn’t: a consumer-grade app covering 120+ countries, plus its own chain.

CEX players are entering too—Binance is the one to watch most. In June 2026, it launched zero-commission trading of 7000+ U.S. stocks and ETFs for non-U.S. users, and then previewed bStocks: minting users’ holdings into BNB Chain-backed 1:1 tokens, trading 24/7. The first batch includes Nvidia, Tesla, Circle, Micron, and SanDisk. The capital flows already tell the story: Binance’s top 30 days incremental tokenized stock funding was $300 million+, while xStocks was $33 million over the same period, and Robinhood was $13 million.

One-sentence summary: among the three on-chain players, Robinhood’s underlying structure is the weakest (a debt claim without proof of reserves, versus the other two’s 1:1 custody model), but it has the strongest consumer distribution. It bets that the app-funnel and its own chain matter more than legal purity. Ondo and Backed bet the opposite. Binance is the variable: it’s playing the same distribution card as Robinhood, but its funnel is much larger—bStocks funding flows have already outperformed everyone. Whoever wins the bet will be answered by the RWA TVL numbers in the next two quarters.

Risks, open questions, and conclusion

2. Perps only launched halfway. Arcus’s RWA and stock perps are still queued up; on day one, only Lighter crypto perps are available.

3. Perpetual liquidity started from zero. The Lighter integration is a dedicated USDG order-book instance; depth must be fed with incentives and cannot inherit Lighter’s USDC main book. Thin depth in the early stage is a real, concrete risk.

4. The base layer of stock tokens. Roughly 1:1 “hedging” is only a statement; there’s no confirmed proof-of-reserves. And for private company tokens, it’s explicitly not 1:1 and not redeemable.

5. Geographic restrictions. Perps and stock tokens exclude the United States; Lighter perps also exclude the UK, Canada, Switzerland, the UAE, and Singapore—cutting off the largest retail market.

6. Centralization. A single sequencer whose operator isn’t disclosed, an in-house market maker (Pleiades), and no published decentralization roadmap.

7. Savings yield. About 7% annualized is floating and driven by demand, sourced from loan interests in Spark, Ethena, and Maple markets. Higher yield means higher risk; insurance only covers loopholes, not depegs and market volatility.

8. Quality of activity. Early trading volume and most users are largely meme rotation; RWA assets are only about 4% of TVL. The bullish logic requires meme liquidity to convert into stock token and savings balances—but that conversion has not been proven. Base basically didn’t achieve it back then.

Our conclusion: this infrastructure ledger already adds up. Robinhood keeps 90% of chain revenues, controls the sequencer, and earns USDG float yield. Cold-start was also solved by the meme wave amplified directly by its own CEO. The remaining suspense is whether this chain becomes an RWA chain or just a casino with a brokerage brand. Three things will answer it: (1) whether RWA TVL can rise from about 4% to meaningful scale—an early indicator is whether stock-token DeFi like Arrow can keep compounding; (2) after zero fees and point incentives decline, whether Lighter’s USDG order-book instance can keep real depth; (3) whether Robinhood will provide proof-of-reserves for stock tokens—because compared with Ondo and Backed’s 1:1 custody model, the debt-note structure is its most fragile side wing. This chain has no token, so any view can only be expressed through the ecosystem: ARB (a cut of chain revenue), Lighter, and early ecosystem tokens.