Author: Fintech Brainfood

Translation: Deep Tide TechFlow

Deep Tide Introduction: Robinhood launches a DeFi yield product with a 7% APY, stock tokens, and commodity futures in London, behind which is a hybrid strategy of half real yield and half marketing subsidy. When yield products, stock tokens, and collateralized lending come together, Robinhood is building a prototype of a decentralized wholesale broker, which means new liquidity play for investors and an upgrade in direct competition with Coinbase for the industry.

Robinhood launches 7% APY yield product, mainnet, and futures

At the Greenwich National Maritime Museum in London, Vlad Tenev took the stage wearing a stylish pirate-inspired coat to announce the launch of Robinhood Earn—a DeFi yield product with a 7% APY. Over 120 countries of stock tokens, 24/7 trading, which can be used as collateral in DeFi lending pools. Additionally, it launched commodity, ETF, and forex perpetual futures (gold, oil, euro/dollar) in Europe, with up to 10 times leverage.

The 7% yield product is quite interesting. Where does the yield come from? A Morpho treasury orchestrated by Steakhouse invests in spUSDG, USDe, and SyrupUSDG, providing over-collateralized lending to institutions. Half of the yield comes from native income, and half comes from incentives.

In other words, about 3.5% of the yield is currently marketing expenditure. But as activities expand and USDG trading volume increases, it earns more floating deposits (because it buys more government bonds). USDG will share floating deposits with distributors, in this case, Robinhood. If all goes well, the other 3.5% should start coming from the stablecoin itself.

This is targeted at all 27 million users, not just the DeFi crowd. This 7% yield product will be available to its 27 million users through the core Robinhood app (supported by the embedded wallet provided by Privy).

It is also insured by Lloyd's of London, covering DeFi risks. So it can deal with hacking or smart contract vulnerabilities (which do happen!).

Stock tokens are super interesting. Although technically not stocks and do not represent ownership of benefits. Robinhood is a registered broker-dealer, so it ensures a 1:1 peg with these stocks. It also supports securities lending. Users can take this stock token to secure loans, like for a down payment on a house.

Now combine stock tokens and yield products. If you have a very large "lending to traders" capability, you can earn native yield, plus earn yield through your stablecoin, and now you can lend to those who want to use stock tokens as collateral for loans. Ultimately, what you get looks very much like a decentralized wholesale broker. Very interesting.

Their own chain also has economic benefits. They keep all transaction fees instead of paying them out. These fees are negligible on most modern chains.

The comparison with Coinbase is obvious. Coinbase announced the upcoming launch of "stock tokens," with USDC offering lower yields, and it provides futures through premium and institutional platforms, but not as a core consumer product. Robinhood has more control over stablecoins with USDG, and as a seasoned licensed broker, has greater scale and experience in acquiring stocks and ETFs to support its products.

With a market cap approaching $100 billion, can Robinhood become a trillion-dollar financial company? There are no trillion-dollar financial companies. Not JPMorgan, not Visa, not BlackRock. This company expands its geography and products, entering new areas like prediction markets and AI faster than anyone else. It could be said to be a financial company that needs catching up (or on par with Revolut and Nubank).

Damn, Robinhood's marketing is too clever. They host events for the most active users, with television-grade camera equipment, a whole theme, and then announce possibly ten different products.

We recorded a complete interview with Johann at their event site, which you can find on Tokenized.

Plaid reportedly exploring IPO in the US

According to Bloomberg, the company is said to have had preliminary discussions with banks regarding a potential public listing. Plaid was valued at $13.4 billion during the fintech boom in 2021, then reset to $6.1 billion in 2025. Earlier this year, as investor sentiment towards fintech began to recover, it rebounded to an $8 billion valuation. If Plaid continues to move forward, it could become one of the most important fintech IPOs of this year.

Plaid successfully jumped on the AI bandwagon. They launched integrations with ChatGPT and Perplexity, allowing people to access and view their account data. This turns what was originally a pipeline into a secure channel for AI agents. This could be much larger than for applications or fintech companies.

They are also doing interesting things with custom transformer models on LLM. I spoke with them last week about their new sequence-based models that are more effective than previous machine learning attempts at classifying payments, predicting fraud, or missing repayments. This brings more loans or less fraud to customers, the products get better, and revenue grows.

Plaid has successfully diversified its revenue. Its lending and fraud products now generate significant income for the company. The outcome of "bank payments" remains unclear, but if it happens, Plaid is also well-positioned in this regard.

Wherever Plaid goes, fintech follows. While Stripe, Robinhood, and Revolut are much larger companies, Plaid is the industry bellwether. They acquired This Week in Fintech. In many ways, Plaid is the core of "fintech" from 2019 to now. When they go public, it'll be graduation day. I just hope the window stays open long enough for them to catch up. There is some tight sentiment in the market at the moment.

Erebor targets an $8 billion valuation, deposits reach $4 billion

According to Bloomberg, Erebor's deposits reached $4.5 billion. Their report from March showed $1.1 billion. Nearly quadruple growth in a single quarter for a bank that just received a full national charter in February. They are now negotiating to raise funds at a valuation exceeding $8 billion, up from $4.35 billion in December.

Founded by Palmer Luckey. Backed by Founders Fund, 8VC, and Lux Capital. They took about 9 months to receive a full charter from OCC. Paxos, Ripple, Circle, and Stripe have all applied for trust charters. None of them operate full deposit banks.

This quarter, they added about 400 clients, with deposits growing by approximately $3 billion. Even if every dollar came from new clients, the average account is over $7 million. A few very large depositors, with almost all exceeding the FDIC limit. Does that sound familiar?

The demand for crypto collateralized lending is below their expectations. The winning product is the boring one: a deposit account that remains open. It turns out that a "prevent being strangled" bank may be a key feature for a certain type of client.

Luckey specifically stated that none of this quarter's growth came from his own company. When the CEO preemptively rebuffs this criticism, you know they've really hit the mark for the first time.

Rapid growth, concentrated, uninsured, correlated deposits are exactly what SVB looks like. The current distinction is: according to the first-quarter report, Erebor holds zero loans. Its operations are closer to a narrow bank rather than a lending institution. And it achieved profitability without those loans.

Meta wants to purchase Kalshi

Before instructing employees to build an independent prediction market application, Meta CEO Mark Zuckerberg proposed acquiring Kalshi. NPR reported there was a meeting, but negotiations never progressed beyond that stage. In June 2025, Kalshi and Polymarket were trading about $28 billion per month. A year later, according to The Block, these sites' monthly trading volume approached $220 billion, primarily driven by sports-related betting.

While acquisition talks never advanced, Meta did form a partnership with Kalshi in March, allowing for easy integration of the Kalshi market into the Meta social media app Threads.

Meta appears to excel in M&A more than R&D. WhatsApp and Instagram might be the greatest acquisitions of all time. But considering the questionable AI and "metaverse" investments, and the failed "Facebook Workplace," you can't help but wonder what the next step for this tech giant is, beyond better ad monetization.

Last week I wrote about Meta trying payments again in India. After Facebook Credits, Libra, and countless other attempts, perhaps by hiring the CRED founder, they can add another major monetization channel.

4 companies

1. Novaquant - AI decision governance layer

Novaquant sits between internal systems and AI models, observing every interaction to ensure these interactions are auditable and compliant with policy. Policies, decisions, and information flows can then be compiled into a complete decision history.

If you're regulated and sending data to models, you need this. I first wrote about this idea when I crafted a supervisory framework for Sardine at the beginning of 2025. Companies have hundreds or even thousands of policies; the connections between these policies, your operational processes, and what the models do can be extremely challenging to trace. It makes sense to embed a layer between them. However, I wonder if this is a feature of a larger "enterprise AI control plane" that's emerging rather than a product?

2. F2 - AI-native trade execution in private markets

F2 reads all historical memos, models, and reviews to construct a structured, queryable history of trades created by each company. These insights can then be used to filter new trades, execute deep due diligence in preparation for investment committees, construct models in Excel, and monitor entire portfolios.

F2 specializes in private credit. Companies like Hebbia and Rogo are designed for broader tasks and workflows. F2's differentiation comes from its deep understanding of the credit market, which is embedded into the product. Time will tell if this means they become the preferred choice.

3. Wealthreach - AI content and website engine for RIAs

Wealthreach helps RIAs create custom websites without templates and with strong SEO rankings. Companies can continue to edit and maintain the website, change copy, publish blog posts, using the platform as a blend of a dedicated SEO agency and website agency. Active websites perform better in SEO audits. The platform is designed to create new content in the tone of RIAs based on the input you provide, and to redirect visitors who visit your site but do not fill out forms through email and LinkedIn outreach. Active websites start at $300/month, SEO for $500/month, and retargeting at $500/month.

AI is creating businesses that operate in niche sectors. Imagine building this company five years ago before AI appeared? It would have been challenging. With AI, generating pages and copy is not difficult. Making it sound like an RIA and perform well in that niche's SEO is key.

4. Trustapp - Custody for agents

Trustapp started as an online custody solution, addressing risks in markets that frequently experience high levels of fraud, with fake sellers or bad buyers. This leads to high chargeback rates and high costs. They have since shifted their top-level information to focus more on AI agents and processors for any large transactions or B2B payments. The world needs processors specialized in handling complex transactions. I wonder if being that is the best outcome. They also focus on helping these merchants better enable AI to discover them.

I definitely see two or three pitches about making your "eCommerce AI agent-ready." But auto and B2B markets also need agent cataloging and discovery. This vertical focus may be the wedge for Trustapp.

Recommended Reads

1. A frontier without an ecosystem is unstable

Article link: https://x.com/satyanadella/status/2066182223213293753?utm_campaign=where-robinhood-s-7-yield-comes-from&utm_medium=referral&utm_source=www.fintechbrainfood.com

Microsoft CEO Satya Nadella believes that the value of AI will not come from "who has the best models," but from "who can build learning loops around various models." Can you bring in all the data, your context, and private IP you don’t want to share, and push them to different models based on tasks? The skill for the next decade will be building private evaluation and private reinforcement learning environments to help your AI get better every time it executes what you do.

This is clearly Satya's pitch to have enterprise tools, control planes, and contexts that help these companies build private evaluation and reinforcement learning models. Considering the offerings in the Microsoft suite, they have a reasonable chance. I also strongly agree with this pitch and vision. As we wrote in the operational modeling handbook, tools, control planes, and your ability to create evaluations and private models are the key differentiators between AI owners and non-owners.

That’s all folks 👋

Remember, if you like this content, please tell all your fintech friends to check it out and hit the subscribe button :）

Want more? I also run the Tokenized podcast and newsletter.

(1) All opinions and thoughts expressed here are the author's personal opinions and do not represent the views of any of their employers or employees.

(2) Out of all the companies or assets mentioned, those where the author has personal and/or financial interests are marked with *. The above content does not constitute investment advice, and you should seek independent advice before making any investment decisions.

(3) Any companies mentioned are simply the first that come to mind and are used for illustrative purposes only.

(4) The research team has not rigorously verified this content. Please do not take it as absolute truth—this is a strongly held weak opinion.

(5) Quotes may be missing; I’ve made an effort to cite but will always strive to update and correct the online version. If I quoted you but did so incorrectly, please contact me.