If you're reading this and still haven't signed up, click the subscribe button below!

This week, Simon Taylor and Cuy Sheffield are joined by David Marcus — CEO and Co-Founder of Lightspark, former CEO of PayPal, and the man who led Meta's Libra project before regulators killed it with fire. If you want to understand why the stablecoin infrastructure moment feels different now than it did in 2019, David is probably the most qualified person alive to explain it. He's been on both sides of that wall. 

The clearest message from this episode: the infrastructure for global dollar accounts is now actually built. What took Libra years of political capital and ultimately failed is now being assembled quietly, piece by piece, by platforms that don't need to ask permission. David Marcus spent most of the conversation on a single question — who's going to own the economics of that infrastructure, and who's going to give it away? 

🎙️ Listen to the latest episode of Tokenized here, or 📷 watch on YouTube.

David's core argument with Lightspark Grid Global Accounts is straightforward, but the implications are easy to underestimate. Every major platform — Uber, Airbnb, creator platforms, marketplaces — is currently a tenant on someone else's payment rails. The moment they pay out to a driver in Indonesia or a seller in Brazil, they lose the economic value of those deposits. Everything that happens next belongs to someone else.

Grid changes that. It lets platforms offer their own branded dollar accounts to their users — with a Visa debit card, real-time off-ramps across 65 countries, and now AI agent delegation built in. But the distinction David kept coming back to is important: everyone else building in this space (Stripe, Ramp) is building accounts they sell to platforms. Lightspark is building infrastructure for platforms to build their own accounts under their own brand.

As David put it: "All of those platforms have been tenants on someone else's payments rails and networks. And I feel like, especially on the account side, that's true."

The economic logic follows directly. Platforms like Uber or a creator marketplace sit at the center of economic life for their users. They have the best transaction data. Pre-AI, leaching that data to edge banks was just the cost of doing business. David's argument now: "In the AI world, where you can actually mine that data and build underwriting models like you couldn't do five years ago or a year ago, is actually negligent." You have the data to underwrite credit, build financial products, and generate — his words — "a whole new multi-billion dollar business." Giving that to banks at the edges is a choice, not a constraint.

Meta launched stablecoin creator payouts this week — live in Colombia and the Philippines, using Stripe Link as the receiving wallet, on Tempo, Polygon, and Solana, with plans to expand to 160 countries. Five years ago, this would have been sirens in the streets. It got almost no coverage. Simon flagged that silence as a signal in itself. David agreed, and laid out exactly why the conditions are different now.

Three things had to exist that didn't exist even a year ago:

The strategic question for platforms is what comes next. Cuy framed it well: the creator economy could be 100x bigger with AI. One-person record labels. One-person film studios. A creator who today has basic payout needs could become a genuine enterprise — one person and five agents spending on image models, video models, and data. The platform that owns the financial layer for that enterprise wins. As Cuy put it: "That's a pretty big business."

Simon drew the embedded finance parallel: early embedded finance tried to put cards inside global platforms and it never caught fire, because you need global infrastructure and banking licenses in every market. Stablecoins solve that. No industrial loan charter required.

The agent piece of this conversation deserves its own section, because it's where the product is actually going.

Lightspark built what David calls "bonded scope delegation" — you bond an AI agent to a Grid Global Account with scoped permissions that define exactly what it can and can't do with your money. Multiple agent scopes per account. David has been running OpenClaw with access to a Grid account for the past month.

Simon's framing was sharp: "The CLI is the new UI." Fintech built great mobile apps. Stripe came along and made APIs the developer interface. CLIs are now doing that for agents. He pointed to Stripe's own data: they launched their CLI seven years ago, it bumped along quietly, and then in December 2025 — vertical acceleration. Agents are now using it. The Collisons called January 1, 2026 "the beginning of the singularity."

Cuy's concern — and it's a real one — is the accountability gap. A world where agents have wallets that aren't tied to any human creates bad outcomes. The right model, which both Cuy and David agreed on, is agents as delegated actors under human control. KYC'd at the human level, with scoped permissions flowing down. The FIDO Alliance announced standards for agent trusted delegation just two days before this episode recorded. Mercury launched agent accounts with scoped permissions the same week.

This isn't theoretical anymore. The infrastructure is being built in real time.

Western Union announced on their earnings call that their stablecoin USDPT launches in May 2026 — this month. The rollout follows a pattern the group recognized immediately:

The bigger point Cuy made is about the receive side. Remittance companies have historically made almost nothing from recipients. The recipient picks up cash and walks out of the store. WU has minimal data on them, no products for them after the transaction. A WU wallet where funds land, with a card attached, where recipients can hold dollars — that's a neobank play for emerging markets. As Cuy put it: "Remittance companies can be these emerging market neobanks using stablecoin infrastructure."

David's take on the stablecoin branding question was direct: nobody wants USDPT. Nobody wants your branded coin. "No one cares about your dollar. People want a dollar." The right move is to offer dollar balances, make them maximally fungible — send to USDC on Solana, fine, a dollar is a dollar — and collect the yield from the reserve while the balance sits in your account. Brand the account, not the coin.

This was the sharpest analytical moment in the episode, and it's one the bank lobby may not have fully worked through.

Under the current regulations as currently constructed, you can't pay yield to stablecoin holders in the US. But if you're issuing dollar stablecoins outside the US to customers outside the US, you can. Cuy pushed on the second-order effects: what happens to capital flows if yield is available on dollar stablecoins outside the US but not inside it? David's answer was direct: "People will want offshore dollar accounts to receive yield on their stablecoin deposits — which would be kind of like probably one of the biggest self-goals of the banks here in the US."

Simon drew the Eurodollar parallel directly. Midland Bank in London in the 1960s offered higher yield on dollars than US banks could under the regulations of the time. Dollars flooded into London, then everywhere around the world. The Eurodollar market was born. Now, as Simon put it: "We have the Eurodollar for everybody else." Not just for banks — for anyone with a stablecoin wallet.

The yield debate in Washington is focused on whether stablecoin holders in the US can earn interest. The quieter question is what happens when the answer is no domestically but yes everywhere else. That's a capital flow question, not just a product question.

The final story: Securitize and Computershare — the world's largest stock transfer agent — announced an agreement to enable tokenized shares for US issuers. Share issuers can now include issuer-sponsored tokens as part of their issued capital alongside existing shares in the direct registration system. Computershare processes all corporate actions.

Simon's framing was the most useful one here. Transfer agents bridge the gap between a blockchain record and financial regulation. Under SEC Rule 17ad-10, you have the legal share — not a price-tracking derivative, not a claim on a claim on an SPV. That legal ownership is enforceable in court. A lot of tokenized assets currently on-chain don't have that. This deal starts to change that.

David's view was honest: this isn't Lightspark's focus, but they'd be a "second-order beneficiary" — more tokenized assets means more things Grid accounts can buy and sell directly.

Cuy's working capital angle is worth sitting with. If tokenized money market funds trade 24/7 against stablecoins, and AI treasury agents can automatically hold funds in the highest-yielding asset while maintaining required liquidity — you end up with a lot less idle cash. As Cuy put it: "I really struggle in the yield debate of entering an era of hyper optimization with AI and tokenized securities... you're just going to have a lot less cash at rest, sitting there, not earning yield." The stablecoin yield debate may be somewhat moot by the time the infrastructure catches up.

What this episode marks, more than any single product announcement, is the moment when the Libra thesis — a global, open internet for money — stopped being a regulatory casualty and started being a quiet infrastructure build. David Marcus is doing in 2026 what Libra tried to do in 2019, but without the press conference, without the congressional hearings, and without asking anyone's permission. The rails are being laid one corridor, one platform, one agent delegation at a time.

The question he kept returning to isn't whether this gets built. It's whether platforms will own the economics when it is — or hand them to someone else, the way they've always done.

Thanks so much for reading the Tokenized Newsletter!

Please share this edition or share it with your colleagues if you enjoyed it!

Disclaimers