Foundations of Intelligent Money
“Do you wish to be great? Then begin by being. Do you desire to construct a vast and lofty fabric? Think first about the foundations of humility. The higher your structure is to be, the deeper must be its foundation.”
— Sermon 19 on the New Testament, Saint Augustine, Christian theologian, 354–430.
Our work as venture capital investors is heavily weighted towards consuming large amounts of data to help us uncover patterns, trends, entrepreneurs, startups ahead of the curve. As an example, I read on average over 1500 emails per quarter (each being a newsletter I subscribe to or containing a research report I downloaded). On rare occasions, the themes we chisel, track and refine over time become reality. I strongly believe that what we are witnessing today — in what we read, and in the conversations we participate in across the marketplace — fits this convergence. Namely, what we have already written about — digital wallets, AI and tokenization melding together (see our November 2025 GP letter, where we first framed this trifecta and the Onchain financial services world it leads to) — stopped being a thesis and became reality. Each of the three legs has production deployments, statutes, consortiums forming with dates, Dollar or Euro figures attached.
In other words, our MGV thesis — that all assets will be tokenized and live onchain — is developing into a living organism. Machines are becoming economic actors. Agents are getting cards, then interacting within and across protocols. Payments are becoming productive. This is what we mean by intelligent agents. We are seeing the economics of money slowly moving from issuers to distribution owners — witness the consortium stablecoin wars. At the same time, assets are moving onchain at institutional grade. Tokenized money market funds, short commercial paper and deposits are moving first. Equity tokenization aimed at the retail trade is following. All of this is supported and serviced by named banks and licensed operators in top grade jurisdictions.
Simply put, money and assets are becoming machine readable. Access to money and assets is becoming intelligent, powered by AI. Meanwhile, access to intelligence itself is becoming a monetary flow in its own right — compute and AI tokens are bought, metered and budgeted like any other financial resource. Crucially, all of this is happening within regulated, jurisdictionally compliant frameworks.
Picking through the various individual narrative arcs, we are seeing intelligent agents becoming economic actors with credentials — a future we first sketched back in August 2023 in “Time to Recalibrate”, our early piece on autonomous AI agents. We saw Ramp issuing AI agents tokenized, single-use and spend capped corporate cards, while Stripe countered with Shared Payment Tokens and USDC rails. Stripe then launched its Machine Payments Protocol which quickly became a two-protocol race with Coinbase-backed x402 for the “internet’s payment layer”. Mastercard and Google went the open-source protocol way while Visa positioned itself as the identity and authorization rail for agentic commerce. Finally, Stripe demoed a fully autonomous purchase process while Airwallex raised a mega H round to go after “autonomous finance and agentic commerce”. I will also mention Coinbase which is quickly repositioning itself from an exchange to the Financial OS stack for AI agents, and ChatGPT which is linking to bank accounts via Plaid. All actors, incumbents and insurgents alike have been busy. So have we: our investment in Paygentic — billing and payments infrastructure built specifically for AI-native, agent-driven businesses — is a direct expression of this intelligent agents thesis. Even if the game is still very young, it is afoot. One thing to keep an eye on is how fraud and the fraud stack will be built and taken care of. As of now little is being developed regarding such a major issue. If systems of record are winning the SaaS wars (more on that later), then intelligent orchestration layers as inter-operational middleware may win the intelligent agents wars.
Moving on to intelligent transactions we saw and continue to see equally frenetic activity. Mastercard to acquire BVNK, while Zelle is moving towards stablecoins and The Clearing House (both in the US) is moving to tokenized deposits. Stablecoin infrastructure, whether handling stablecoins or tokens, is becoming the rail of rails when it comes to money. This is the quiet transformation of financial infrastructure we chronicled in July 2025, now playing out at full volume. More recently, and this is a whopper of a story, roughly 140 companies including Visa, Mastercard, Stripe, Coinbase, BlackRock and many banks launched a consortium stablecoin that shares yield with distributors around OpenUSD. This piece of news moved the markets so much that JPMorgan and BoA may be in talks and working towards a rival stack. Transactions are becoming intelligent fast. This means that float economics will migrate from issuers to distribution owners and operators as issuance will fast commoditize. What I find fascinating here with intelligent transactions is that, similar to every other major technology innovation wave, at least 90% of the innovation in stablecoins towards intelligent transactions will be doomed to fail, yet the remaining 10% will rule the world. For us at MGV these intelligent transactions reflect a switch to defensible moats such as compliance, orchestration, fraud prevention, AML tooling, identity verification. Our filter for separating the meaningful 10% from the doomed 90% is the ten moats framework we detailed in our April 2026 GP letter: we underwrite regulatory and compliance positioning, transaction embedding and systems of record, and we avoid the layers that AI and issuance economics will commoditize. We see the same chink in the armor so to speak between intelligent transactions and intelligent agents.
Intelligent assets, nowhere to be seen but in the eye of the dreamer, are now acquiring a name and well-deserved buzz. Picture Revolut, Robinhood, all major crypto exchanges. They are now offering trading on tokenized assets, and in many cases volume has overtaken defi trading volume. This is a thesis we have backed since 2023, when we co-led the seed round of Ctrl Alt, the first UK-regulated alternative asset tokenisation platform, and doubled down on in 2024 with One Trading, the first MiFID II-compliant digital asset perpetual futures trading venue in the EU. SG Forge and Euroclear are moving to explore MiCA compliant stablecoin settlement of tokenized commercial paper while deposits are themselves acquiring a proper tokenized stack — the Canton Network consortium (Goldman, DTCC, BlackRock, Nasdaq among others) on one side, and ZKsync’s Prividium, chosen by five US regional banks with over $600 billion in combined assets for FDIC-insured tokenized deposits, on the other. The SEC, the CFTC are now much more friendly, clarity is coming to the EU, and experimentation is running apace with most of the main protocols. Again, we are early in the game but big names on both sides of the Atlantic are laying down their cards. We are even seeing governance solutions reaching tokenized equities (with voting rights moving onchain). One note of caution, even if tokenization is institutionalizing, it remains to be seen whether token holders will see automatic benefits and where value will flow.
I would be remiss if I did not mention Intelligence (and I use the word carefully, knowing there is nothing really humanly and intelligently creative about AI). We did discuss AI in previous GP letters — most recently in our April 2026 GP letter on AI, software and the future of moats —, so this will be an augmentation of sorts. The market repriced many publicly traded equities while most prominent newsletters from Sequoia to Bessemer called for the death of SaaS. It seems that repricing is not only happening at the equity price level for publicly traded assets, but also at the business model level. Take note of Sequoia which called for services being the new software. Token spend and token budgets, token integration, token ROI (not the tokenized asset kind, but the tokens used with Claude or ChatGPT) have taken on a life of their own. AI user costs are now rivaling cloud computing bills (our investment in Cloud Capital c.2 years ago – a start-up financialising compute and AI infrastructure spend – was made in anticipation of this developing and now accelerating trend).
From a fintech point of view I will note that the worlds of underwriting, pricing risk, calculating spreads, going long or short, stock picking, investment decisions, financial analysis will never fully be the same. Even if AI adoption is still early, we are seeing job displacement (jobs lost while many more are created) as opposed to job destruction. It seems to us that underwriting agility, data ownership and positioning across a value chain are being repriced higher while switching costs, which were an important moat pre-AI, are depreciating faster. This is vindication for the ten moats analysis in that same April letter, which fleshed out new defensible moats such as data ownership, licensing in financial services as positioning, closeness to monetary transactions and systems of record.
I will end this GP letter with four points, each of which carry their own promises and risks.
First, sovereignty has invited itself into tech, and therefore into financial services and fintech too. The volatility we see coming from the US (embargo on an LLM model, tariffs, wars of choice in the Middle East, belligerent rhetoric towards traditional allies) is having an impact on AI — every country or bloc of countries is treating AI as a sovereign asset — and as a result money is now being treated similarly. Stablecoins have already become an instrument of geopolitical might, something the EU is slowly coming to terms with. For us at MGV this means that jurisdiction and licensing is now a product. Regulatory domicile is even more important for value creation than before and will continue to be so.
Second, fintech is rushing to acquire licenses (financial licenses for market operators, payment operators, which is nothing new) and charters or bank licenses (which is new). We see this unfolding in the US, and we are hopeful that with a change of the guard at the ECB in the coming months we will see this in the EU. Circle, Klarna, Kraken, Revolut, Sony, X Money… all want to become a bank at some point. This trend will not go away and is a logical outcome of what we are seeing in intelligent transactions, assets and agents all supported and augmented by AI. An AI tool, however mighty, will never be able to disintermediate a bank license. We examined this dynamic — banks’ enduring regulatory moat, and their parallel risk of marginalization — in our February 2026 GP letter.
Third, compute and cloud are the foundation and Achilles heel of the entire edifice, with the sovereignty angle included. Capex is exploding ahead of demand, the bubble is simmering as we speak. Will a burst of said bubble kill the momentum? Maybe in the US but certainly not in Europe where infrastructure deployment is lagging.
Fourth, the liquidity window has reopened, tentatively. OpenAI, Anthropic, Plaid, Revolut, Ramp, Brex, PayPal are all either preparing for an IPO in the footsteps of SpaceX or will be acquired by hungrier incumbents. As long as the cloud infra bubble and AI consumption holds, we may be in for positive surprises which will percolate through the entire alternative assets market space. Add the diminishing prospects for higher interest rates, at least for the EU, and one could become cautiously optimistic.
Saint Augustine reminds us that the loftier the structure, the deeper the foundations must run. This quarter, the foundations of intelligent money were poured into production — and we intend to keep building on them with you. If any of the themes above resonate, or deserve a challenge, we welcome the conversation; we are always available for a call or a coffee.
With conviction,
Is there more to add here in terms of what we do and how we approach our investing activities and framework to avoid the false positives and focus on that 10% meaningful innovation?
