On May 18, Sygnum Bank announced the first live AI-agent-driven digital asset transactions executed by a regulated Swiss bank.

A client issued plain text instructions. The Sygnum AI agent independently planned each step, reviewed the smart contracts, flagged the risks, and prepared the transactions. The client signed each one from a self-custodial wallet on their own device. The agent handled stablecoin transfers, asset swaps, on-chain lending positions, token wrapping, and liquidity provisioning.

The pilot was built on a Model Context Protocol (MCP) server developed in-house by AI@Sygnum, using Anthropic's Claude as the underlying AI model.

Thomas Frei, Head of AI and Data Analytics at Sygnum, framed the trajectory directly.

The architecture is asset-class agnostic. The same plumbing that handles stablecoin transfers handles tokenised equities, gold, and securities. Sygnum's product roadmap will require FINMA approval before general client deployment, but the pilot has been completed.

Why this matters.

For forty years, the deposit franchise of every developed-market bank has been protected by a single variable. Friction.

The saver does not check the rate. The treasurer checks but does not move. The cost of switching exceeds the yield differential at most scales. Money market funds have existed for fifty years but have not killed the bank deposit franchise.

That protection is being removed in production.

Sygnum is not a fintech. It is a FINMA-regulated digital asset bank.

The architecture demonstrated on May 18 is not theoretical. It is operational on Mainnet, in plain text, with the client retaining custody and consent at every step. The technology is general purpose. The underlying AI model is the same one millions of users access daily.

A retail saver does not yet have an agent that compares deposit rates, money market fund yields, stablecoin rewards, and tokenised T-bill returns and executes when the differential clears a threshold. They will. The corporate treasurer is closer. The institutional treasury function is closer still.

Sygnum is two days ahead of the rest of the industry. The rest of the industry is months behind, not years.

The lazy tax does not survive the agentic age.

The Bank for International Settlements published Working Paper No. 1270 last year, revised earlier this year: "Stablecoins and safe asset prices."

The headline finding is structural.

During periods of US Treasury bill scarcity, a $3.5 billion stablecoin inflow compresses 3-month Treasury yields by 5 to 8 basis points.

That is roughly double the baseline impact under normal market conditions. The marginal investor in short-dated US sovereign debt is now, at the margin, a stablecoin reserve manager.

The mechanism is straightforward. Stablecoin issuers hold reserves predominantly in short-dated US Treasury bills and repo collateralised by Treasuries. As stablecoin supply expands, the reserves expand. The reserves are deployed into the bill market. When the bill market is constrained, that deployment moves the price.

As of January 2026, USDT alone holds $186 billion in circulation with 63 percent of reserves in T-bills. USDC adds another large multiple. The combined stablecoin reserve demand for short-dated US Treasuries now sits at scale comparable to medium-sized sovereign wealth funds.

Why this matters.

When the BIS publishes a paper, the central banking establishment is signalling that something has moved from policy curiosity to structural consideration.

The BIS is the central bank for central banks. Its working papers are not retail analysis. They are the intellectual scaffolding for what the world's monetary authorities are thinking about next.

The implication is layered. Stablecoin growth is no longer a niche payments story, it is a story about the demand curve for the safest asset in the global financial system. The US Treasury benefits from the structural demand. Other sovereigns notice. Australia notices.

For the Australian bank franchise, the implication is sharper. Stablecoin reserve demand is rerouting capital that historically deposited in the global banking system into US Treasuries via stablecoin issuers. The capital does not disappear. It is intermediated by a different rail.

That rail is not the Australian banking system. It is not any banking system. The disintermediation is structural and the BIS has now quantified it.

Money market funds did the same thing to US bank deposits in the 1980s. The mechanism is identical. The wrapper is just faster.

Jamie Dimon does not lose deposits if he can help it.

On May 6, 2026, JPMorgan's blockchain unit Kinexys, working with Mastercard, Ripple, and Ondo Finance, completed the first cross-border, cross-bank redemption of a tokenised US Treasury fund executed on a public blockchain.

The asset leg cleared on the XRP Ledger in under five seconds. The fiat leg moved through Mastercard's Multi-Token Network into JPMorgan's correspondent banking infrastructure and arrived as US dollars in a Singapore bank account.

Outside US banking hours. Outside Singapore banking hours. In near real time.

Ondo Finance redeemed its OUSG token, a tokenised representation of short-dated US Treasuries with $250 million in assets under management. The settlement asset on the XRP Ledger was Ripple's RLUSD stablecoin. JPMorgan delivered the actual dollars.

For JPMorgan, this was the first time its private blockchain rails connected directly to a live public Layer 1. Kinexys has processed more than $3 trillion in total transactions on its private infrastructure since launch. The May 6 pilot was the moment the private rail joined the public one.

Why this matters.

For two years the Australian banking sector has framed stablecoins as an external threat. Something to be opposed in regulatory submissions. Something to be regulated out of competitive relevance. Something that happens to other people's deposits.

JPMorgan has read the same regulatory landscape and reached a different conclusion. If tokenised treasuries are going to settle on public blockchains, the bank that operates the cash leg captures the institutional flow. If stablecoins are going to be the settlement asset, the bank that integrates with them keeps the customer. If 24/7 cross-border settlement is the future, the bank that builds the bridge owns the bridge.

Asked about stablecoins on JP Morgan's July 2025 earnings call, Dimon was direct.

Dimon then built the bridge. And two weeks ago, he used it.

The depositor who is about to be optimised by an agentic AI does not care whether the receiving wallet holds a stablecoin or a tokenised Treasury or a deposit token. The depositor cares about the rate, the speed, the auditability, and the regulatory standing. The May 6 transaction was bank-grade on all four.

The Big Four operate four payment rails (BPAY, NPP, the wholesale RTGS network, and SWIFT for cross-border). None of them runs 24/7. None of them connects to a public blockchain. None of them has settled a tokenised Treasury redemption with a Singapore counterparty over a weekend in five seconds.

The competitive gap that opened on May 6 is structural, not technological. JPMorgan has the infrastructure.

October 28, 1991.

The Andrea Gail, a 72-foot commercial fishing vessel out of Gloucester, Massachusetts, was working the Flemish Cap, 1,200 nautical miles east of her home port. The catch had been poor for weeks. The crew had pushed further than planned and they had finally filled the holds.

Three weather systems were converging behind them.

Each had been tracked. Each had a name and a forecast. Each on its own was survivable.

The convergence produced waves measured at over 100 feet. The largest waves ever recorded by an offshore buoy in the North Atlantic. The Andrea Gail's last transmission was at 6pm on October 28. The boat was never found.

The Australian Big Four are not a fishing boat.

They are well-capitalised, well-managed, and the most profitable banks in the developed world. The crew is experienced and the structural buyer of their equity is the largest compulsory pension system in the developed world.

Three forces are now converging on the spread that funds the franchise.

This is the analysis of those three forces. What each does. Why each alone is survivable.

Why the convergence is not.

THE FIRST FRONT: STABLECOINS

The first front of the storm is the one I have lived professionally.

Two decades on investment bank trading and treasury desks, pricing the paper that funds these franchises. Since 2021, on the stablecoin side from a risk and advisory standpoint.

Stablecoins are not ‘crypto.’ They are digital money market funds.

A USDC token is a claim on a reserve held by the issuer. The reserve is short-dated US Treasury bills, repo collateralised by Treasuries, and bank deposits. The composition is identical to a regulated money market fund. The wrapper is a token rather than a fund unit. For the majority of the market, the yield is captured by the issuer rather than passed to the holder.

For now.

The 1980s precedent is exact.

Money market funds attacked the spread that funded US bank lending. The mechanism was a yield-bearing alternative to a transaction deposit, regulated as something other than a deposit, drawing balances out of the banking sector into instruments that did not fund bank lending.

By 1990, money market funds held over $400 billion in assets that had previously sat in US bank deposits. The deposit base of the US banking system was structurally repriced and did not recover.

The 2026 mechanism is the same. The wrapper is just faster.

The GENIUS Act prohibits passive yield on regulated stablecoins. The CLARITY Act Section 404 preserves activity-based rewards, cashback, fee rebates, and loyalty programmes that are economically equivalent to yield in everything but legal classification.

The American Bankers Association has spent the past quarter trying to tighten that language. They have lost. The architecture is preserved. The political battleground has narrowed to the design of the rewards programme rather than the existence of the spread-sharing model.

Brian Moynihan, CEO of Bank of America, told an earnings call last quarter that interest-bearing stablecoins could siphon up to $6 trillion in deposits from the US banking system. That figure is 30 to 35 percent of total US commercial bank deposits. A global systemically important bank CEO does not quote that number on an earnings call lightly.

Apply the migration ratio to Australia. The ADIs hold approximately $3 trillion in household and business deposits. A 30 percent migration is $900 billion. A more conservative 10 percent is $300 billion. The Big Four's combined deposit base would shrink by an amount larger than CBA's entire mortgage book.

The replacement is wholesale funding.

A transaction deposit costs the bank close to nothing, around 10 basis points. The bank deploys it at the cash rate and keeps almost the entire spread. When that deposit migrates, the bank must replace the funding in the wholesale market, where major bank senior paper currently prices at around 55 to 75 basis points over swap. The bank still earns the cash rate on the asset, so the true compression is not the full wholesale yield. It is the loss of the near-free funding benefit: in the order of 150 to 250 basis points of margin on every dollar that moves from a transaction account to wholesale replacement.

A 10 percent migration of transaction deposits is approximately $80 billion. At a 150 to 250 basis point compression, the annual earnings hit is in the order of $1.2 to $2.0 billion across the four majors. That is roughly 4 to 6 percent of combined Big Four annual cash earnings.

And that is the conservative case.

It assumes only a tenth of the exposed deposits move, and it assumes human-paced migration

That is Front One in isolation.

THE SECOND FRONT: AGENTIC AI AND THE LAZY TAX

Money market funds have existed for fifty years. They have not killed the bank deposit franchise.

The reason is friction.

The retail saver does not check their deposit rate. They opened the account in 2003. They have not looked at it since. The corporate treasurer checks, knows the bank rate is below the alternative, and does not move. The cost of moving (operational risk, audit trail, counterparty diligence, payroll integration) exceeds the yield differential at most scales. The SME owner has no time. Payroll runs through the bank account. The mortgage is at the bank. The relationship manager calls every quarter. Moving the operating account is a project that does not get prioritised.

This collective inaction is the lazy tax. It is the spread that funds the whole business model.

The Productivity Commission has used the term to describe the higher rates Australian banks charge existing customers compared to new ones. Reserve Bank speeches have referenced it in the context of consumer banking competition. ASIC has reported on it in deposit and mortgage markets.

The lazy tax is not a thesis. It is a documented feature of the Australian banking system.

For forty years the lazy tax has protected the Big Four from money market fund competition. The fund yields more. The deposit pays close to nothing. The depositor does not move.

That protection is being removed by software.

Agentic AI is the term for systems that act on outcomes rather than commands. Give the agent a mandate (preserve principal, maximise yield within a defined risk envelope, maintain operational liquidity above a threshold), and the system continuously compares deposit rates to money market fund yields to stablecoin rewards to tokenised T-bill returns.

When the differential clears the threshold, the system executes.

The agent does not get tired. The agent does not have a relationship manager. The agent does not feel the inertia.

The connection between agentic AI and the lazy tax has not been drawn cleanly in published research.

This is not a 2027 thesis. We already covered Sygnum Bank’s agentic pilot earlier. And that it used Anthropic's Claude as the underlying AI model.

The vendor ecosystem is already commercial. Kyriba launched its Treasury AI platform commercially in late 2025. Bottomline launched Bea for cash management. Embat raised €30 million in Series B last week to scale agentic treasury across the UK and Ireland. TIS launched AI Discovery for Payments through AWS Marketplace. PwC's most recent treasury survey reports that 74 percent of treasury teams are expanding or actively using AI.

The lazy tax does not survive the agentic age.

The Big Four's free-float deposit base has been protected for forty years by the friction of switching. The friction is being removed in production right now.

That is Front Two.

THE THIRD FRONT: THE AUSTRALIAN BUDGET

The third front is specific to Australia and very recent.

On the evening of May 12, 2026, Treasurer Jim Chalmers delivered the Federal Budget. He described it as "the most important and ambitious Budget in decades." It arrived against a global oil shock driven by conflict in the Middle East, with headline inflation forecast to reach 5 percent by mid-2026.

Two structural changes were announced.

The first is negative gearing reform. From July 1, 2027, losses on established residential properties acquired after 7:30pm on May 12, 2026 will only be deductible against residential property income and capital gains, not against wages or other income. Unused losses can be carried forward. Properties owned on Budget night are grandfathered under the existing rules until they are sold. Newly constructed properties remain fully negatively geared, as do investments supporting affordable housing programs.

The second is capital gains tax reform. From July 1, 2027, the 50 percent CGT discount on assets held for more than 12 months will be replaced with cost-base indexation plus a 30 percent minimum tax on net capital gains. The reform applies to all CGT assets held by individuals, partnerships, and discretionary trusts. Not just property. The 30 percent minimum tax on discretionary trusts is explicit in the package, which matters because trusts are a primary vehicle for high-net-worth holding of bank shares. Superannuation funds retain their existing CGT concessions.

The headlines have focused on the property market. The mortgage book implication for the Big Four is real. Treasury projects house price growth slowing by around two percentage points. Mortgage brokers report meaningful reductions in borrowing capacity as banks strip negative gearing tax benefits out of serviceability calculations.

The largest, highest-margin segment of the Big Four mortgage book, investor lending into established housing, is being repriced before any property sells.

That is the visible front.

The less-discussed implication is for the bank shares themselves.

The 50 percent CGT discount has been a structural feature of Australian portfolio construction since the Howard government introduced it in 1999. Self-funded retirees, family offices, SMSFs in accumulation phase, and high-net-worth individuals have used Big Four bank shares as a long-duration yield instrument with favourable tax treatment. The CGT discount applies and dividend franking applies. The combination has produced a post-tax IRR that no equivalent global bank share offers.

The Treasury's own analysis notes that 83 percent of the CGT discount benefit currently flows to the top 10 percent of taxpayers by income. The same 83 percent of new investor loans in 2025 were for established housing. The Government has named both concentrations as targets of the reform. The structural buyer of Big Four shares at the personal portfolio level is exactly the cohort being repriced.

Cost-base indexation plus a 30 percent minimum tax changes the long-term holding economics of every long-held Australian bank share in non-super wealth. A holder who bought CBA at $50 in 2010, sitting on a large paper gain in 2026, faces a materially different after-tax outcome on sale under the new rules. The discount that would have applied to half the gain is gone. The cost base is indexed by CPI instead. In a low-inflation environment, indexation is a worse outcome than the old discount.

Superannuation is exempt. That is structurally important. The compulsory pension system that has bought Big Four shares for two decades retains its concessional CGT treatment. The pension flow into the banks does not change.

What changes is the demand at the margin from non-super wealth. Family offices. SMSFs in pension phase. Direct retail shareholders. Discretionary trusts. The cohort that has provided incremental demand above the structural super flow is being repriced, and the trust-held portion is being taxed directly.

This is Front Three.

The May 12 Budget did not target the banks. It targeted housing affordability and the inter-generational fairness of the tax system. The bank franchise is collateral damage in a tax reform designed for a different purpose. That is what makes the timing structural rather than political.

The convergence of all three fronts begins on July 1, 2027.

THE CONVERGENCE

Each front is survivable on its own.

A 10 percent deposit migration to digital money market funds is uncomfortable for the Big Four but not existential. The funding base shrinks, wholesale replacement is more expensive, NIM compresses by an amount that resembles a bad quarter rather than a structural break. The Big 4 have absorbed worse shocks.

Agentic AI removing the lazy tax is also survivable in isolation. The deposit franchise loses some pricing power. Treasury cash optimises faster. Some balances migrate. The bank responds by paying competitive rates on a larger share of the deposit base. NIM compresses again. Earnings recover at a lower run rate.

The May 12 Budget reforms are survivable too. Mortgage demand softens. House price growth slows. Non-super demand for bank shares moderates. The structural super flow remains. The dividend continues. The franchise continues.

Each independently survivable.

The convergence is not.

The reason is mechanical. Each front attacks a different part of the same balance sheet, simultaneously, beginning at the same time.

Front One attacks the funding base.

Specifically the low-cost transaction deposit segment that funds 45 percent of CBA's net income. A 10 percent migration here is not a 10 percent earnings hit. It is the replacement of approximately $80 billion of effectively-free funding at a compression of 150 to 250 basis points. In the order of $1.2 to $2.0 billion of annual cash earnings, lost.

Front Two accelerates the migration that Front One enables.

The Moynihan number of $6 trillion offshore assumes human-paced switching by retail savers and corporate treasurers. The agentic AI thesis is that the switching is no longer human-paced. The Sygnum pilot demonstrates the architecture. Once deployed at scale, the migration timeline compresses materially. The bank's ability to manage the funding transition through standard treasury practice depends on time the storm does not give it.

Front Three reprices the equity holder at the same moment.

The lower earnings that Fronts One and Two produce arrive in a market where the structural buyer at the personal portfolio level is being repriced. The 50 percent CGT discount that has supported holding bank shares for two decades is gone. The negative gearing flow that supported mortgage demand is gone. The marginal demand for bank equity weakens at the same moment the earnings base weakens.

Fronts One and Two are already at sea. Front Three is legislated for July 1, 2027. That is the convergence moment.

The structural buyer that has held the bank multiple at 25x is the compulsory super system. Super remains. The CGT exemption for super remains. The structural buyer is preserved.

What changes is everything outside super.

The non-super wealth holder. The family office. The SMSF in pension phase. The high-net-worth direct holder. The trust. The cohort that has provided the incremental demand that pushes the multiple from "structurally supported" to "structurally elevated."

CBA at 25x is not the structural-support multiple. The structural-support multiple is closer to 18x, the average of the other major banks. CBA's premium to the structural support has been provided by the cohort the Budget is now repricing. Westpac at 18x carries a similar but smaller premium. NAB and ANZ trade closer to the structural support level.

The convergence is not "the banks fall to zero." It is "the bank multiple compresses back to structural support" while earnings compress for distinct reasons. The two compressions multiply rather than add.

A 15 percent earnings hit applied to a 25x multiple that compresses to 18x produces a share price decline materially larger than either factor would suggest in isolation.

The Andrea Gail did not drown because of one weather system. She drowned because she met three at once, in the same week, in the same square mile of ocean, with no time between waves to recover. The crew was experienced, the boat was well-built, but none of it mattered.

The convergence is the storm.

WHAT THIS MEANS FOR PORTFOLIO CONSTRUCTION

This is not investment advice.

It is the framing of a question that needs to be answered.

The Australian super fund balanced or growth option is overweight ASX 200 financials by construction. Index-weighted exposure to the Australian market is structurally overweight the banks because the banks are the largest stocks in the index. The conservative super option, favoured by members within ten years of retirement, holds a higher allocation to Australian equities and is therefore even more concentrated.

For a fund with $100 billion in a balanced option and 22 percent in Australian equities, approximately $7 billion sits in the four major banks plus Macquarie. That is the structural floor of the position.

Above the structural floor sit three active decisions.

The first is overweight or underweight against the index benchmark. The second is the allocation to bank hybrids and major bank senior unsecured. The third is the credit overlay on the broader corporate bond book given that two-thirds of Australian investment grade corporate debt is issued by the same four institutions.

Each of these is exposed to the convergence in a different way.

Equity overweight is the cleanest exposure. If the multiple compresses to structural support and earnings compress from the convergence, an active overweight produces alpha that is the inverse of what it has produced for a decade. The question is not whether to be underweight. The question is whether the active manager has the conviction to be underweight a sector that has compounded at double-digit returns annually for ten years, in a market where the structural buyer remains structurally supportive.

Hybrid spreads are exposed differently. The hybrid market for major bank capital has priced spreads at levels that reflect the franchise quality. Spread widening of 50 to 100 basis points is a reasonable expectation if the convergence plays out. The current owner of major bank hybrids takes mark-to-market pain. The future buyer at wider spreads takes the running yield. The question is whether to be early reducing the position or late buying the widening.

Investment grade corporate credit is the most complex exposure. The Big Four issue most of the senior unsecured paper in the AUD corporate bond market. The cost of replacement funding for the bank that the convergence is repricing is partly the spread these instruments will need to clear. The credit overlay decision is not "do I own bank senior." It is "what spread do I need to be paid to own it through this cycle."

The conservative super member is the cohort with the least optionality. The member close to retirement, holding the conservative balanced default, has the highest concentration in Australian financials at exactly the moment the convergence is hitting. Their trustees have the structural buyer mandate. The member has the lifetime accumulation. The trustee's mandate and the member's interest may diverge in the next two years in a way they have not diverged in the past two decades.

This is the question the next two years will ask. Not whether the Big Four can survive the convergence. They can. The question is how the equity holder, the hybrid holder, and the conservative super member should be positioned through it.

Five of the six largest companies on the ASX. The most expensive bank stocks in the developed world. Three converging fronts beginning July 1, 2027.

Each independently survivable. The convergence is the storm.

THE NEXT ISSUE

The Andrea Gail could not adapt to the storm.

The Australian big banks will be forced to.

Their franchises are too large, too institutionally embedded, and too well-capitalised to disappear. What changes is the response. The challenger banks, the digital banks, the stablecoin-native rails, and the agentic AI platforms force a change in their playbook.

Issue 10 is out next week and will game out their likely response.

Moneyball: “Adapt or die.”

Thanks for reading.

Mark McKendry

Stablecoin State

P.S. If you would like to contact our team just reply to this email - we read every response.

If this issue made you think differently about stablecoins, forward it to a CFO, treasurer, asset manager or finance leader who should be reading it.

The conversation is just getting started.