US national debt has passed $40tn. The figure is quoted often enough to have lost its force, so consider what accompanies it: long-dated yields at their highest in roughly two decades, the 30-year printing above 5.33% this month, and an interest bill near $1.4tn a year. Trillions issued during the zero-rate era are refinancing into a 5% market, while foreign official institutions now hold only around 12% of Treasuries outstanding.

What makes this cycle different is that Washington is no longer the only borrower exposed to that curve. The AI buildout has quietly become one of the largest debt-financed capital programmes in corporate history. The five largest US hyperscalers issued roughly $121bn of bonds in 2025, against an average of about $28bn a year between 2020 and 2024, and bond issuance as a share of their capex has risen from around 2% in 2022 to roughly 17% — a structural shift from a self-funded buildout to a borrowed one.

The consequences are visible in credit: the combined weight of Meta, Alphabet, Amazon and Oracle in the Bloomberg US Corporate Investment Grade index nearly doubled in the year to April, and long-dated hyperscaler paper has widened to roughly 118 basis points over Treasuries.

Scott Bessent's framework was clear: grow the economy at 3%, compress the deficit toward 3% of GDP, and add three million barrels a day of domestic energy production — fiscal repair through growth and supply rather than austerity.

Each leg has slipped. The deficit ran at 5.7% of GDP in 2025 and remains wide of target, and disruption in the Strait of Hormuz has pushed oil higher rather than lower, leaving the Treasury Secretary arguing that expensive oil is a headline-inflation problem while core eases. Growth is the one leg still standing.

When you cannot repair the fiscal position, you manage the price of the debt instead. That is a genuine shift in posture, and the measures arrived quickly. On 19 August, moving off its own published calendar, the Treasury said it would at least double its liquidity support buybacks in the 10-to-20 and 20-to-30-year sectors, lifting the ceiling from $2bn per operation to at least $4bn, effective 9 September through 4 November. The next day Bessent indicated operations could be larger still, that buybacks would become routine, revealing that part of the purpose was to send a signal.

How it would be funded was left open. Most assumed bill issuance, as Bessent's own “Treasury Twist” description implies. On Monday, two senior officials indicated that the General Account, the government's operating balance at the Fed, $935bnon 20 August, is considered available.

That is a materially different instrument. Bill-funded buybacks swap long duration for short and are self-limiting, since they shorten the maturity profile into a refinancing wall. TGA funding removes duration without any offsetting issuance, while transferring cash from the Treasury’s account at the Fed into the banking system, the mechanism that drained liquidity during the 2023 TGA rebuild, now running in reverse.

There are two important caveats. This is not QE (quantitative easing): it redistributes existing reserves, is finite, and needs to be rebuilt before the debt-ceiling bind arrives this winter or early spring. And the $935bn is a stock, not available firepower. The previous administration targeted $550–600bn, leaving roughly $350bn of additional capacity. Still significant, but a long way from a trillion.

Call it buybacks, liquidity support, or maturity management. When a sovereign carrying $1.4tnof annual interest expense begins deploying cash reserves of this size to contain its own borrowing costs, the label has stopped describing the function.

The reaction was immediate and highly selective, which is what makes it informative. The dollar was dumped, falling to a three-month low against the euro and posting its sharpest decline in three weeks. It has struggled to recover since, even as economists upgraded their Q3 US growth forecasts. That is the tell: good growth news would normally support the currency, and the market chose to price the fiscal implication over the cyclical one.

Bonds got relief, then gave it back. The 30-year fell more than ten basis points to 5.184% on the announcement and the 10-year to 4.637%. Within a week it had returned to almost exactly where it started, and it continues to test the market near 5.25%. Each intervention buys a smaller and shorter reprieve.

Equities did not enjoy it as one might expect. Rate-sensitive value names rallied, but the leadership did not participate: the Nasdaq-100 underperformed and the Morgan Stanley US Momentum Index traded lower through the week. This is where the financing story closes the loop. A debt-funded capex cycle is a rate-sensitive one, and momentum in this market is largely the AI trade. A day of lower yields does not offset the cost of capital those balance sheets now carry.

What did benefit were the assets nobody can issue. Gold rose roughly 6% over the week and has extended to around $4,675, its highest since May, with silver near $69. Bitcoin posted its largest weekly gain since March 2023, up some 23%, moving above $80,000 and reclaiming its 200-day moving average for the first time in a year. Ether gained around 28%.

A weaker dollar, a bond market that will not settle, momentum unable to rally, and a simultaneous bid across gold, silver, bitcoin and ether. That is not a growth trade or a rate-cut trade, but rather a currency trade.

For those who believe the money debasement trade has further to go, which “scarce assets” should be considered? And how to define them?

The word is used loosely, so it is worth being precise. The conventional measure is the stock-to-flow ratio: inventory already held, divided by annual production. A high ratio means the existing stock dwarfs new supply and cannot be replenished quickly — gold is the classic case. A low ratio means the asset is abundant and easily produced, which is why most industrial metals and agricultural commodities do not qualify however volatile their prices become. The implication is simple: when money is easy to create, own what cannot be printed.

Gold is the purest expression and has hedged fiat debasement for as long as fiat currencies have existed. Silver carries the same monetary property with industrial demand layered on top, which makes it more volatile and, at points in the cycle, considerably cheaper relative to its own history.

Bitcoin belongs in the category by design rather than geology: supply capped at 21 million units, an immutable monetary policy, neutral settlement, and property rights that do not depend on any jurisdiction honouring them. It has also been declared dead in 2011, 2014, 2018 and 2022, with drawdowns from $32 to $2, $1,000 to $200, $19,000 to $3,000, $69,000 to $15,000, and most recently from roughly $120,000 to $65,000. The volatility is not a defect in the design; it is the design, and a real constraint on how much a portfolio can carry.

Ether qualifies differently. Ethereum's supply is dynamic rather than fixed: new issuance pays validators, while a portion of every transaction fee has been permanently burned since the base-fee mechanism arrived in August 2021. Net issuance therefore depends on network usage, and under sufficient usage supply contracts. Where bitcoin is sound money by fixed rule, ether is called ultrasound money, scarce as a function of adoption rather than decree, and correspondingly the more contingent holding of the four.

The fact that real yields might stay high for longer is one of the major arguments by the bears. Historically, scarce assets such as precious metals have underperformed when real yields were positive.

This is precisely the change most likely to be missed by investors applying the model that has worked for forty years. Rising real yields are supposed to hurt gold by raising the opportunity cost of holding an asset that pays nothing. That relationship began breaking down in early August and has stayed broken: gold has risen while the 30-year sits near 5.25% and the 10-year around 4.70%. Nor has the market quietly priced easing, it has priced the opposite. August's flash composite PMI came in at 56.0, the strongest in more than four years, and futures at one point implied roughly a 40% probability of a rate hike in September. Gold is not merely tolerating a hostile carry environment; it is advancing into one while the market debates tightening.

The explanation lies in the composition of those yields rather than their level. Long rates increasingly reflect fiscal anxiety, debt supply and debasement risk rather than stronger growth or tighter policy. Once that is true, the old rule inverts: higher yields and higher gold can coexist indefinitely, because both are symptoms of the same deteriorating confidence in fiscal sustainability. Gold no longer needs falling real rates to work. It simply requires only the continued impression that the fiscal position is being managed rather than repaired.

Scarce assets remain one of the most effective tools for protecting a portfolio against monetary debasement, and the case has strengthened over the past month. The largest borrower in the world has an overwhelming incentive to keep its funding costs contained and has begun assembling the instruments to do so. The AI capex cycle has precisely the same interest in a lower long end. Neither pressure reverses quickly, nor is resolved by anything currently on the policy agenda. An asset whose supply no authority can expand under political pressure does something in that environment that no bond can do.

The diversification argument is separate and equally valuable. These assets carry low correlation with traditional equities and fixed income and, importantly, with one another. Gold, silver, bitcoin and ether express the same thesis through very different mechanisms: geological scarcity, industrial demand, fixed protocol supply, and network-dependent issuance. Held together, their combined volatility is materially lower than any individual component would suggest, which is what makes a basket more sensible than a concentrated position. Sizing should reflect those differences: gold carries the lowest path risk and the deepest institutional bid and is the core holding, while silver and the digital assets are higher-beta expressions that belong as satellites rather than substitutes.

What would weaken the case is worth naming. If the TGA is never meaningfully drawn and September's operations are funded with bills, the announcement was a free option and much of the recent move reverses. If the intervention succeeds and long yields stabilise below 5%, the evidence for fiscal dominance fades. In addition, a resolution in Hormuz would remove part of the geopolitical bid while reviving the energy leg of 3-3-3.

The comeback of scarce assets is not a bet on inflation. It is a bet that the institutions responsible for the value of money have a growing incentive to let it be worth somewhat less, and are assembling the tools to do so without ever having to announce it.