Can Gov Deficits crowd out AI Capex? No! So, higher household savings required. Plus, evidence of more friction in funding capital account surpluses

• Rising real yields reflect structural shifts: Long-end yields are climbing not just due to inflation fears but rising real rates, suggesting expectations of stronger trend growth, higher risk premia, or tighter global capital supply.

• Surge in capex demand: A coming wave of AI-driven investment, especially in the U.S., could significantly boost corporate capex, mirroring the dot-com boom and pushing the corporate sector into a net borrowing position again.

• Large fiscal deficits projected in most major economies: It’s not just a US story: China running an 8% government deficit through to 2030, and worsening budget positions in Germany and Japan.

• Global savings-investment imbalance tightening: Post-GFC changes, like reduced capital mobility, mean countries are increasingly reliant on domestic savings to fund investment and deficits—requiring higher interest rates.

• Sustained higher equilibrium rates likely: With large fiscal deficits across major economies and higher private investment needs, the bond market may be signaling a long-term rise in equilibrium interest rates (r*), ending the era of cheap money.

For better or worse markets had (more Trump trade headlines as I write) substantially reduced estimates for a recession this year. Betting markets have reflected the same sentiment. Our own measure of cross market implied growth has recovered smartly pushing implied vols back down in many markets.

But across all major markets back end yields have continued to trend higher - a process that began post Covid, paused for a bit and then started again last year before accelerating over recent weeks.

This is a bit odd when we look at front end yields which have been falling over the past couple of years as headline inflation declined. Japan is an outlier here (which partly explains the particularly acute sell off in long dated JGBs compared to the rest).

There are good cyclical reasons for this latest sell off: the risk of recession owing to outlandish “sudden stop” tariffs levels is seen as having declined, but Mr Trump is still causing import prices to rise (in the US, but surely one side effect is lower prices overseas as US import demand falls) and US official and non-official labour supply to decline.

Cyclically the economy remains at or close to full employment and measures of the output gap indicate little spare capacity. And now fiscal policy will remain loose! How can the Fed cut rates in that sort of environment - in fact may they need to raise them? But those factors aren’t at play to the same extent in the rest of the world (ex Japan which now has core inflation over 3% - which is a really big, and currently ignored, story for global yields).

It’s also interesting that US implied swaption vol doesn’t appear to have spiked suggesting this is an orderly rather than a panic sell off or conversely that real net selling is happening rather than hedging or playing the swaption vol surface.

And finally it is very interesting that the increase in nominal yields has been mostly comprised of increases in real yields usually a sign of stronger trend growth (r* anyone?), funding squeezes, or increasing risk aversion. It is true that US Sov CDS has increased during the recent tariff kerfuffle - but that has not been the case for other sovereigns.

Sure, where the US yield curve moves the rest of the world follows and US spreads to other major markets have increased so perhaps this is a purely US phenomenon cyclical/supply phenomenon which is merely dragging other countries along.

But Andy and I wonder if this relatively orderly increase in real yields over the past few years is telling us something important about global investment vs savings across government, corporate and households and something about a retrenchment in global investment and savings flows (or how easy it is to finance current account deficits).

Everyone (and their dog) is now bored of saving/investment identities. But we know a few things now about the outlook.

First, that the GOP fiscal plan will either worsen or leave the US structural deficit position worse off. Certainly if there is a recession in the next few years this is a terrible starting point for the subsequent deficit balloon.

Second, if tech folks and consultants like McKinsey’s are right, we are at the start of very large global capex increase to roll out machine learning LLMs and agents.

This points to an increase in net corporate investment in the US. lower taxes and lower government spending. If Mr Trump wants to reduce the capital account surplus at the same time then only higher household savings can square the circle. Perhaps bond yields are telling us something about the future interest rate required to move the US domestic savings market toward a higher rate of flow.

Its worth thinking about the capex side for a moment. It might surprise you but the US actually only adds to its net capital stock at a run rate of about 3% of GDP (see chart below).

Gross capex is a lot higher but the vast majority of it is replacement capex to deal with depreciation. Yes, these are GDP stats from the BEA so they are estimates but it is a reality that mature economies with large capital stocks need a lot of capex just to stand still. And, as newer capital goods have much faster depreciation rates than real world stuff capital consumption will keep rising.

Therefore if McKinsey’s estimate of $7trn by 2030 to build out AI infrastructure is in the right ball park, and half of it happens in the US then about $700bn would need to be spent per annum. That represents about a 66% increase in net capex and a 16% increase in gross capex.

Using the most recent GDP numbers net capex would rise to about 6% of GDP from below 3% now. No way you say? May I refer you to the dot.com boom - you can see it in the chart above for net capex. Corporate net capex peaked out at around the same rate then and the corporate sector saving/investment balance went negative for almost 5 years.

But there’s more. Consider the IMF’s government deficit projections till the decade end for the major bond markets (chart below). I hope you are as surprised as I am. China is projected to run deficits of 8% through to 2030, France around 6% and Germany and Japan are both worsening too. The UK stands out, but Mr Starmer is rowing back fast.

And of course if everyone is going AI and robot capex crazy then corporate sector capex everywhere should rise too leading to either smaller corporate surpluses or deficits. That need not require an increase in household saving rates if countries can run capital account surpluses easily and (relatively) cheaply using funding from excess saving countries. But is that still the case structurally let alone cyclically (with China running lower capital account deficits and still controling the thing, stopping the inevitable wall of money that would like to “diversify”)?

The charts below are fascinating (I would think that!). They are scatter charts of a cross country panel of savings and capex rates as a % of GDP from the World Bank database.

In a completely closed capital account world you really have no choice but to fund your own investment with your own savings. You’d therefore expect to see a very strong correlation between domestic savings and investment across economies - a 45 degree line and a high r^2 for your trend line.

In a completely open capital account world there need be no relationship between the points. Kent’s saving rate doesn't determine it’s capex rate. So you’d expect a poor correlation between national savings and national investment rates. Clearly this doesn’t mean, if everyone has their own currency, that interest rates would be the same everywhere. Instead they would clear along with the currency to ensure, globally, that savings = investments. If there’s a global central bank helping along (ahem, the Fed?) then so much the better.

I compare three years: 1996, 2006 and 2023 (the last data points available). In 1996 the correlation was pretty good with outliers - you basically didn’t have a completely free and frictionless capital market system.

But 2006 - picked to be just before the start of the GFC - starts to hint at a very strong change in the relationship. The correlation had dropped substantially. Some may recall at the time discussions about this - capital accounts opening, easy trade finance, and western banks running large overseas USD loan books. Funding deficits was easier and high capex countries could go for growth.

But the 2023 relationship is really quite worrying. It shows an even better fit than the pre-2000s. National savings and investment rates are more correlated now than they were almost 30 years ago. We understand some of the reasons why; the roll back of international banking post GFC is one. But others a lot more complex.

In this world funding domestic government and corporate deficits is either more expensive, since you have to attract foreign capital which has lower elasticity and less market depth (ie larger price moves to attract flow), or because you have to increase your domestic savings rate through higher interest rates to clear domestic S-I balances without any central bank policy moves.

Yes, central banks could monetise government deficits instead, re-entering a QE phase, but they are (mostly) independent and we are not in a deflationary environment any more. This isn’t likely. In fact any attempt to do so could be catastrophic for currencies and long dated yields.

So yes, there are clearly cyclical reasons for a steeper curve and bar a recession its tough to see a route to much rates soon - with Japan playing a key role in the recent sell off. But if this analysis is correct the bond market may be telling us something very important about the demand for funds and the international supply of those funds on a trend basis.

If this is the trend then higher equilibrium rates are here to stay for a long time (with the normal cyclical overlay). Having had a long period of cheap money leverage levels across economies will need to fall as a result.