Financing AI Like It's 1870

Big investment booms tend to bring a matching boom in financial engineering. In Net Interest, Marc Rubinstein argues that AI is following the pattern because the amounts involved leave no other option.

The size of the gap

Estimates for AI infrastructure capex through 2030 run from $7.6 trillion to $8.0 trillion. KKR analysts think traditional US public investment-grade bond investors can supply about $1.7 trillion, which leaves a $6.3 trillion gap. KKR expects the rest to come from private investment-grade debt, asset-backed finance, infrastructure debt, structured solutions "and financing structures that may not even exist yet."

Rubinstein points to a historical parallel. AI capex could reach 3.63% of GDP through 2032. US railroad capex from 1870 to 1890 came to 2.24% of GDP. He quotes historian Richard White's Railroaded on how railroad investors moved from government bonds to secured bonds, then convertibles, first and second mortgage bonds, land grant bonds and income bonds. In White's words, these were "bonds on anything and everything that investors might accept as collateral," sold by bankers who acted like "carnival barkers."

What financial engineers do

Rubinstein admits financial engineering is hard to measure, but he says how much of it is going on is a useful signal of where we are in the cycle. It tends to thrive when ordinary balance sheets can't keep up with an investment boom. The same structures that keep the money flowing can also hide where risk is building up. The engineers' job is to satisfy companies, investors, regulators, auditors and rating agencies at once, splitting cash flows and risk among the parties most willing and able to carry them.

Long-term government borrowing costs have risen across the rich world. Matthew Klein argues that France has been hit hardest, and that the cause is the gap between its yields and its growth, not just the level of its yields.

A global reset, mostly matched by growth

Before the pandemic, long-term yields were near zero or negative in Japan and "core" Europe. They were about 1% in the U.K., about 2% in the U.S., Canada, Australia and Korea, and about 3% in Poland. Today they are about 3% in Japan, about 3.5% in Germany and Sweden, 4% in Canada and Spain, close to 5% in France and Korea, and a bit over 5% in the U.S., U.K., Australia, Czechia and Poland. Overall, rich-country yields sit . Switzerland, Singapore, Taiwan and, to some degree, Greece are the main exceptions.

Financial literacy is supposed to be basic. A survey that has now run for ten years suggests much of the US adult population still lacks it.

What the index measures

The TIAA Institute–GFLEC Personal Finance Index is a 28-question survey given to large samples of Americans every year for the past decade. Eight of those questions serve as its public face. A June 2026 report by Paul J. Yakoboski, Annamaria Lusardi, Andrea Sticha and Fran Mastry, "A decade of tracking financial literacy in America," reviews the results.

The finding is simple. On the eight headline questions, roughly half or fewer of respondents answer correctly, and that share has barely changed in ten years. Scores on the full 28-question test have also held fairly steady over the decade. When the data is broken down by group, some do better than others. Older respondents tend to score higher, for example. But, as the Conversable Economist puts it, "no subgroup is especially covered in glory."

A HumbleDollar reader asked a reasonable question. Over the past century, stocks returned about 10% a year and bonds about 5%. From 2016 to 2025 the gap was wider: nearly 15% a year for the S&P 500 against under 2% for bonds. So why should anyone own bonds?

The usual answer is sequence-of-returns risk. If stocks fall early in retirement, a retiree has to sell shares while prices are low to cover spending, and too many forced sales can drain the portfolio. The reader, "Tom," thought this risk was overstated. His reasoning: a retiree withdrawing 4% a year only "realizes" the loss on that 4%. A 50% crash would then cost just 2%, and long-run stock returns should easily make up for that.

What the numbers show

Adam Grossman ran a simulation using the standard 4% rule: withdraw 4% of the portfolio in year one, then raise the amount each year with inflation. He started with a bad case. Someone who retired in January 2000 with 100% stocks went straight into three losing years: −9%, −12% and −22%. Their portfolio fell nearly to zero within 25 years. At a 4.5% withdrawal rate it ran out within 20 years. Tom's arithmetic misses this: the losses hit the whole portfolio, not just the year's withdrawal.

Still, 2000 was close to a worst case. Over the past 100 years, only two other retirement start dates did as badly over 25 years: 1929 and 1969, the second just before a decade of flat markets and rising inflation. From 1926 to 2025, an all-stock retiree ended with more money than they started with in 91% of rolling 25-year periods. In about half of those periods they ended with five times their starting balance.

The University of Michigan's consumer sentiment index is lower today than at any point in its history, which goes back to the early 1950s. Ben Carlson lists what it now sits below: 9% inflation in 2022, the Covid pandemic, the 2008 financial crisis, the early 1980s (15% inflation, 18% mortgage rates and two recessions in three years), 1970s stagflation, and 11 recessions and 11 bear markets in total.

Other data look nothing like that. Stocks are at all-time highs, unemployment is 4%, household net worth is at a record, and the U.S. hasn't had a real recession in nearly 20 years.

The legitimate grievances

Carlson lists reasons the mood is sour:

• Price levels, not inflation rates. Trailing 12-month inflation is 3.4%, but consumer prices are up about 30% since the start of 2020. Wages rose too. But the jump came so fast that nobody adjusted, and the old prices are still fresh in people's memories. Cars and houses cause the worst sticker shock.
• Housing lock-in. About 65% of households own their homes, so most are shielded from higher prices. But homeowners with 3% mortgages won't take on 7.5% rates, so they feel stuck where they are. Young buyers are priced out. Carlson says this is why housing activity is at financial-crisis levels.
• Wealth concentration. The top 0.1% holds 15% of household net worth, up from 8.6% in 1989.
• Energy and war. The war has pushed national gas prices toward $5.
• Politics and social media. Polarization wears people out, and social platforms reward doom.