The Money Broke

Between 1850 and 1894 the United States ran an experiment in monetary regimes and left a clean record of the result. The NBER wholesale price index, benchmarked so that 1910 to 1914 equals 100, opened the period near 83 and moved in a band roughly between 80 and 115 through the 1850s. War finance broke that pattern. Once the Union began issuing unbacked greenbacks, the index more than doubled, peaking at 225 in 1864. The three decades that followed reversed all of it and more. As the country worked its way back to a hard money standard, wholesale prices fell 69 percent from the greenback peak, closing 1894 at 69, below where the series began in 1850. The decline was gradual and interrupted, with a partial recovery into the early 1880s that gave way to further softness. It also ran alongside one of the most rapid industrial expansions in American history. That combination is the point. Falling prices did not prevent capital formation or growth, which undercuts the standard argument that a monetary system permitting deflation is unworkable. For anyone holding a fixed-supply asset today, this is the closest historical analogue available. A savings instrument that gains purchasing power over decades is not a theoretical construct. The United States lived under one for a generation, and the burden of proof belongs to the side arguing it cannot happen again.

The left panel plots the natural log of the US price level from 1790, using both the GDP deflator and CPI, and the two series tell the same story: prices are 38 times their 1790 level, and nearly all of that increase arrived after 1940. Before then the price level moved in both directions. Wars pushed it up around 1814, during the 1860s, and again around 1919, and each spike was followed by a decline that gave back most of the ground. By 1940 the price level stood only modestly above where it started 150 years earlier. The right panel measures the break directly. Ten-year average inflation from the GDP deflator swung between roughly negative 6 percent in the 1820s and nearly 8 percent around 1920. Since 1940 it has not gone negative once, an 85-year run with no precedent in the prior 140 years. It reads 2.9 percent today. Values before 1913 on the left and before 1939 on the right are traced from published Johnston and Williamson figures and spliced to FRED data thereafter, so read the early path as accurate in shape rather than to the decimal. The practical consequence is that holding cash stopped being a round trip. Under the old regime, waiting out an inflation spike worked because the reversal came. Under the current one, only the rate is ever in question, never the direction.

The Bureau of Labor Statistics tracks what a consumer dollar actually buys. Indexed to January 1913 at 1,020.4, that measure stood at 29.9 in June 2026. The decline is 97.07%, which leaves the 1913 dollar with 2.93 cents of its original purchasing power after 113 years.

The path is not a smooth slope, and the interruptions are instructive. The First World War cut the dollar nearly in half, to 481 by June 1920. It recovered through the 1920s to about 590, then gained again during the Depression, peaking at 800 in May 1933 as prices collapsed. That was the last time holders of cash were meaningfully rewarded. Wartime and postwar inflation took the measure down to about 445 by the late 1940s, the 1950s and 1960s ground lower in steady steps, the 1970s fell steeply, and since 1983 the index has spent every month below 100.

The useful reading is not that inflation is high in any given year. It is that the unit most savings are denominated in has depreciated in almost every decade for over a century, and the two periods when it did not were a world war aftermath and a depression.

Holding wealth in that unit is a position, not a neutral resting state.

The Federal Reserve describes 2% as a symmetric target, meaning misses above and below should balance out over time. The record since the pandemic reads differently. CPI crossed above 2% in March 2021, peaked at 9.1% in June 2022, and has not printed below the target since: 65 consecutive months, averaging about 4.5% over the stretch. The July 2026 reading of 3.4% arrived after a reacceleration through the spring, which makes it hard to describe the latest stretch as the tail end of a smooth descent. A miss that persists for more than five years and averages more than double the target is better understood as a revealed policy preference. Debt of the size the US now carries is easier to service in a currency that gives up 3 to 4% of its purchasing power every year. For anyone holding savings in that currency, the floor at 2% is the advertised loss rate, and the last five years show the realized rate runs well above it.

Since 1960 the growth rate of US M2 has spent most of its time above the 10-year Treasury yield. The extreme reading came in 2020, when the money supply expanded 19% while the 10-year yielded under 1%, a gap of more than 18 points. It is not an isolated episode. M2 growth ran above the yield through most of the 1960s and early 1970s, again in the 1990s and 2000s, and again from 2019 onward. The clearest exception is the early 1980s, when the 10-year reached about 14% and the yield sat above money growth for most of a decade. Since 2023 the two have converged near 4% to 5%, with 2026 shown year to date. Both series come from the Federal Reserve, M2SL and GS10. The point for anyone holding capital is that the nominal risk-free rate has rarely been a hurdle worth clearing. If the unit of account is expanding faster than the safest instrument pays, then holding Treasuries preserves the number and not the claim on real assets. That reframes what counts as a conservative position. The default cash allocation has been quietly losing ground for most of living memory, and the 2020 reading was the most expensive version of that trade rather than a new phenomenon.

Federal debt stood near 57% of GDP in the first quarter of 2000 and drifted up to roughly 64% by 2008. The financial crisis moved it to about 100% within five years, where it plateaued for most of the 2010s. The 2020 shutdown pushed it to a peak of 132.7%. It has since eased to 122.6% as of the first quarter of 2026, still more than 20 points above the pre-2020 plateau and more than double the level of 2000. The 2021 to 2025 average sits 11% above the 2016 to 2020 average. Data is FRED series GFDEGDQ188S, retrieved August 2026. The shape is the argument. Debt does not oscillate around a mean here. It steps up at each emergency and settles at a new, permanently higher floor, because the spending enacted during a crisis rarely reverses and the borrowing is never retired. Two shocks 12 years apart moved the ratio by roughly 65 points in total. Nothing in the record suggests the next shock behaves differently. For someone holding or deploying capital, the question is not whether the debt gets repaid in nominal terms, since it will. The question is what the currency it is repaid in will be worth, and the ratchet in this chart is the reason to hold assets that cannot be issued.

Net interest took about 12.4% of federal outlays in fiscal 2000, then fell for a decade as rates declined, bottoming near 5.3% in 2010. It sat in a 6% to 6.5% band through the middle 2010s, rose to about 8.4% in 2020, and dropped again to roughly 5.2% in 2021 as emergency spending swelled the denominator and rates sat at zero. From there it has gone straight up, reaching a forecast 14% in fiscal 2026. The FY2026 figure is a forecast, not an actual. Data is from OMB via FRED. Interest is the one line in the budget nobody votes on. It is set by the size of the debt and the level of rates, and it is paid before anything else. At 14%, roughly one dollar in seven of federal spending buys no service, no transfer, and no investment. Every increment it takes has to come from somewhere: higher taxes, lower discretionary spending, or more borrowing at the prevailing rate. Historically the path of least political resistance has been to let nominal growth and inflation shrink the real value of the obligation. Anyone holding long-duration nominal claims on the US government is on the wrong side of that preference, and the pressure behind it strengthens with every year the interest share climbs.

From January 2000 through June 2026, hospital services rose 291%, college tuition 197%, and medical care 150%. Average hourly earnings rose 136% over the same period. Housing rose 115% and food and beverages 109%, both below wages. The three categories that outran pay are the three a household can least easily avoid, postpone, or substitute. Medical care and college tuition have beaten wage growth in every year since 2001, so this is not a single episode that a later stretch of real wage gains offsets. It is a 26-year trend. All series are rebased to January 2000 and drawn from FRED and Bureau of Labor Statistics data. The arithmetic is uncomfortable. A household whose pay has grown 136% while its hospital bills have grown 291% is not poorer on paper, but its discretionary income has been quietly reallocated to the least negotiable line items. Savings capacity is what gets squeezed. For capital already accumulated, the relevant measure of preservation is not the headline inflation rate but the growth rate of the specific obligations the money is meant to fund. Hospital costs have grown more than twice as fast as the pay meant to cover them. A portfolio benchmarked to headline inflation is quietly losing ground against the bills that actually come due.

Between 2014 and 2023 the average US seed round grew 3.1x, rising from about $700,000 to roughly $2.5 million after peaking near $2.6 million in 2022. US M2 grew 1.9x over a longer stretch, from about $11 trillion in 2014 to above $20 trillion by June 2026. The median seed round traced the same arc at a lower level, moving from roughly $270,000 to about $1 million. Seed pricing, in other words, inflated faster than the money that was supposedly driving it.

The second panel explains why that is strange. World Bank Doing Business data, a series discontinued after 2020, shows the cost of starting a business in low and middle income economies falling from 143% of income per capita in 2004 to 25% by 2020, with high income economies drifting from about 14% to about 5%. The real cost of forming a company collapsed while the checks written to fund one tripled.

For anyone deploying capital into early-stage companies, that gap is the entry-price problem stated plainly. Larger rounds at lower operating cost mean more of the eventual return has to come from valuation rather than from the business clearing its first milestones, and it leaves entry price, not company quality, as the variable most likely to decide the outcome.

The share of mature companies whose interest costs exceed their profits for three consecutive years rose from 4% in 1987 to 15% in 2020, according to Congressional Research Service and Bank for International Settlements data. The climb was continuous rather than crisis-driven: about 6% by 1997, 8% by 2002, 11% by 2012, and 14% by 2018. Each easing cycle left a larger residue of firms that could service debt only because money was cheap.

By 2020 roughly one mature company in seven fit that description. These are not startups burning venture money toward a milestone; they are established firms whose earnings do not cover their interest bill and have not for years.

For anyone holding broad equity or credit exposure, this is a statement about what the index actually contains. A zombie share near 15% means a meaningful slice of listed corporate America depends on refinancing conditions rather than operating performance, and that dependence only shows up when rates or credit availability move against it. The population of firms that cheap money kept alive is also the population that a sustained repricing of capital has to work through, which is a very different risk than the earnings risk most portfolios are built to absorb.