The Statist Peril of Techno-Asset Inflation

The Mises Institute published a Mises Wire essay this month arguing that the most dangerous asset bubbles in history share a specific signature: cheap money arriving at the same moment as a genuine technological breakthrough. The claim matters right now because the United States has spent 2025 and 2026 running both conditions at once. The Federal Reserve cut its policy rate three times between September and December 2025, bringing the target range to 3.50 to 3.75 percent, while hyperscaler capital expenditure on artificial intelligence infrastructure ran past a combined $400 billion annualized across Microsoft, Alphabet, Amazon, and Meta. The essay's warning is not that AI is fake. It is that a real revolution plus a debased unit of account produces a price signal nobody can read, and that when the reckoning arrives, the state will blame the technology rather than the money.

The Sixteenth Century Template

The pattern the Mises writers describe starts with the Spanish silver of the 1500s. Potosi opened in 1545. Over the following century, roughly 16,000 tons of silver moved from the Americas into European circulation. Prices in Spain rose something on the order of 300 percent across the sixteenth century, modest by modern standards but violent for an economy that had known centuries of stable coin. The silver did not arrive into a static world. It arrived alongside the printing press, oceangoing navigation, and double-entry bookkeeping spreading north from Italy. Capital chased the new. Antwerp and Amsterdam became financial centers on the back of technologies that were genuinely transformative and valuations that were not.

The template repeats with unnerving regularity. Britain's railway mania of 1844 to 1847 followed the Bank of England's easy discount policy after the 1844 Bank Charter Act; Parliament authorized 9,500 miles of new track in 1846 alone, roughly triple the mileage that then existed, and by 1850 railway share prices had lost about two thirds of their peak value. The 1920s produced radio, electrification, and the automobile alongside a Federal Reserve that expanded credit through the decade; RCA went from about $85 in early 1928 to 549 in September 1929 to 10 in 1932. The late 1990s produced the internet alongside a Fed that cut rates during the 1998 Long-Term Capital Management rescue and flooded the system ahead of Y2K; the Nasdaq Composite gained 85.6 percent in 1999 and fell 78 percent from its March 2000 peak.

In every case the technology was real. Railways did reshape Britain. Radio did reshape America. The internet did reshape everything. The bubble was never a referendum on the technology. It was a referendum on the currency.

Why Cheap Money Corrupts Technological Judgment

The Austrian argument here is more precise than the popular version. It is not simply that low rates make people reckless. It is that the interest rate is the price that coordinates production across time, and when a central bank sets it below the rate that savers and borrowers would have found on their own, the error concentrates in the most time-distant projects. Ludwig von Mises framed this in 1912 and Friedrich Hayek elaborated it through the 1930s: malinvestment does not spread evenly. It clusters in long-duration capital.

Nothing is more time-distant than a technology whose payoff is asserted to arrive a decade out. Discounted cash flow arithmetic makes this mechanical. A business whose cash arrives in year fifteen sees its present value swing enormously on the discount rate; a business whose cash arrives next quarter barely moves. Cut the long rate by 200 basis points and the speculative growth asset reprices far more than the cash-generative one. Investors experience this as a signal that the future has become more valuable. It is instead a signal that the measuring rod has changed length.

This is the specific peril the Mises essay names. Techno-asset inflation is not identifiable in real time by looking at the technology, because the technology genuinely works. It is identifiable only by looking at the money. And the money is the one variable the professional class of forecasters is institutionally disinclined to examine, because the institution setting it also employs a large share of the profession's economists.

The 2026 Ledger

Run the current numbers against the template. US federal debt held by the public passed $30 trillion during 2025. Interest expense on the federal debt crossed $1 trillion annualized in 2024 and has not come back down; it now exceeds defense spending. The Congressional Budget Office's baseline has deficits running above 6 percent of GDP through the decade in an economy at full employment, a peacetime configuration with no precedent outside wartime.

Against that fiscal backdrop, the AI buildout has become the largest private capital program in modern American history. Nvidia's data center revenue ran above $130 billion for fiscal 2025 and kept climbing. Microsoft, Alphabet, Amazon, and Meta have collectively guided to capital expenditure north of $600 billion for 2026. OpenAI's Stargate program with Oracle and SoftBank was announced in January 2025 at a headline $500 billion over four years. A meaningful share of this is being financed with debt rather than operating cash flow, and an increasing share is being financed through special purpose vehicles and vendor arrangements that keep the leverage off the balance sheets investors actually read.

That last detail is the tell. Vendor financing, where the supplier of the equipment also funds the buyer, was the defining feature of Lucent and Nortel in 1999 and 2000. Circularity in an investment boom is the mark of a market where the marginal buyer is no longer a customer with independent economics. It is a market where capital is abundant enough that the participants can manufacture demand for each other.

The Case Against the Bubble Thesis

The opposing view deserves a fair hearing, because it is not stupid.

The bulls, whose most disciplined representatives include the research desks at Goldman Sachs and the venture partners at Sequoia and Andreessen Horowitz, make three arguments. First, unlike 1999, the spending is being done by companies with enormous free cash flow. Alphabet and Microsoft generate more cash annually than most sovereigns collect in tax. This is not Webvan burning venture money. Second, the revenue is showing up. OpenAI's annualized revenue crossed $10 billion during 2025 and Anthropic's grew several-fold in the same period. Enterprise software companies are reporting real AI-attributed line items, not vaporware. Third, the Fed's balance sheet has been shrinking, not growing, since 2022. Quantitative tightening reduced holdings from roughly $8.9 trillion to under $6.6 trillion. It is hard to call this a monetary bubble when the central bank is draining reserves.

The bears, who include Michael Burry, various strategists at Societe Generale, and a growing chorus at the Bank for International Settlements, answer that the balance sheet is the wrong gauge. The relevant number is total credit, and the fiscal deficit has been supplying the economy with the stimulus the Fed withdrew. Treasury issuance skewed toward bills has kept short-term liquidity abundant regardless of the Fed's holdings. They note that depreciation schedules on AI hardware assume useful lives of five to six years for chips that are functionally obsolete in three, meaning reported earnings are systematically overstated. And they note that the free cash flow argument was made about Cisco in 1999, when it was profitable, cash-generative, and briefly the most valuable company on earth before losing 86 percent of its value.

Both sides are describing the same object. The bulls are describing the technology. The bears are describing the financing. The Mises essay's point is that the argument is unresolvable in those terms, because techno-asset inflation is precisely the condition where a real technology and a corrupt financing structure occupy the same asset.

The Statist Reckoning

Here is where the essay's title earns its weight. What follows the bust is not a return to sound policy. It is an expansion of the state.

The 1929 crash produced the Securities Act of 1933, the Glass-Steagall separation, and the Reconstruction Finance Corporation, but it also produced Executive Order 6102 in April 1933, which criminalized private gold holdings above five ounces and forced surrender at $20.67 an ounce before the dollar was revalued to $35 in January 1934. The state's diagnosis of a monetary crisis was that citizens held too much hard money. The 2000 bust produced Sarbanes-Oxley and a Fed funds rate cut from 6.5 percent to 1 percent, which seeded the housing bubble. The 2008 bust produced Dodd-Frank, $700 billion in TARP authority, and a permanent expansion of the Fed's balance sheet from under $1 trillion to over $4 trillion, an emergency measure that became furniture.

The pattern is consistent. The state creates the monetary conditions for the bubble, disclaims responsibility when it bursts, identifies the private sector as the culprit, and takes new powers as the remedy. Expect the same script when the AI cycle turns. The candidates are already visible: compute licensing regimes, mandatory model registration, capital requirements on AI-exposed lenders, and a strategic case for nationalizing or subsidizing domestic chip capacity that will be framed as competition with Beijing rather than as a bailout. The EU's AI Act, whose general-purpose model obligations took effect in August 2025, provides a ready template. The US Congress, which has produced nothing coherent on AI so far, will produce something fast the week after a major AI-linked credit event.

Bitcoin as the Instrument That Cannot Be Diluted

This is the case for Bitcoin, and it is not a hedge argument. It is a measurement argument.

Every episode in the techno-asset inflation sequence was made possible because the unit of account was elastic. Spanish silver, Bank of England discount policy, Federal Reserve credit expansion, and post-2008 quantitative easing all worked the same way: they added claims without adding goods, and the added claims went hunting for the highest-duration asset available. A monetary base that cannot be expanded removes the mechanism. Bitcoin's supply schedule is not a policy. There is no committee, no dual mandate, no emergency facility. The April 2024 halving cut issuance to 3.125 BCH per block; the next, expected around April 2028, cuts it to 1.5625. Roughly 19.9 million of the 21 million coins now exist. Nothing about a financial crisis changes that number.

That property is what makes Bitcoin useful for pricing a technological revolution honestly. If AI capital spending is genuinely productive, it should generate returns measured in a fixed unit. Priced in a currency whose supply grows with every deficit, an investment can appear to succeed while destroying real capital. Priced in a currency with a fixed terminal supply, the arithmetic is unforgiving. The Austrians have argued for a century that the business cycle is not an inherent feature of markets but an artifact of monetary manipulation. Bitcoin is the first instrument that lets an individual opt out of the manipulation without asking permission from a government that has demonstrated, in 1933 and repeatedly since, that it will confiscate hard money when hard money becomes inconvenient. Self-custody is not paranoia. It is the historically literate position.

What to Watch

Depreciation disclosure. Watch whether any of the major hyperscalers shortens the assumed useful life of AI accelerators in its 10-K filings. Microsoft and Amazon both extended server lives in prior years to boost reported earnings. A reversal would knock billions off net income and would signal that management no longer believes the assets hold value. This is the single most informative number in the cycle.

Private credit and AI vendor financing. The BIS and the Financial Stability Board have both flagged private credit opacity. Watch for the first sizable default at a neocloud operator, the GPU-leasing firms financed against hardware collateral. CoreWeave-style balance sheets depend on residual values for chips. If Nvidia ships a generation that halves the value of the prior one, the collateral evaporates faster than the loan amortizes.

The Fed's reaction function. If the FOMC cuts in response to equity weakness rather than labor market deterioration, that confirms the put is intact and the malinvestment continues. Watch the dissents. Watch whether the phrase "financial conditions" migrates from the minutes into the statement.

Fiscal dominance signals. Watch the ratio of net interest to federal revenue. If it approaches 20 percent, expect increasing political pressure on the Fed to cap yields. Any discussion of yield curve control, however academic, marks the point where monetary policy formally becomes debt management.

Bitcoin's behavior in the drawdown. In March 2020 and again in 2022, Bitcoin sold off with risk assets. The interesting question is whether the next equity drawdown produces the same correlation or a decoupling. A decoupling would mark the transition from speculative asset to monetary asset. That transition will not be announced. It will only be visible afterward, in the chart.

---

Source: Mises Institute