Clocks Without Hands
Economic Cycles, Gave's Energy Physics, and the Eternal Return of the Infrastructure Bubble
A Sequel to the Death Notice
In a previous essay on these pages, I argued that the recession had been quietly euthanized — killed not by competence, but by political intolerance. Central banks anesthetized the business cycle. Fiscal deficits replaced organic demand. Volatility was outsourced to the balance sheets of tomorrow.
But declaring the recession dead is not the same as declaring cycles dead. They are not the same thing.
A recession is a symptom. A cycle is the underlying physiology. And physiology does not negotiate with policy makers. It operates on its own timeline — indifferent to the Federal Reserve's dot plots, unmoved by congressional stimulus packages, and entirely unimpressed by the consensus forecast at Davos.
This essay is about the cycles that survived the great suppression. Not Juglar's classic credit machine — which, as we saw, has been blunted by deindustrialization and permanent stimulus. But two older, more stubborn forces: the energy cycle, as articulated with unusual clarity by Charles Gave, and the infrastructure cycle, which has produced some of history's most spectacular manias and most useful wreckage.
These are not cycles you can cancel. They run on physics and human psychology — a combination no central bank has yet found a way to print against.
The Taxonomy of Time: Not All Cycles Are Equal
Before we go further, a brief cartography of cycles — not to be encyclopedic, but because the confusion between timescales is the most common analytical mistake.
The shortest cycle, described by Joseph Kitchin in 1923, runs roughly 40 months. It is a creature of inventory management: firms overshoot production, warehouses fill, orders collapse, layoffs follow. Clean, mechanical, and increasingly irrelevant in a service economy where the "inventory" is often a line of code or a streaming license.
The Juglar cycle — named for the French economist Clément Juglar, who described it in 1862 — is the classic "business cycle" of 7 to 11 years. Its engine is capital investment and credit. Expansion builds optimism; optimism invites leverage; leverage funds projects of declining marginal quality; eventually, the credit machine reverses, and the liquidation phase destroys the weakest positions. Alvin Hansen identified roughly a dozen such cycles in the United States between 1837 and 1937, averaging just over 8 years. The mechanism is not subtle. It is almost biblical in its regularity: expansion, temptation, overshoot, reckoning.
Beyond Juglar, Nikolai Kondratiev postulated a much longer wave — 40 to 60 years — tied to technological revolutions and the construction of new production paradigms. The steam engine. Electrification. The automobile. The internet. Each wave begins with genuine innovation, scales through credit and institutional support, reaches structural saturation, and gives way to the next order. Schumpeter formalized this as "clusters of innovation." What matters for our purposes is the timescale: these are generational rhythms, not tradeable cycles. You position for them, you don't time them.
And then — largely absent from the academic literature but present in the work of practitioners like Gave — there is a cycle that sits between Juglar and Kondratiev: the energy cycle. Approximately 25 to 30 years. Driven not by human decisions but by the physics of resource extraction. This is the cycle that, paradoxically, the most modern economies remain the most vulnerable to. Because energy is not software. You cannot refactor it at marginal cost. It has weight, it has lead times, and it has the patience of geology.
The Pig Cycle and the Price of Oil
Charles Gave's contribution to macro-financial thinking is not always rigorously academic — and he would probably take that as a compliment. But among his most durable insights is the application of the "cobweb model," familiar to agricultural economists as the "pig cycle," to the energy sector.
The original cobweb: when pork prices are high, farmers breed more pigs. But pigs take time to grow. By the time the expanded supply reaches market, prices have already fallen. Overproduction follows. Low prices discourage breeding. Supply contracts. Prices rise again. The cycle is self-reinforcing precisely because the supply response is delayed relative to the price signal.
Transposed to oil, the mechanism is identical — but the timescale is an order of magnitude longer. The decision to drill an exploratory well, develop a deep-water field, or ramp up shale production and the moment that production actually reaches the market are separated by years, sometimes a decade. The industry cannot react quickly, because geology does not respond to quarterly earnings guidance.
Gave's schematic is worth walking through carefully:
Phase one. Energy is cheap. It has been cheap for years, perhaps a decade. Producers, burned by previous losses, have cut investment. The financial community has declared fossil fuels uninvestable — on ESG grounds, or on the thesis that peak demand has arrived, or simply because the last cycle ended in value destruction. The result: underinvestment compounds quietly while demand continues to grow.
Phase two. The gap between constrained supply and growing demand becomes impossible to paper over. Energy prices begin to rise. This is not a geopolitical event. It is arithmetic. The rise accelerates as traders reprice the scarcity.
Phase three. Energy prices double, triple, or more. And here is where the transmission mechanism bites: energy is not a discretionary input. It is foundational. A doubling of energy costs compresses margins across the entire non-energy economy, erodes household purchasing power, and produces the classic stagflationary signature — inflation and stagnation simultaneously. The Juglar cycle, already suppressed by fiscal stimulus, finds a new avenue for its destructive impulses.
Phase four. The price spike finally triggers investment. Exploration budgets expand. Rig counts rise. New fields are developed. Shale fracking responds, given its shorter lead times. After several years, the supply surge arrives — often coinciding with a demand slowdown — and prices collapse. The cycle resets.
Gave has observed this pattern in long-run historical data on the S&P 500 / oil price ratio, identifying major troughs in the early 1920s, around 1950, in 1980, and around 2010. The periodicity is roughly 25 to 30 years — not a law, but a regularity that has been stable enough across different geopolitical regimes and monetary systems to warrant serious attention.
What matters for today: we entered the 2020s after a full decade of energy underinvestment. ESG mandates, collapse of the oil price in 2014–2016, COVID-era demand shock, and the political toxicity of fossil fuel investment all reinforced the pattern. The setup for Phase Two — the supply gap — was not canceled. It was deferred. The structural inflation argument in the previous essay rests, in part, on precisely this foundation.
Infrastructure Manias: When Capitalism Goes Evangelical
If the energy cycle is slow, geological, and indifferent to human enthusiasm, the infrastructure cycle is its opposite: fast on the upside, loud, exuberant, populated by genuine visionaries and spectacular frauds, and somehow always ending in the same place — wreckage, recrimination, and infrastructure that outlasts its builders.
The pattern is consistent enough across two centuries that it is almost embarrassing. An innovation arrives that is genuinely transformative. Early projects prove the concept. Returns are extraordinary. The financial press declares a new era. Capital floods in, leveraged by credit that bears no relationship to realistic demand projections. Promoters issue stock. Parliamentarians approve hundreds of schemes simultaneously. At the peak, valuation is based not on discounted cash flows but on the sheer momentum of collective belief.
Then demand disappoints, or capital markets tighten, or both. The weakest projects collapse first, then the second tier, then even some of the first movers. Stock prices fall 80% or more. Bankruptcies cascade through the supply chain. Investors lose fortunes. Politicians hold inquiries.
And then — here is the twist that redeems the whole spectacle — the physical infrastructure that was built during the mania survives. Often purchased for pennies on the dollar by opportunistic buyers, it goes on to become the backbone of the next era's economy.
The Canal Mania of the 1790s in England is the first legible example in the modern era. The success of the Bridgewater Canal — which dramatically reduced the cost of transporting coal to Manchester — demonstrated a genuine productivity gain. Between 1790 and 1793, Parliamentary authorizations for new canal companies exploded, with capital authorized rising from roughly £90,000 to nearly £2.8 million — an increase of more than thirty-fold in three years. Many of these projects never paid a dividend. Several were never completed. The promoters prospered; the shareholders, often small savers seduced by promises of 10% returns, did not. But the canals that survived became the circulatory system of the early Industrial Revolution.
The Railway Mania of the 1840s was the same story at a larger scale, with better PR. In 1846, at the peak of the frenzy, Parliament authorized 263 to more than 270 new railway companies, representing approximately 9,500 miles of proposed track. Roughly a third of those lines were never built — the victims of insolvency, fraud, or the simple exhaustion of capital that followed. Recent academic analysis of this episode (including a 2025 SSRN working paper examining the dynamics of the boom and bust) distinguishes two phases: a first, arguably rational repricing as liberalization genuinely expanded the potential of the railway network; and a second, speculative phase driven by momentum and the entry of less-informed retail investors. The bust was then triggered by a combination of enormous capital calls, the Irish famine, and monetary tightening. A familiar sequence.
The Telecom and Fiber Optic Bubble of the late 1990s replayed the same template in the register of the information age. Driven by genuine revolution — the internet was real, and its implications were genuinely transformative — operators and new entrants invested hundreds of billions of dollars in backbone networks, undersea cables, and fiber infrastructure. The demand projections assumed infinite and exponential growth in bandwidth consumption. The financing structures were creative, to put it diplomatically. When the bubble burst in 2000–2001, entire companies evaporated. WorldCom, Global Crossing, 360networks — the graveyard is long.
What survived? The dark fiber. The millions of kilometers of optical cable, installed at enormous cost, sat underutilized for a decade. Then, as streaming, cloud computing, and mobile video arrived, that overbuilt infrastructure became the physical substrate for the next phase of the digital economy. Netflix runs on the ruins of Global Crossing. Every Netflix subscriber is, in a small way, a beneficiary of the irrational exuberance of 1999.
This is the insight formalized by the economist Carlota Perez in her work on technological revolutions: each great wave of capitalism follows a two-phase structure. First, the installation phase — dominated by financial capital, characterized by speculative excess, and marked by the physical construction of the new infrastructure. Then, after the inevitable crash and the repricing of assets to realistic levels, the deployment phase — where productive capital takes over, and the economy learns to actually use what was built during the mania. The bubble was not a failure of capitalism. It was its delivery mechanism. Expensive, wasteful, and apparently indispensable.
The AI Data Center and the Ghost of Dark Fiber
We are, right now, in the installation phase of the next great infrastructure cycle. The object of desire is artificial intelligence: its training infrastructure, its inference capacity, and the electrical grid required to power it all.
The numbers are, by historical standards, extraordinary. Capital expenditure commitments by the major hyperscalers — Microsoft, Google, Amazon, Meta — are estimated at $391 billion for 2025 alone. Projections for total AI-related infrastructure investment through 2030 range from $5 to $7 trillion. Data center construction is consuming a meaningful fraction of global copper and steel output. Power purchase agreements are being signed at scale not seen since the deregulation of electricity markets in the 1990s.
Several observers have drawn the parallel to the telecom bubble explicitly. The structural homology is hard to ignore: genuine enabling technology, proof-of-concept established by early movers, financial conditions (until recently) accommodating, and demand projections that assume not current utilization but the utilization of a future that has not yet arrived.
The counterargument — and it deserves respect — is that the monetization timeline for AI infrastructure is considerably shorter than it was for fiber optic cable. A GPU cluster generates revenue within months; a transatlantic cable in 1999 required a decade. The deployment cycle may be faster this time. Carlota Perez's framework does not prescribe a fixed lag between installation and deployment.
But the risks are familiar. Concentration of investment in a small number of dominant providers creates a fragility in the supply chain for chips and power. Electricity demand from data centers is growing faster than grid capacity in most Western markets — producing the uncomfortable irony that the digital economy, designed to be weightless, is colliding with the most physical of constraints. And if the revenue from AI applications does not scale proportionally with the CapEx, the second phase of the infrastructure cycle — overcapacity, margin compression, and debt restructuring — will arrive on schedule, regardless of the enthusiasm of the participants.
The "dark fiber" of this cycle may be GPU capacity rented at prices that cannot be sustained. Or it may be a grid of renewable energy that was built to power the AI economy, but whose intermittency limits its utility. Either way, the infrastructure will outlast the bubble. That is both the consolation and the historical record.
Reading Cycles as an Investor
The practical implications of this framework are not subtle.
The energy cycle, by Gave's template, suggests that a decade of underinvestment in hydrocarbon production creates the structural conditions for a price squeeze — regardless of the near-term demand outlook. The current price weakness in oil (the Brent curve near $60 at the time of writing) does not refute this; it is precisely the kind of price environment that deepens the underinvestment and sets up the next spike. Commodities exposed to energy — base metals, fertilizers, energy services — are cheap on a structural basis for exactly the same reason. The pig cycle is not over. It is in its under-breeding phase.
The infrastructure cycle counsels selectivity rather than abstraction. The error is not to invest in infrastructure during the installation phase — some of those investments are generational. The error is to confuse the quality of the underlying technology with the rationality of the valuation. Railways were transformative. That did not prevent railway stocks from losing 80% of their value. AI is transformative. That is not the same as saying Nvidia at 40x revenue is cheap.
Carlota Perez's framework, applied forward, suggests that the deployment phase — when the infrastructure built during the current mania becomes the affordable substrate of the next economy — is likely to be deeply productive for companies that consume AI services rather than build them. Just as the most valuable companies of the 2010s (Airbnb, Uber, Spotify, Netflix) were the ones that monetized the cheap bandwidth and cloud capacity bought out of the telecom wreckage, the most valuable companies of the 2030s may be those that leverage commodity AI infrastructure at marginal cost.
One structural conclusion threads through all three cycles: real assets, which are the physical expression of cycles, cannot be made obsolete by monetary policy. You can suppress the credit cycle with QE. You cannot suppress geology, or the physics of electricity, or the fact that a fiber optic cable, once laid, exists. In a world of fiscal dominance and structurally higher inflation, the bias toward tangible claims — commodities, energy, infrastructure — is not a trade. It is an ontological position.
The recession may have been canceled. The cycles it expressed were merely deferred. And deferred cycles, like deferred maintenance, tend to arrive with compounding interest.