Crackpots #54: Perpetual Motion Machine (Patreon)

Hey everyone, it's Liz. Let's do this.

The first time someone laid fiber-optic cable across the United States, they buried more glass than the future really required. In the 1990s, flush with funny money and the promise of an infinite internet, telecoms raced to trench conduits beneath highways, rail lines, and riverbeds. Then the dot-com bubble burst, bandwidth demand hit the breaks, and a lot of that cable ended up just sitting there. Whole networks went dormant, miles of pristine fiber sealed in polyethylene, waiting for a signal that never came. We call it dark fiber: infrastructure built in anticipation of use, now a cautionary tale of over-investment, proof that capital’s most optimistic visions of connection can easily get ahead of themselves.  

But the dark fiber remained. It became a kind of geological layer beneath the digital economy, proof that capitalism’s preferred mode of building is to overbuild: to anticipate value before it exists, to throw a bunch of shit into the future and pray it sticks. Every boom since has followed that same logic, whether energy pipelines, shipping ports, data centers or entire social platforms. Each one is a bet that tomorrow will arrive fast enough to justify the debt of today. Beneath this runs a confidence that scale itself can bring about use, that the creation of connection will organize its own demand.

Now, in the age of AI, the fiber is glowing again. The same companies are trenching, building entire server campuses the size of towns, with their own private gas-fired power plants set to keep the lights on. We hear of efficiency, of compute, of the coming leap in productivity. But the deeper faith is temporal: that by spending quickly and massively enough—by layering one dark network ontop another—they can finally escape the lag between investment and return. Dark Fiber is now the master metaphor for the dream of the present: that value can travel faster than time.

Marx begins Captial Vol. II with a deceptively simple question: how does capital move? Not how is it produced, but how does it circulate, how does value reproduce itself through time. He writes that capital’s motion is always a circuit, a looping of money into commodities then into production and back into money again. But no loop is instantaneous. Every stage (production, transport, then sale) takes time, and during that time value is inconveniently frozen. The measure of success isn’t simply how much surplus labour yields but crucially how fast it can be realized.

This is why he spends so much time on depreciation—the slow wearing out of machines, buildings and tools—as the key to the system’s internal clock. Fixed capital, he writes, “does not give up its value all at once, but piecemeal.” Each cycle transfers a small portion of the machine’s value into the product while the rest stays fixed in the means of labour. That portion is then set aside as a fund to replace the machine once its spent, accumulating as quick as it can but still not fast enough to keep up with the deterioration of the machines. Value then becomes split between the present and a speculative future.

Marx called this the problem of turnover time, the rate at which capital complete’s its circuit. If turnover slows, accumulation follows; if it accelerates, profit multiplies. “The more often capital can be turned over,” he wrote, “the more surplus value it produces in a given period of time.” Capital has to move faster and faster to sustain itself but the stuff of the material world — machines, workers, energy, geography — gets in the way. The result is a permanent war on time.

David Harvey takes this as the essence of modern capitalism, recasting Marx’s analysis as a theory of space-time compression, what amounts to a constant effort to annihilate distance and delay through technology, logistics and finance. Capital, he writes, “must overcome barriers by spatial and temporal fixes,” pouring investment into various networks promising to speed up circulation. These temporary fixes solve immediate crises, sure, but only by creating deeper ones: every new channel of acceleration demands a heavier front-load, greater debt and the faster realization of value to service it. The entire economy, then, is a machine for borrowing against the future.  

Harvey's temporal fix reads prescient against the landscape of data center buildouts and global AI roadmaps. As each new wave of technology pushes the circuit further bit by bit, time compresses between production and realization. Nick Land would of course mythologize this drive as destiny, imagining capital itself becoming an artificial intelligence racing to escape the drag of reality. But the contradiction was there from the start: acceleration may shorten turnover but it can never abolish it completely. Every escape ulimately depends upon the time it denies. All those VCs who fetishize Land's mania mistake this treadmill for transcendence, celebrating speed without understanding the constraint that defines it.

The new infrastructure of AI advertises itself as frictionless: liquid, instantaneous, continuous. In reality, it looks more like the circuit under stress or a system that can’t pause without losing value. So the trillion dollar AI build out spends to stay in motion, each model update a minor act of arbitrage, time borrowed from the future to cover the present gap between investment and return.  

The numbers are staggering. Google, Meta, Microsoft and Amazon have spent more than a hundred billion dollars on capital expenditures in a single quarter. According to The New York Times, the data center developer QTS—owned by Blackstone—recently securitized $3.46 billion of its own facilities, bundling them into commerical-mortgage backed securities. Other hyperscalers are turning to the same menu of instruments: corporate debt, private credit, off-balance-sheet vehicles. Meta’s new Louisiana site is financed through a $30 bullion special-purpose entity designed to keep debt off the books, while Elon Musk’s xAI is structuring a $20 billion SPV to buy and lease Nvidia chips back to itself. Bank analysts now talk of opportunity in the A.I. CMBS market—not sure that requires additional comment.

In the 19th century, borrowing opened up new worlds of production. Now it simply keeps the existing one on life support. Each new round pledges the same asset, the same bet that more speed can pass for growth. Maybe it can. The securities flooding the AI market are funding circulation — the space between hype cycles and the patience of investors — not actual creation. Capital no longer borrows to build, it borrows to stay in motion. We have an economy financed, refinanced and then refinanced again, a loop of deferral we mistake for expansion.

Is this that big of a deal? Yes and no. The state, after all, keeps the tempo. “Capital,” Marx wrote, “is command over unpaid labour yet to be performed.” Today that command extends into time itself: policy calendars, energy targets and fiscal cycles are increasingly set to the refinancing schedules of capital. When OpenAI suggested that Washington might need to “backstop” its trillion dollar build out, people rightly got pissed off, but the tech companies are just saying the quiet part out loud. We already socialize risk through subsidies and tax credits, not to mention an increasingly loose regulatory apparatus, while pretending not to. The state already steps in — not to plan, but to maintain liquidity.

Above all that financial circuitry runs an even faster one: the equity market. The AI build-out is financed less by debt than by stock itself. When Microsoft or Nvidia add a trillion in market value, that paper wealth transforms into working capital. Stock-based pay, acquisitions and cheap credit is secured by the confidence the hyperscaler’s valuations confer. The market isn’t rewarding investment so much as it is enabling it as rising valuations convert expectation into liquidity, giving these companies the space to spend years of projected profit in advance. The state, in turn, uses this as proof so when the Nasdaq climbs, confidence in the AI Economy hardens into real world policy: more tax credits, more deregulation, more and faster permitting. Everything works in tune to protect these valuations because they now anchor everything from employment to retirement to national prestige. The market sets the tempo, deciding how long the economy can keep pretending to build up into the future.

But data centers need power faster than the grid can deliver it, so the industry builds its own. Open AI and Oracle’s Stargate complex have built up their own natural gas turbines in West Texas while Elon’s Colossus sites have their own dedicated plants. The Wall Street Journal calls it a “Bring Your Own Power” boom, like a reversion to 19th century practice where industrial titans generated their own electricity to get ahead of the state that couldn’t keep up with their insatiable demand. But even these fixes can’t erase time, just rearrange it. All those turbines need permits and crews and eventually replacement. Every attempt to abolish delay just rebuilds it elsewhere.

Marx treated depreciation as more than a simple accounting convention. Each machine, he wrote, “transfers its value to the product in proportion as it loses its use-value.” All this sounds very orderly, but it never stays that way. Technological change ensures that what Marx calls moral depreciation—obsolescence through innovation—outpaces physical decay. So the machine still works, but its value has diminished because a faster one exists. “All fixed capital,” he warns, “becomes a fetter upon production as soon as it has ceased to transmit its full value.” Depreciation in this sense, is the mechanism by which capital devours its own future. Each round of investment shortens the life of the round before it. Machines devalue machines, speed devalues speed. The very progress that drives productivity simultaneously erodes the basis of accumulation. Capital has to replace itself quicker and quicker, piling up debt and excess capacity in the hope that realization can keep up with obsolescence. When physical turnover hits its limit, credit steps in to stimulate: pretending the decay has been overcome and tomorrow’s profit can pay for today’s expense. But the bill always comes due and this fix only defers crisis by entrenching a deeper one. Harvey puts it succinctly: “The problem of capital is not that it cannot produce; it is that it cannot rest.” The machine has to keep spinning faster just to stay still.

The credit cycle and the depreciation cycle now share a schedule, each keeping the other alive. When one fails, the other follows. When that happens, the state will have to backstop both—including the investment and the ensuing collapse it was designed to prevent. Nowhere is the divide in approach to this problem sharper than between the United States and China. Both are building the skeleton of artificial intelligence but they are working no incompatible clocks.

In the United States, capital chases speed through finance. The future is collateralized and resold through SPVs, CMBS, off balance sheet debt—all devices for turning tomorrow into liquidity today. China runs at a slower pace, not because it avoids debt but because its time is administered. The state coordinates depreciation through policy banks and five-year plans, rolling over losses inside a bureaucratic calendar instead of a market one. The abundance liberals and Andreesenite futurists mistake the spectacle of megaprojects for frictionless acceleration when it’s really a planning horizon extended through state control of energy and infrastructure. As Adam Tooze notes, China now has roughly half again the PV capacity needed for a global Net-Zero path—a surplus that depresses prices and locks in dominance. The velocity that Americans envy is actually the visible face of patience, time administered at scale.

While American capitalism treats time as something to be cleared—a cost to minimized or latency to be erased—the Chinese model treats it as an asset to be administered. Each system organizes duration according to its own metabolism: one converts delay into speculation, the other folds it into planning. What we are watching isn’t simply competition for markets or territory but a contest over how the future should be timed.

America still bets on the instant—on finance, narrative, the liquidity of attention—to outweigh China’s command of material continuity. China wagers that the long rhythm of infrastructure, the slow addition of cables and kilowatts, will outlast America’s speculative heat. For now, the two coexist uneasily, like metronomes beating at different speeds in the same room.

But capital hates parallel clocks. Its metabolism requires synchronization, and the only way to do that is for the fast one to simply eat the slower one. That is the American project now, exporting tempo: we can’t build a power plant without a securitization scheme but we can beam our culture anywhere and everywhere at the speed of a push notification. China may export steel but we’re selling time, the style of living the world has learned to envy in which speed is treated as freedom and attention is the only currency that never debases.

As the FT writes, that logic of speculation now governs communication. The same equations that price volatility in markets now structure the flow of news and politics. Platforms understand that division and emotional whiplash hold attention the longest and after enough time, the public agenda reorders itself around those returns. The effect is more drift than persuasion, attention getting pulled toward whatever moves fastest until every subject competes on the same plane of reaction time. In that environment, perception itself adopts the logic of the feed, with each user moving through their own privately timed loop. Gestures get recorded and recycled and refined to elevate the next level of engagement. The feed predicts the next moment from the last, turning experience into data and data back into experience. But what we lose is the feeling of any shared tempo, as each life unfolds on its own schedule of refresh. Taken together, those loops have become a new civil sphere. The feed is its infrastructure: vast, continuous, almost entirely unmediated. What once passed through parties or unions now moves directly between the user and the state. Civil society hasn’t withered, it’s been rewired into an always-on network that converts reaction straight into politics.

Dylan J. Riley describes fascism as the product of “overdeveloped civil societies”—societies dense with organizations but incapable of collective coordination. That seems to be whats happening now, only digitally. Civil society hasn’t thinned, it has metastasized. A vast, professionalized middle class reproduces itself through an endless ecology of institutions—universities, NGOs, startups, media platforms—all claiming to mediate between ordinary citizens and the state. Social media is just the newest organ, a real time network of associations that performs connection while actually diffusing it. The result isn’t atomization but saturation. Too many publics, each demanding immediacy, none actually able to hold time.

In his conversation with John Ganz, Riley noted that Trumpism drew strength from this excess. It offered relief from the exhaustion of mediation, the fantasy of direct action unburdened by institutions. That fantasy has since become the norm. Governments now govern in the idiom of the feed: representation reduced to presence, decision to update. The problem is that the same middle class that sustains these networks cannot abolish them without abolishing itself; its power and identity are built from the institutions that now stand in for politics. What Riley calls “authoritarian democracy” names this adjustment—power that justifies itself through constant motion, borrowing its legitimacy from the very mediating class that was once meant to restrain it.

Civil society was once the buffer that slowed the state, now it's the engine that drives it. The institutions that were meant to support pluralism—media, nonprofits, universities—have all converged on one main goal: to keep participation continuous, producing a politics that mistakes ceaseless activity for life. And so the feeds keep perception moving at the market's pace: brief, volatile, always updating. Attention comes to match this tempo, thought follows a beat behind. Life begins to feel like a squence of tiny wagers, each scroll or swipe a bet on the next hit of recognition. A person raised inside this asylum feels life as a run of small, stochastic thrills: continuous, forgettable, hopefully profitable. 

The same dynamic that turned financial markets into a perpetual game of roulette now manages orindary experience. Work, intimacy and even belief have reorganized around the logic of the wager—minimal effort, maximum variance, the hope of a sudden jackpot that justifies continued play. In our casino culture, attention and desire are both chips on the table. The meme-stock trader or the influencer, they all live on the same clock: a string of micro bets chasing volatility for its own sake.  

The thing is, desire—slow, relational, durable—takes time. It needs absence, risk, the slow work of approach and refusal. But the feed can't afford that. What it can sell is arousal, instant and scalable, easily quantifiable. We learn, then, to accept stimulation in place of wanting. Gooning is the laboratory form of this. In Harper’s, Daniel Kolitz calls it “a bleak, and virtually inevitable end point for the lonely and poorly socialized internet generation.” People sit in blackout rooms surrounded by screens, edging for hours, their bodies “feedback instruments for stimuli.” Its all very grotesque but it is at least precise: the frictionless delivery of neurochemical reward while getting rid of everything that makes desire human: delay, otherness, the possibilty of failure. 

On device AI will close the loop, delivering infinite personalized erotica with endless mimetic variation. Every fantasy, all at once, at zero marginal cost. What once required encounter or imagination will be completely satisfied by algorithmic substitution. Scarcity disappears, and with it the need for want. Our bodies learn the tempo of the network and we call it pleasure just as our minds refresh instead of remember. We become perfectly adapted to the temporal regimes that produce us, efficient cogs in a machine that transforms our attention into value. Withdrawal, the drift toward solitude that breathless reports in The Atlantic now register as epidemic, is not moral failure. It's the rational adjustment of people to a world where stability no longer pays, where jobs vanish and institutions thin out and the only viable exits are the algorithmic jackpots that promise to change everything overnight. Arousal survives here because it can be measured and optimized. Desire withers because it cannot.

Traits that once implied depth like patience, restraint or discretion are all liabilities now. To stay visible—that is to stay valuable—inside a system that constantly updates, the self has to update with it. Personalities harden into performance and sincerity folds into a style of presence. What we used to call shamelessness is now a kind of savvy fluency. The successful convert exposure into protection, keeping emotion constantly in circulation. Shame is too slow and hesitation costs reach. The world now rewards whoever can cycle affect the fastest—quick: outrage to irony to contrition without pause. This is what charisma passes for under collapsed time.

You can see the same logic reflected in the state. Institutions built to plan now perform like traders—reactive, data driven and addicted to short term movement. Trump is really the prototype. A literal casino man, he learned early on how to turn risk into theatre and how to leverage debt into proof of vitality. His personal style became a kind of political grammar: each outburst or Truth’d decree or insane executive pivot keeps the wheel spinning buying more time. Planning can’t exist when the moment demands improvisation and so the government behaves like the market it once claimed to regulate, running on volatility unable to rest.

This is why everyone is so shameless now. Embarrassment only halts action when the moment demands acceleration. When nothing is disqualifying everything is possible. The endless feed of scandal then becomes its own source of power. This all began as a personal adaptation but now it’s a mode that defines entire institutions: bureaucracies tweet, departments chase meme cycles, diplomacy only wins if it goes viral. 

All the figures that dominate the moment—the influencer, the populist, the craven careerist—are variations of the same type, shaped by the environment that sustains them. They move easily across worlds, fluent in volatility, their constant improvisation less a style than a survival skill. Each one performs the function the network demands: visibility without pause, emotion without continuity, a willingness to live as content. The line between private ambition and public office has narrowed. All the old mediating structures have collapsed into a single plane of instantaneous address. Everyone speaks to everyone all the time, and in that saturation the distinction between governing and going viral disappears.

So politics is now direct-to-consumer. The state competes on the same field as the rest of us for the same scarce resource—attention—and must update itself just as often. The populist dream has been realized: representation without delay and participation without mediation. We call it democracy but maybe it's an overgrowth, an ecosystem of constant feedback where we mistake visibility for consent and speed for popular mandate. 

The rhythms of production and media and governance have all merged, each burning through the time it needs to survive. The same loops that organize attention now organize labour and policy and the state, the market and the feed all pulse in unison. Democracy at this speed and saturation turns fascistic, and acceleration becomes the method of rule: an economy, a politics and a culture all measuring vitality by how little delay they can endure.

Thomas Friedman calls this condition the polycene: a world where “everything interacts in real time.” This is capitalism narrating its own exhaustion as triumph, mistaking simultaneity for progress. The dream of instant valorization—of value realized before it's made—has found its theology. But beneath the optimism, the machinery keeps overheating: the grids, the networks, the minds. Everything hums at the edge of capacity, synchronized to a clock that can't be stopped, only broken. Time is being rewritten everywhere as an accounting fiction but the wear is real even if the books insist on perpetual motion. They say eventually it will all line up again, and the correction will hit like crisis. But each collapse restores time only long enough to forget it. The contradictions continue to hold, nothing resolves. The system keeps time by simply running out of it.

xx

L