Boom, or Mirage? The same numbers, two opposite verdicts

 

In the first half of 2025, roughly 92% of U.S. real GDP growth came from a single category that made up just about 4% of GDP: investment in information-processing equipment and software. That breakdown comes from Harvard economist Jason Furman, using Bureau of Economic Analysis (BEA) data. Strip those categories out, and GDP growth for the period runs at an annualized rate of about 0.1%.

Pick up the first lens — the nominal one, which puts dollars spent and dollars recovered on the same ledger. The verdict is clear: the visible return on hyperscale AI buildout lags far behind the investment. Through this lens, that 92% is a column propped up by IOUs. Growth is borrowed.

Now switch lenses. Take the same compute spending and run it through the quality-adjusted framework the Peterson Institute for International Economics (PIIE) published this spring. Deflate by the collapse in inference prices, and the "quality-adjusted output" of that same compute grows by over 2,000% per year in both 2024 and 2025 — more than twentyfold. Through this lens, the 92% isn't borrowed growth at all. It's a boom that conventional accounting systematically undercounts, because the national accounts simply don't have a ruler long enough to measure it.

Same numbers. Two lenses. Two opposite conclusions. Some read the echo of the 1996 fiber-optic bubble; others read the Solow paradox dissolving on the eve of an electrical revolution. Whichever lens you stand behind, you'll likely read the same verdict — and the problem isn't the reader. It's the lens.

This piece doesn't take sides. It dismantles one thing: the conclusion is a function of the lens, not of the data itself. And the dangerous part is that the question disguises itself as an objective report requiring no choice at all.

Two rulers, measuring two different things

The BEA ruler has no dedicated AI deflator. AI hardware, software, and cloud services sit folded into existing product categories, deflated by indices built for technologies that moved far more slowly. BEA's own research papers concede that AI activity is hard to isolate in GDP statistics. So when you see a report showing a negative internal rate of return (IRR) on AI investment — the person reading that minus sign did nothing wrong. They have only this one ruler, and its markings were cut to the rhythm of a previous generation of technology.

The PIIE ruler asks a different question. Not "how many dollars in, how many out," but "how much more can one dollar of compute do this year than last." Per-token prices fall roughly 97% a year — about a 35× efficiency gain — so a dollar buys more than ten times the inference it did a year ago. Strip out that "price decline" component, and the remaining "quality-adjusted output" explodes exponentially.

Each ruler is internally coherent. The reader who sees a negative IRR through the BEA lens and the reader who sees twentyfold growth through the PIIE lens are looking at the same negative. Only the developing chemicals differ — and the photograph comes out reversed.

The third lens: turn the gun around

The first two lenses each have their blind spot. BEA measures who made money; its blind spot is who saved money — when AI passes its gains downstream as steep price cuts, that enormous consumer surplus never appears on its scale. PIIE measures how much capability grew; its blind spot is how much of that capability gets captured and paid for. Between capability and revenue, a bridge is missing.

But the third lens is the sharp one. It takes PIIE's own preferred lens and turns it back on the side PIIE would rather not illuminate.

PIIE quality-adjusts only the output side. Inference prices crash → quality-adjusted output soars → output is undercounted. But what if you symmetrically put the same lens on the input side — marking up the book value of capital stock, since a dollar of GPU compute this year is also several times stronger than last year's?

Total factor productivity (TFP) is a residual: total output growth, minus capital's contribution, minus labor's, with technological progress as what's left over. Quality-adjust capital as aggressively as you quality-adjusted output, and the marked-up capital stock balloons, swallowing the entire residual. The measured TFP then falls — possibly turning negative.

This isn't hypothetical. Studies of early-20th-century electrification dissected it precisely: aggressively quality-adjust capital, and the residual method reclassifies all "progress" as capital deepening, making TFP vanish in the data. "Technology racing, total factor productivity falling" isn't a paradox. It's an artifact of adjusting one side and not the other.

PIIE's authors know this. They label their estimates an upper bound, not a point estimate — a first approximation. That twentyfold figure holds only when every assumption lines up and you quality-adjust output alone.

So the third lens exposes not PIIE's math — the math is fine — but its framing choice: which side gets the lens, and which doesn't, is itself a silent inference.

History's reassurance comes with two conditions

You might be thinking: history has shown that even when micro-level investors are wiped out, the macro economy thrives anyway. Half of that is true. The other half rests on premises that don't hold for AI.

Railways. Britain's 1840s Railway Mania saw share prices spike then crash by some 60%, with hundreds of companies going bankrupt. Yet by Crafts's restatement of Hawke's estimates, the social rate of return on British railway investment was about 15%, against a private return of roughly 5%. In the very years micro investors bled out, society at large was making a fortune.

Electricity. Electrical equipment shipments surged, yet total factor productivity fell for decades — Paul David's "productivity paradox." Only after factories were rebuilt and organizational capital reorganized did TFP finally erupt in the 1920s. Decades separated the technology's diffusion from its appearance in the output statistics.

Canals. Most projects of the 1790s canal mania failed, ruining their investors. But the liquidated, cheaply reorganized waterways carried decades of cheap fuel to the first factories — becoming the physical foundation of Britain's Industrial Revolution.

All three cases prove the same thing: in the buildout phase of a general-purpose technology, negative micro IRR and macro prosperity are two faces of the same coin, separated only by time. But the "micro dies, macro lives" reassurance stands on two premises.

First, durable assets. Railway tracks and rights-of-way depreciate over decades or centuries; when an operator goes bankrupt, the right-of-way sells cheap and the trains keep running. AI? Under standard accounting, GPUs carry accelerated depreciation schedules of just three to five years, and each new generation crushes the value of the last; model weights have minimal salvage value against open-source substitutes. A bankrupt railway still runs trains. A bankrupt AI operation leaves behind chips that probably aren't worth powering on.

Second, natural monopoly. Once a canal was dug, no equivalent route existed before the railway; once rails were laid, geographic monopoly formed. AI runs in virtual space, unbound by geography — a frontier model faces every competitor on earth, including open-source substitutes at a fraction of the price. The money micro investors lose settles onto a track with no moat.

Durability and monopoly — railways had both, canals had both, electricity had them partly. AI? Under the currently observable industry structure, the evidence for both premises is thin. Not necessarily absent forever — organizational reorganization, industry standards, state-level compute consolidation could each partly substitute for the old premises in the future. But that's a possibility with open variables, not a realized fact. Cover a structural gap with a variable, and the moment that variable shifts, the whole chain of reassurance has to be recomputed.

The lens's verdict, or yours

Stack the three lenses. The first says growth is borrowed — blind to who saved money. The second says the boom is undercounted — blind to the unbuilt bridge between capability and revenue. The third says even the framework claiming to fix those blind spots is making a silent framing choice of its own.

History's reassurance is withdrawn too. "Micro loses, macro wins anyway" is well documented — but it depends on durable assets and natural monopoly, and AI holds neither.

So if you ever believed "this boom is real," or "this growth is borrowed" — the question worth asking may no longer be which number is right. Across the lenses we've swapped layer by layer, the same numbers never once delivered a verdict.

And the version you believed — did the data tell you that, or did the lens you chose?

Three Years on Paper

 

How AI's capital boom quietly rebuilt a structure the telecom bubble already lived through once.

At some point, the numbers on a contract stopped being a solemn promise and became something that could be taken back at any time.

Our generation grew up trusting the signed document. Once you signed, you had handed over a stretch of your future — handed over, with no coming back. But spread a few of the latest prospectuses side by side, and that old faith quietly begins to loosen.

One number, two faces

The filing is explicit: a frontier model company pays roughly $1.25 billion a month, locked in for three years — about $45 billion all told. Grand and imposing, as if the next three years were nailed firmly to the page.

Scroll down a few lines, though, and another sentence sits there quietly: either party may terminate upon 90 days' notice.

It isn't an isolated case. Between a major search company and the same compute provider, the headline reads roughly $920 million a month — and deeper in, the same 90-day exit.

The $45 billion carries the fanfare; the 90 days keeps the exit. One speaks to the market; the other is only talking to itself.

Curiously, not every contract looks this way. When the large cloud operators stop being tenants and become buyers of compute, the terms harden at once: five-year terms, totals of nearly $10 billion or even over $17 billion, around 20% prepaid, and termination granted only if the counterparty fails to deliver. The buyer who wants out cannot leave.

So a counterintuitive picture emerges. The end with the most uncertain returns and the thinnest foundation — model companies that have yet to prove they can earn steadily — signs the contracts that are easiest to walk away from. The end with the most cash and the most room to maneuver locks itself into years of minimum-payment obligations. The ones who should leave can't; the ones who could leave anytime can. Hold that mismatch — we'll come back to it.

A borrowed balance sheet

The deeper layer hides in three words: who owns this?

The chips that power training and inference don't sit on the user's books. A private-capital firm builds a special-purpose shell, buys roughly $35 billion of custom silicon in a staged structure, and leases it back. The hardware sits inside the shell; the user only pays rent; its balance sheet looks clean and light. Growth was borrowed — and now even the balance sheet that carries the growth is borrowed too.

This is no one-off cleverness but a pattern that has set. By one major rating agency's estimate as of year-end 2025, the largest cloud operators held about $969 billion in data-center lease commitments — roughly two-thirds of it, over $660 billion, not yet on their own balance sheets. Those off-book obligations alone already exceeded the adjusted debt they had publicly disclosed. The lightness on the books is real. It's just that the weight lifted off hasn't vanished — it has only moved elsewhere, and it's waiting.

One more point is worth pausing on. The party providing credit support for the largest senior tranche of that debt, per cross-reports from multiple financial outlets, turns out to be the very party that sold the chips. So the tranche is priced not on the young model company's own credit, but on the veteran chip supplier's investment-grade rating — and the interest cost comes down. The shovel seller backstopped the debt the shovel buyer took on to buy the shovels.

What that support is, legally, and which dollar it covers, the public filings don't yet make clear. So we can write only as far as "credit support was provided" — the rest stays blank. Yet those two words alone are enough to recall an old story.

An echo from twenty years ago

Around the turn of the century, telecom equipment makers did the same thing.

Fiber and switches were the hottest things going; demand seemed bottomless. To move their own switches and routers, several equipment giants simply lent money to carriers that couldn't afford to buy — letting them purchase the vendor's own gear with borrowed funds. On the books: orders, revenue, a growth curve climbing ever higher. Nine equipment makers carried roughly $25.6 billion in customer guarantees.

Then the tide went out. The carriers couldn't pay, and those receivables once booked as "revenue" came back one by one into the makers' own laps. Nortel alone reported about $2 billion in bad debts and write-downs in its 2001 losses. They too had been certain demand would always be there — and that certainty didn't survive a single winter.

Set then against now and it's nearly the same story in a more respectable suit: from the telecom closet to the data center, from coaxial cable to custom silicon, the structure unchanged. The only difference, perhaps, is that this time the balance sheet is thicker — thick enough to defer the loss longer, long enough for everyone to have time to believe this time is different.

And that "longer" is precisely what should worry us most.

The quietest irony

Lay the three pieces together: a wispy contract you can exit in 90 days; a multi-year lock-in with the door welded shut; books made light while the weight waits off-balance-sheet — an arrangement already rehearsed in full twenty years ago. Three things that should unsettle. Yet the unease almost always lands in the wrong place.

People watch the end that looks most fragile. But precisely because it's fragile, visible, and quick to unwind, its economic loss and its accounting loss arrive almost together — nowhere to hide. The losses buried deepest and longest sit inside the contracts called "rigid" and "rock-solid": payments keep flowing, the books look unchanged, impairments held year after year beneath the waterline — until one quarter, when a contract isn't renewed, they surface all at once.

People lose sleep over the first kind, then hand their money, quite comfortably, to the second.

What have we been pricing all these years — a business that's genuinely growing, or an arrangement engineered carefully enough to defer the reckoning, but bound to wake?

The question doesn't need answering in a hurry. Only, next time you see those grand numbers that nail years to the page, it may be worth a little more patience — to turn to the lines further down, and see whether there, too, a quiet sentence sits, saying something only to itself.

One Order, Two Readings

 

5:21 in the afternoon. A letter.

On June 12, 2026, Anthropic said it had received a directive from the U.S. government, issued under national-security and export-control authority: its two most capable frontier models, Fable 5 and Mythos 5, were to stop serving any foreign national at once — including the company's own foreign employees. To comply, both models went dark for everyone that night.

By most accounts, it was a first: a government directly shutting down a publicly released, running commercial AI model. Not a fine, not an after-the-fact investigation — just a sentence: now, switch it off.

And quietly, it flipped over the word everyone had been using for months.

The sovereign premium, inverted

For half a year, the market had been paying frontier AI companies an extra markup. The logic: once a model is strong enough to assess systemic financial risk or surface national-grade cyber vulnerabilities, it becomes part of national security — the government needs it, protects it, walls out rivals. Its valuation floats free of ordinary software gravity, up into something called the "geopolitical premium." The deeper the government is embedded, the thicker the premium.

The letter is the same fact, seen from the other side.

The government really is deeply embedded — deep enough to zero out your most valuable product line in a single afternoon. The very capabilities that make you "sovereign-grade" are not the deepest part of the moat; they are the trigger. Put plainly: when the government can protect you, it simultaneously holds, at no cost, the power to switch you off. In the language of finance, that's a free put option.

The sovereign premium and the sovereign kill-switch are two faces of one relationship. You paid for the first. Almost no one has booked the second.

One order, three onlookers

The named party sees collapse. Rewind to 2020: SMIC was added to the U.S. Entity List, its advanced-node processes hit with a "presumption of denial." Even as plenty of export applications were later approved, the market kept it inside a permanently steeper geopolitical-discount frame. That discount stuck to the valuation and would not peel off.

The compliant survivor sees a gift. Under the same Entity List, TSMC — the advanced-node leader with no comparable geopolitical constraint — absorbed the high-end customers flowing out. One order falls; the named party carries an indelible discount, the unnamed one collects the migrating clients. Same shadow, two opposite faces.

But there's a third onlooker — and this is where AI differs from every license story before it. Spectrum, casinos, banks are bolted into the ground. An AI model is software: it can be routed around, swapped out, designed out. In 1999, after a satellite-tech leak, the U.S. pulled commercial-satellite export authority back into the stricter munitions framework. The result wasn't a domestic windfall — Europe rolled out "ITAR-free" satellites built specifically to bypass U.S. parts, and the U.S. global share slid from roughly 51% to about 41% over the following years. Control too hard, and global customers don't wait around — and swapping a model is far easier than building another satellite. The "survivor premium" you meant to defend can simmer down into a discount of self-marginalization.

Three onlookers, three readings. One order, three contradictory photographs.

So — premium, or discount?

The honest answer: we don't know yet. But between knowing and not knowing sits one very specific variable — whether a named exemption appears.

If the next comparably capable model gets a whitelist with a name on it, the market will read it as selective licensing and pay survivors a premium. If it gets switched off with no exemption, the market will redefine the whole field as "revocable at will" — killing multiples first, then financing.

So the thing to watch isn't "will there be a second shutdown," but "did the first comparable case get a written, named exception." That's the earliest lamp to light up at this fork.

It hasn't lit yet — because two load-bearing cards are still face-down. One is the prospectus: both Anthropic and OpenAI have filed confidentially, the risk sections unpublished. How they word "our flagship could be switched off at any time," and how they disclose conflicts with major shareholders, will decide whether the market prices them as high-growth software or as a license that can be voided overnight. The other is the $35 billion financing structure that pledges chips and leases them back — its contract terms aren't public, and how rent, debt and guarantees behave under a regulatory shutdown is nowhere to be found.

With the cards face-down, you don't write the ending.

A darker thread

The shutdown, per multiple reports, was partly triggered by an Amazon security study — a finding that the model could be jailbroken for cyberattacks, escalated to the White House. Keep this within the evidence: officials never named Amazon, the wording was "partly," and at least five companies had voiced concerns. It's not a verdict; it's a strong but unconfirmed thread.

Still, it's sharp. Because Amazon wears three hats here: Anthropic's major shareholder, its cloud provider, and its competitor. When such a company's "security finding" can pass through the machinery of the state and become an order to shut down your product, "safety evaluation" stops being only a tool of public governance — it can also be an interface for rearranging market rank. The thing prized as a moat — "the strongest safety capability" — flips, from this angle, into an attack surface.

Yet the one who swings the blade rarely walks away clean. After Huawei's 2019 Entity-List addition, Ericsson was supposed to inherit 5G share — but once Sweden moved to ban Huawei, Ericsson promptly warned of backlash in China, its China revenue share falling from around 11% to about 3%. The party expecting to collect the spoils paid a price too.

The question, handed back

Back to that order, delivered at 5:21 in the afternoon.

It looks like a regulatory headline, one company's bad week. But zoom out and it asks: when a company's most valuable asset is no longer the model it trained, but the permission slip for who it's still allowed to sell to — are you pricing a business, or a license? And the issuer of that license also holds the power to void it.

The fork has opened; it hasn't closed. Anyone who's already written the ending — "frontier models are finished" or "winner takes all" — is reading a version that comes not from the event, but from the side they already wanted to believe.

The next order like this will come. It will be disguised as an ordinary headline, so you'll think it has only one reading.

Which side will you read it from?

Track Change, or Exit

 

In early 2026, a batch of U.S. federal spending figures looked like a migration. Homeland security spending more than tripled year over year. Health-sector AI appropriations nearly sextupled. Total federal AI investment rose close to eighty percent. Line those numbers up and anyone would write the same sentence: public finance is pivoting toward artificial intelligence.

But switch the lens, and the same figures point in only one direction: retreat.

Not a single number changed. What changed was the ruler used to measure them. The conclusion has never been a function of money — it is a function of the lens.

---

The tsunami was produced by division

The federal ledger accumulates by fiscal year, leaving a snapshot each month. But reporting lags — a payment can take days or weeks to land in the books. A snapshot taken in the first few days of a month is mostly empty.

Compare that seven-day snapshot against a fully settled month from the prior year: the numerator is nearly hollow, the denominator full, and the growth rate shoots through the roof. The tsunami was not produced by surging money. It was produced by division.

At the time, the people watching these signals had nothing but that seven-day snapshot. Reading a surge under that lens was not obtuseness — it was what the data honestly showed under that methodology. We can call it an artifact today because we later waited for the reporting period to settle and performed a line-by-line drill-down. Spreading out cards that only developed later, then laughing at someone for misreading the hand, is the cheapest kind of hindsight.

Three lenses, three layers deep

The first examines the numerator. Switch from "latest month" to "settled months aligned on both sides," and that more-than-double homeland-security tsunami falls to a 3.8% decline. The direction reverses entirely. That astonishing growth rate had a June barely seven days old sitting in its numerator.

The second examines the ruler itself. When the ledger first pulls a month, if it captures a partial value, that value freezes permanently — never updated even as real money flows in later. So even the "comparison baseline" carries a defect. That seemingly credible –28.2 that appeared to have already corrected things? Itself an artifact. Unfreeze and align, and the true value turns out flat. When the ruler is crooked, the more carefully you calibrate, the more confidently you are wrong.

The third examines the classification — and this is the real mechanism. It does not measure where money flows. It measures how many times "artificial intelligence" appears in an application form. A state applies for rural healthcare funds and, to make the proposal look modern, writes in "AI-assisted diagnostics," "telehealth platform," "intelligent triage." The keyword scanner fires, and the rural-reform money gets filed under AI. Whether you win depends on keyword density. This is not a ledger — it is a lottery.

The first two lenses expose inflated numbers; that is old news. The third exposes a classification that does not hold — and that is the real foundation of the track-change illusion.

Drilling through: the same money, misidentified twice

Three seemingly independent "AI net additions," traced to the deepest layer, all converge on a single agency — the Centers for Medicare & Medicaid Services — and a handful of its existing programs. The load-bearing piece: a rural healthcare reform package totaling over $970 million split among five states (Kansas, Georgia, Idaho, Washington, Maryland). It had nothing to do with AI.

Of those five, Georgia, Idaho, and Washington were flagged not only by the AI lens but again by the cybersecurity lens. Combined: $586 million, same appropriation ID on both lists, amounts matching to the cent. The same money for fixing rural hospitals, counted once under AI and once under cybersecurity — a pure case of mistaken identity, filed twice.

Yet the lens was not indiscriminately mislabeling everything. The NSF's $53 million in computer-science grants also appeared on both lists, but legitimately — computer security genuinely straddles both fields. Right and wrong can be cleanly separated: 91% of the double-counted total falls on the single rural-reform line.

The money did not move — the label did

The third lens works by quietly equating "mentioned AI in an application" with "this is AI spending." The real causation runs in reverse: states packed their rural-reform applications with trendy keywords, and the lens flagged them. Not a cent moved; only the name changed.

Writing that way was a rational act under existing rules, not grant fraud. The question is not about the applicant — it is about a ruler that treats "mentioned" as "did" and "wording" as "flow of funds." The problem has always been in the classification mechanism, not in the people being classified.

Under settled, auditable figures, in all probability no net new public dollar going toward AI can be found. Loosen the variable — let seven-day snapshots and literal keywords count — and the track-change picture appears instantly. What was uncertain was never the direction of the money, but which lens you are willing to trust.

What the settled ruler actually measures is plain to the point of being dull: science-foundation funding dropped by more than half; defense AI procurement retreated; the homeland-security tsunami never crested. What declined truly declined. What was flat was truly flat. The few entries that looked like gains, once drilled through, turned out to be neither AI money nor new money — just old rural-hospital funds, flagged once by each of two lenses.

---

The same numbers, three lens changes, two opposite conclusions. Under one lens, public finance charges toward AI with blazing momentum. Under another, it is quietly exiting, and the only steps that looked like forward motion were names mistakenly assigned again and again.

This spectrometer can tell you what color each lens refracts — which cell was produced by dividing an empty month, which by a frozen baseline, which by application wording passing itself off as fact. But it will not flip that cell for you.

The next batch of reports will come. The next impressive growth figure will not carry a footnote reading "I was produced by dividing an empty month." Every time, that same question of which lens to choose will reappear, disguised as a fact that requires no choosing.

If "public money is surging toward AI" is a sentence you once believed, then what is truly worth asking may no longer be which number is right or wrong. Under every settled, auditable methodology examined here, that net addition never appeared. And the version you once believed — from which lens, exactly, was it read?

The Invisible Tax

 

For two years now, everyone has been waiting for a tipping point—waiting for AI to steal office jobs, for entry-level positions to go dark one by one, for a visible moment you can point to and say, "That's it."

But the bill already arrived. It didn't make the evening news; no one held a press conference; no one stepped forward to claim it. It traveled through the supply chain—from a silicon wafer, to a production line, to a memory chip quietly swapped out, all the way to the phone you paid for. While everyone stared at layoff notices out front, a receipt in the back office had already been silently handed to those who can least afford it.

---

AI's appetite, these past two years, craves only one thing: high-bandwidth memory. Data centers need it, accelerator cards need it, every expansion of every model needs it. So the three giants that nearly monopolize global memory made the most rational decision imaginable: redirect their best wafers and most expensive packaging lines to the highest-margin premium memory, and leave ordinary people's standard memory in the margins of the fab.

Those margins went empty. That emptiness is the vacuum.

The numbers are not gentle. The consumer-facing share of standard memory was slashed thirty to forty percent in a single year; some fabs shifted as much as seventy to eighty percent of capacity to premium. Once supply drained, prices came off the leash—standard memory contract prices jumped ninety percent in one quarter, then sixty the next; flash storage followed in lockstep at seventy percent per quarter.

Regular viewers will recall the previous chapter: the biggest hidden hand behind global inflation wasn't wages or the money supply—it was electricity consumed by data centers, an energy shock that quietly drove up the marginal cost of every AI inference. That was a bill hidden upstream, buried in the electric bill. This time the direction reversed: not cost-side pushing up, but supply-side pressing down. Price hikes flowed like water, trickling down the supply chain into the phone you just bought. One pushes upstream toward cost, the other downstream toward the end user. The same externalization, a different direction.

---

Pushing costs to the end of the line—to those who can least afford it—is not new.

Three years ago, during the auto chip shortage, limited chips went first to the highest-margin pickups and luxury SUVs; thin-margin entry-level models simply had their lines shut down. Ford cleared out the Fiesta, killed the EcoSport; Tesla's promised $25,000 car was shelved before it ever rolled off the line. The ones squeezed out of the new-car market were always the thinnest wallets—they flooded into used cars and drove those prices up twenty to forty percent themselves.

Go back another half century. The 1973 embargo and the 1979 revolution pushed real oil prices up more than sixfold in eight years. In the winter of 1978, the elderly and the poor poured thirty to fifty percent of disposable income into the hole of basic energy costs. President Ford, urged to impose a gasoline tax, saw clearly how it would crush the poorest—and vetoed it; by some accounts, he even broke with the advisor who proposed it. Half a century later, the same regressive pattern is still playing out. Only this time, no one is willing to give it a name.

---

This bill is collected two ways. One overt, one covert.

The overt kind is a price hike. Average phone prices climbed from $440 past $510; Sony, Nintendo, and Microsoft all quietly abandoned the old art of the mid-cycle price cut, choosing instead to squeeze profit from the machine you already own.

The covert kind is a spec downgrade. Same price, but memory shrinks from 16 GB to 8. Entry-level phones revert to 4 GB, as if time went backward. The most candid example: a laptop giant hiked the starting price for 16 GB by $400, then released an 8 GB version to bring the entry price back down—except that machine no longer meets the company's own memory threshold for an "AI PC." It admitted as much in its own press release.

Why is the covert tax more insidious? Because it bypasses the alarm circuit in your brain. Psychology quantified this threshold long ago: we don't notice something shrinking until it passes roughly eight to ten percent. So manufacturers calibrate each reduction to land just below that tripwire. The same product shrinks; the unit price rises twelve percent; yet sales climb six percent the following year—because no one noticed. The overt tax faces hearings; the covert tax can't even produce evidence. That is precisely its advantage.

---

Who gets taxed first?

In India, sub-$100 ultra-low-price phones saw sales plunge nearly sixty percent in a single year—not because people didn't want to buy, but because memory price hikes squeezed out every last drop of margin. Manufacturers simply gave up. Analysts passed a cold verdict: even if prices come back down, this tier has permanently lost its commercial viability. A market of 170 million units was quietly erased.

In Europe, phones under €200 shrank to a record-low quarter of shipments. In the U.S., the cheapest prepaid phones are being pushed off shelves one by one, while the premium iPhone edges up against the trend. The poor man's Android is retreating; the rich man's flagship is advancing. Shipment volumes fall, yet total market value rises—sell fewer, sell pricier. That is the entire story of this year.

When the bill comes down, it is never evenly split. The first to lose a phone is not the office worker worried about being replaced by AI. It's the person who saved for months just to buy the cheapest smartphone—who won't even know that buried in the price, a portion was subsidizing a data center on the other side of the planet.

---

Inside the vacuum, it's not as if no one stepped in. A wave of low-cost newcomers seized the vacant space, ramping up volume at a stunning pace—a rare pressure-relief valve.

But the gap filled, and prices kept climbing.

The steel industry had one ending: old giants ceded the lowest-end rebar, and a newcomer built on scrap steel seized that tier, ate its way upward, and drove century-old mills into liquidation—small eats big. The display panel industry had another: old giants ceded large-size LCD, and the newcomer who took over monopolized that tier but toppled no one, leveraging concentrated capacity to control volume and prop up prices. Incumbents held the top-tier technology gap; newcomers locked down the commodity end. Each stayed in its lane.

This round of standard memory, so far, looks like the latter. The vacuum was filled, but what's been rewritten is market share, not price—the hardest technology gap at the top, the premium memory that feeds AI, remains untouched. The newcomers fill share; they can't rewrite the playbook. Unless, one day, the backfiller earns a ticket to the top tier—only then will this landscape be rewritten.

---

Why is everyone waiting for "white-collar job loss" but no one looking at "the poor can't afford a phone"?

Because the former is a natural story to tell. It has a clear face—a laid-off programmer, a shuttered office, a frozen job posting. Its causal chain is short enough for a single sentence: "AI stole the jobs." The latter is the exact opposite: diffused across billions of receipts, each too small to mention, its causal chain absurdly long—wafer to production line to memory chip to price tag—so long that people just lump it into the basket of "everything's expensive lately." No one to blame, no one to speak for it. It sinks into the noise of inflation, an unclaimed levy collected in the back office.

For a covert tax to detonate, someone first has to name it. France's Yellow Vests sparked from a fuel tax hike—but once it was framed as "the urban elite dumping the bill on the rural poor," the fury burned nationwide. California's famous tax revolt was the same: specific property tax bills reading "you're about to be forced from your home," plus an organizer who welded scattered complaints into a single slogan.

Right now, no one has named it.

---

Last episode argued that much of this boom's growth was borrowed, with the bill passed to the future—externalization along the dimension of time. This episode is about the same bill heading a different direction: not only passed to the future but also pushed downstream, to those standing right now at the very end of the supply chain, those who can least afford it—externalization along the dimension of space. One borrows from time; the other presses on space. One makes tomorrow pay for today; the other makes the poor pay for the rich.

The problem was never that AI produces no political consequences. The problem is that the consequence it produces first may not be the one the political system is willing to acknowledge first.

This invisible tax will most likely not be refuted. It will be quietly miscategorized—filed under "inflation," "brand strategy," "consumer downgrade"—and forgotten. Until one day, some price event loud enough yanks it out of the noise, and people pause: so that's what we've been paying for all along.

And until it is seen, it is being paid.