The Trillion-Dollar Capex Bet

24 Oct 2026 — the yield back under 5%, and Gartner’s estimate. The debt page brings its reading of the ten-year Treasury yield up to 22 October. The yield closed above Ruchir Sharma’s 5% line on 20 trading days in a row from 23 September, peaked at 5.31% on 5 October, and fell to 4.96% on 22 October as investors sold technology shares and bought government bonds. At that yield the debt page’s floor is back within $1B of this page’s $1.30T. The reading page gains an entry on Gartner’s forecast of 14 October: about $7T of AI infrastructure spending worldwide over 2026–30, $8.5T of revenue needed, no more than $3.6T expected, and less than half of that reaching the firms that own the data centres. Set beside this site’s figures, it agrees with the floor’s revenue per dollar of spending and finds the buyers’ bill cautious. Nothing on this page changed.

23 Oct 2026 — each builder’s revenue against its share. The pages for Alphabet, Amazon, Meta, Microsoft and Oracle each gain a chart of the company’s revenue growth since 2023 against its share of the floor, the share coming due a little more each year to the end of 2030. Growth is read two ways, as on the tracking page’s gap chart: all of it, and only the part above the company’s own pace before the AI spending began. The charts read the same figures as the gap chart, so they move with each set of results. No figure changes.

22 Oct 2026 — Meta’s third quarter. Meta reported on 21 October, a week before the date the site’s data feed expected. Revenue was $64.6B, up 26%, above its own range. Its capital spending, $34.9B, was more than the $32.8B of cash its business generated, so free cash flow was minus $2.1B, and it created a non-voting share class it says it expects to use to sell shares. The watch list, the gap and strain charts and the Meta page are updated. Meta narrowed its 2026 plan to $135–145B, so the five’s plans are now $842.5B, not $840B: the tracking page’s counter, the past-booms page and the buyers’ bill follow. Meta’s share of the five’s spending is now 17%, the bill’s first setting, so the bill at its starting values is $1.82T a year, from $1.83T. The floor does not change.

22 Oct 2026 — the IMF’s October figures. The IMF published its October forecast on 13 October, and the site now runs on it. The advanced economies produced $69.6T in 2025, not $69.4T, and the IMF now expects them to grow 2.1% this year, not 1.8%. Today’s pace is that forecast, so §6’s dial starts at 2.1%. Growth would now have to average about 3.4% a year to 2030 to pay for the bet, against 3.2% before, but the gap over today’s pace narrows from 1.4 points to 1.3. Total pay keeps its old pace from about 2.9%, not 2.7%. The wage pool stays at $25T. The IMF’s forecast for 2027 is 1.8%, so the dial may now start from an unusually strong year. The Growth Saviour, the FAQ and the reading list follow, and Ireland’s figure is now checked against its own statistics office, because the World Bank’s does not match Ireland’s national accounts.

16 Oct 2026, late evening — the buyers’ bill in the menu. The buyers’ bill moves from the Context group to the top of the menu, after Tracking, and takes the name the site already used for it; the page was called “What AI’s customers must pay”. It is the site’s second measure of the bet, not background. The floor stays the front page’s only measure, and nothing in either calculation changes.

16 Oct 2026, evening — a page for Anthropic. Anthropic joins the earnings pages, the first for a company that has filed nothing: every figure on it is what Anthropic told investors on 14 October, as the FT and Reuters reported it. It reads the lab two ways. As the builders’ customer, it spent about $9.9B on computing in the quarter, about 3% of the floor at that pace. As a seller to final customers, its $1.70 of revenue per $1 of computing is the first measured view of the middle layer that the buyers’ bill has to assume; applied to every firm in the middle it would give a bill of about $2.06T a year, against $1.83T at the defaults. Nothing in the model moves. When the public prospectus is filed, its figures replace these.

16 Oct 2026, later — the FT’s figures on Anthropic. The Anthropic row now carries what the Financial Times reported from the same investor day: an 11% adjusted operating margin, cash-flow positive, $1.70 of revenue for every $1 spent on computing, the last annualised revenue ($65B at the end of July) and SemiAnalysis’s estimate of about $74B since, and revenue from rival labs training on Claude cut back from 5–7% of the total. The FT and Reuters differ slightly on the second quarter ($11.5B against $11.6B) and the first ($4.8B against $4.7B); the row gives both, where it had called the difference a revision.

16 Oct 2026 — Anthropic’s third quarter. The watch list gains a row for what Anthropic told prospective investors on 14 October: revenue of $16.8B in the third quarter, up 45% from $11.6B in the second, and an adjusted profit with no figure given. The Financial Times reported it first and Reuters confirmed it from people who attended; the figures are press-reported, not filed, and the row says so. Reuters now gives $11.6B for the second quarter where its September report gave $11.5B, and the row notes the difference. Anthropic stays outside the BIS’s five, so nothing in the model moves.

13 Oct 2026 — Bain’s $6 trillion, beside the bill. The reading page gains Bain & Company’s Technology Report of 29 September, which asks $6T a year of AI revenue by 2031 to sustain $1.5T a year of infrastructure spending, on the rule that capital spending is a quarter of revenue. The entry sets that beside the buyers’ bill, $1.83T, and finds Bain 3.3 times larger: about 2.6 times for counting every builder in the world against the five, and about 1.3 times for asking four dollars of revenue per dollar of yearly spending where the bill asks 3.1. A chart draws what each account says would pay: Bain’s $1.8T of existing uses, $1.5T of new markets named with a figure and $2.7T with no source named, beside the floor and the firms in between, with the $1.2T software pool as a line across both. Where the two agree: today’s uses come to about that pool, and it takes about one extra point of growth. What Bain adds that §3 has no place for: a third pool, cost savings in the physical economy. All of it is Bain’s claim, and no figure in the model changed.

8 Oct 2026 — SpaceX’s bonds, after the $40B plan. On 6 October the FT reported, and Reuters confirmed, that SpaceX is in talks to borrow $40B more to buy Nvidia chips, on top of the $25B of bonds it sold in June. The next day the cost of insuring its debt against default reached a record of about 195 basis points, from around 110 when the contracts began trading in June, and the premium on its 2056 bond over the Treasury yield widened to about 2.4 points from the 1.75 it was sold at. The SpaceX page gains a dated section with the figures and their sources, and the debt page, whose §5 said no published figure yet showed the builders’ premium widening, now carries this one: 0.6 of a point on its second slider, about $21B a year on the floor if every builder paid it. From the same FT report, Oracle’s five-year insurance cost stands at 244 basis points, up 40 since the start of September. One rule changed: a figure reported by the FT, Bloomberg or Reuters now stands on that one report, where before every figure needed two sources that agree; the FAQ says so. All of it is press-reported, not filed. SpaceX stays beside the model, and no figure in the model changed.

5 Oct 2026, later that night — a glossary, and the FAQ brought up to date. A new glossary defines the terms this site made up, such as the revenue floor, the buyers’ bill and the capture rate, and the economic and financial terms the pages rely on. Each entry can be linked to on its own, and the terms the BIS coined follow the report’s own wording. It sits in the nav after the FAQ. The FAQ gains an answer on what the floor is and why customers pay more than it, and its answer on where the numbers come from now states the rule that every figure needs two sources that agree. One correction: the FAQ’s footer still gave the US as about $30T of $62T, with a purchasing-power share near 40%, after the answer above it had moved to $30.8T of $69.4T and 37%. It now matches. The glossary was written on 3 October and joins the site now, so its entries on the bet, the two pools and the growth dial follow §3 and §6 as they have stood since 4 and 5 October. No figure in the model changed.

5 Oct 2026, late night — six company pages point to the tracking page. The scoreboard moved from this page to Tracking the bet on 11 September: the clocks, the gap chart, the strain chart, the jobs chart and the watch list. Six pages written before the move still called parts of it the front page’s. The Apple, Amazon and SpaceX pages said so of the patience clock, the gap chart, the jobs test and the test set for Amazon. Five pages quote the growth of about 40% a year that the revenue floor requires: Alphabet, Microsoft, Amazon, SpaceX and Oracle. This page no longer states that figure, and the watch list does. Each of these sentences now names the tracking page or links to the part of it that it means. The Microsoft page’s sentence on 2027–29 points to the second clock, and “what the front page calls the depreciation wall” there became “what this site calls”. On the SpaceX page “the one it calls demanding” became “and treats as hard to keep up”, because no page uses that word now. The SpaceX page still says that the capital spending total is this page’s, which it is. Dates such as “lands on 7 August” stay, because each page is written as of its own results. No figure changed.

5 Oct 2026, night — past booms get a page. §5 has always said, in one sentence, that the BIS compares this boom with canals, railways and dot-com. Has there been a boom like this before? puts numbers on that. On size: the five’s plans for 2026, $840B, are 2.7% of what the US economy produced in 2025. The telecom build-out of 2000 was 1.4%, dot-com investment peaked at 4.4%, and Britain’s railway boom of 1847 at about 7%. On speed, which is the BIS’s own measure: the five’s spending was 2.7 times its 2023 level by 2025 and would be 5.4 times on their 2026 plans, where dot-com investment took five years to reach 1.8 times. The page also sets out what repeats and what does not. The main difference is the one §2 starts from: this equipment lasts about five years, and a railway line lasted a century. The share-of-the-economy measure is this site’s addition, not the BIS’s, and the page says so. No figure on this page changed.

5 Oct 2026, late evening — the Nvidia page’s jobs link goes to the tracking page. The Nvidia page ends by naming the results due next. One of them was the August US jobs report. The page linked it to the front page and called it the front page’s test of whether AI revenue and jobs are moving apart. That test moved to Tracking the bet on 11 September, with the rest of the scoreboard. The link now goes to the jobs chart there, and the sentence says “the tracking page’s test”. Its date, 4 September, stays, because the page is written as of Nvidia’s results of 26 August. No figure changed.

5 Oct 2026, later in the evening — the debt page’s chart names its bottom axis. The chart on the debt page was written to carry a title under its bottom axis, “ten-year Treasury yield”, below the numbers 3% to 7%. The title was drawn just past the chart’s bottom edge, where a browser cuts it off, so it has not been seen since the page went up on 11 September. The chart is now about 20 pixels taller and the title shows, on a desktop and on a phone. Nothing else in the chart moved, and no figure changed.

5 Oct 2026, evening — two strain readings on the Nvidia page dated. The Nvidia page said that Amazon’s capital spending is about 107% of its operating cash flow and Oracle’s about 174%, and it called the strain chart the front page’s. Both figures were the chart’s readings when Nvidia reported on 26 August. Since then Oracle’s results of 10 September moved its reading to 161%, and the chart moved to Tracking the bet on 11 September. The sentence now opens “When Nvidia reported” and links to the chart where it is. The Nvidia page reports one set of results, so its two readings stay as they were that day, and the chart has the latest. No figure changed.

5 Oct 2026, later again — the debt page’s chart, readable on a phone. On a phone, the chart on the debt page printed two labels on top of each other when its sliders were at rest: the one for the current point and the one for the front page’s floor. The amber line also ran through both. The point’s label now sits beside its point, clear of the amber line and of Sharma’s 5% line. On most phones it gives the floor alone, and the yield is read off the axis. The floor’s label is shorter on a phone and goes to whichever end of its line is free. In a few slider positions neither end is free and it is left out; the readout above the chart gives the same figure. Nothing changes on a desktop, and no figure changed.

5 Oct 2026, late afternoon — what higher spending would ask of wages, or of growth. The entry on the five’s borrowing on the reading page gains a second chart. The first drew the five’s spending reaching $5.7T by 2030 if it only holds at the level expected for 2027, against the $3.5T this page’s slider starts at. The second carries that through §4 and §6. At $3.5T customers must gain $5.2T a year by 2030, which is 65 million jobs’ worth of pay or growth of 3.2% a year. At $5.7T they must gain $7.8T: 98 million jobs’ worth of pay, with workers’ share of income falling from 60% to 50%, or growth of 3.8% a year, against 1.8% today. Any mix has to add up to the same sum. These are this page’s own sums run on a larger floor. The slider keeps the BIS’s range, and no figure on this page changed.

5 Oct 2026, afternoon — the builders’ borrowing, and the yield above 5%. The reading page gains an entry on how the five are borrowing, built on the FT’s Big Read of 28 September and read with the Bank of England, KKR and Apollo. The five have sold about $200B of bonds this year and are expected to sell more than $1T more, and lenders now charge them more than similar companies because nobody knows how much is still to come. Apollo’s chief economist says the lending rests on one forecast, that the five’s operating cash flow rises from $600B in 2025 to $2T in 2030; the strain chart’s own data gives $603B for 2025. The entry finds the §2 spending slider low against newer forecasts, and draws the higher path in a chart: with $1.3T spent or planned for 2025–26 and $1.1T expected in 2027, spending that only held level would pass the slider’s top of $4.5T in 2029 and reach $5.7T by the end of 2030. §2 now says so in a sentence under the slider. The slider keeps the BIS’s range, so the floor is unchanged. The entry also questions the wording of the debt clock, which is left as it is. The debt page records that the ten-year Treasury yield has crossed Ruchir Sharma’s line. It first closed above 5% on 16 September, has closed above it on all eight trading days from 23 September to 2 October, and stood at 5.28% on 2 October. The page’s sliders still start at the reading of 10 September, so its panel still starts at the front page’s floor; at 5.28% the floor is $1.31T, about $12B more. A new paragraph there sets out the Federal Reserve chair’s view that the builders’ borrowing is itself part of why the government’s rate has risen, and the case against it. The entry on FT Alphaville’s three estimates now says that Goldman Sachs’s chart and table are built on different amounts of spending. No model figure on this page changed.

5 Oct 2026, later still — a stale Oracle figure corrected on the tracking page. The note under the strain chart on Tracking the bet that explains the SpaceX flag still gave Oracle’s capital spending as 174% of its operating cash flow. That was the reading before Oracle’s results of 10 September. The chart, the note above it, the Oracle page and the debt page have said 161% since then, and the sentence now agrees with them. It also says two things those already say: the figure is about 244% without the money Oracle’s customers paid in advance, and Oracle has sold new shares as well as borrowing. No figure is new, and the chart did not change.

5 Oct 2026, later — the growth dial now runs to 2030, on the bet’s own numbers. §6’s dial ran fifteen years and assumed that capital’s income doubled. Neither came from the spending. The floor falls due in 2030, not 2040, and the doubling was an illustration this page had chosen, so the dial’s headline readings (workers’ share at 49%, wages keeping pace at 3.2% growth for fifteen years) rested on an assumption, not on the bet. The dial is rebuilt on the figures the rest of the page already uses. It takes the $5.2T a year that §4 says AI must be worth to its customers by 2030 and shows how much of it growth pays for and how much comes out of wages. At today’s 1.8% growth all of it comes out of wages and workers’ share falls from 60% to 53%; from about 2.7% total pay grows as fast as it used to, while the share still falls; at about 3.2% growth pays for all of it. The slider for capital’s income, the four small fifteen-year charts and the FAQ item on the doubling are gone. The reality-check chart’s dashed lines are now 2.7% and 3.2%. The BIS’s forty-year chart stays, as the long-run view. On The Growth Saviour, the button that extended the country chart fifteen years at 3% a year now shows the growth that would pay for the bet instead: 3.2% a year to 2030 and today’s pace after, beside today’s pace throughout, to 2035. No figure in §2, §3 or §4 changed.

5 Oct 2026 — the growth is stated as a level to reach, and by when. §3 and The Growth Saviour said the bet could be paid for by growth of 3.2% a year “until 2030”. That left two things unsaid. Only the level reached matters, not the path: the advanced economies would need to produce about $81.1T a year by the end of 2030, where today’s pace reaches $75.9T, and after that growth could return to its old pace. And the date matters, because the spending’s bill falls due then. A new section, If the growth comes late, works that out. Spread over ten years, the growth needed is only 2.4% a year, within the range of the studies the BIS surveys, but by 2030 it supplies under half of what the spending needs. Growth that comes late can spare workers and still leave the first investors unpaid. No figure on this page changed.

4 Oct 2026, later — the charts now fit a phone. On a phone every chart on the site was drawn wider than the screen and had to be swiped sideways, with nothing to show that it could be, so readers saw the left half of a chart. Each chart is now fitted to the width of the screen with its labels at normal size; a few labels are shorter on a phone, and where a legend already names the lines the names are left off the chart. The loop diagram in §5 is the exception: its boxes hold text, so it still scrolls sideways, now with a line saying so. Nothing changes on a desktop, and no figure changed.

4 Oct 2026 — growth is named as the other way the bet could be paid for. §3 said the revenue could come from only two pools, software budgets or wages. That left out a source that does not exist yet: growth. If AI helps people produce more instead of doing their work, the revenue can be paid out of new output and no wages have to go, which is the BIS’s own “productivity boost” scenario. §3 now says so and sizes it: for the builders to receive the floor, customers must gain $5.2T a year by 2030, which is 65 million jobs’ worth of wages, or growth of about 3.2% a year against 1.8% today, or a mix. The Growth Saviour’s second chart is rebuilt to show that trade; the one it replaces ran §6’s dial backwards. Three smaller changes. §4 now sets its 25% capture rate beside evidence (two published AI prices against the cost of a human-handled service contact, and William Nordhaus’s estimate of how little of an innovation’s value its makers keep), which suggests 25% is generous to the AI sellers. §6’s dial gains a slider for how far capital’s income rises, and says plainly that its doubling is an illustration of a full payoff, not a figure from the BIS or from the spending. And §5 corrects an attribution: “euthanasia of the labouring classes” is Nordhaus’s phrase, which Adam Tooze cites; the page had given it to Tooze. The FAQ has three new items on these points. No figure in §2, §4’s arithmetic or the dial’s defaults changed.

3 Oct 2026, night — the Growth Saviour. The economies page is renamed The Growth Saviour and gains a second chart that runs §6’s dial backwards. If capital’s income doubles over fifteen years, wages keep their historical 2% pace only if the 43 advanced economies grow 3.2% a year (§6’s wage-par rate), and workers’ share of income holds at 60% only if they grow 4.7% a year, as fast as capital’s claim. They averaged 1.9% from 2016 to 2025. With everyone else growing 1.5%, the US alone would need 6.1% and 10.2%. Four sliders change the doubling, the years, pay’s historical growth and the growth of everyone but the US. No figure on this page changed.

3 Oct 2026, night — the evening’s changes, combined. Two sets of changes made the same evening are now on the page together: the buyers’ bill moved to its own page, and the economy-wide figures moved to today’s economy ($69.4T) with the US weighted at 37%. The one figure that depends on both is the share of the white-collar wage pool, which is now $25T: §4 here reads $5.2T of wages, 21% of the pool and about 65 million jobs (the entry below says 24%, against the old $22T pool), and the bill’s page reads $7.3T, 29% and about 91 million. The jobs figures do not depend on the pool, so they do not move.

3 Oct 2026, later — what customers must pay moves to its own page. With two measures and a switch between them, the front page had become harder to follow than its argument. The floor, the BIS model, is again its only measure: the headline, §3’s drawing and §4’s jobs arithmetic are drawn from it, so §4 reads $5.2T of wages, 24% of the wage pool and about 65 million jobs again. The buyers’ bill moves whole to What AI’s customers must pay: its three sliders, the check against Anthropic’s accounts and what it leaves out. That page also reruns the jobs arithmetic on the bill, which gives $7.3T of wages, 33% of the pool and about 91 million jobs at its starting values. §2, §3 and §4 each say in a sentence that the floor is what the builders receive, that customers must pay more, and where that is worked out. The switch and the ?model=bis link are gone. No constant changed.

3 Oct 2026, later — the US counts for 37% of the group’s growth, not 45%. §6’s growth chart works out the line for everyone but the US by taking the US out of the IMF’s figure for the whole group. That needs the weight the IMF used to build the figure: each country’s output adjusted for what money buys at home, on which the US was 37% of the group in 2025 (IMF). The chart had used 45%, which matches no measure (in dollars the US is 44%). With 37%, the rest of the group grew between 1.0% and 1.8% a year from 2023, not 0.7% to 1.7%, and is forecast at 1.5% for 2026, not 1.4%. Keeping wages on pace now needs the US to grow about 6% a year, not 5%, if the others keep their current pace. The FAQ’s calculator starts at 37% too (it started at 47%), and its slider runs 30–50%. A correction on the way: the chart’s description said neither line reached 3.2% outside 2021, but the rest of the group grew about 3.5% in 2022, still part of the rebound after Covid; it now says so.

3 Oct 2026, later — the page starts from today’s economy, $69.4T, not $60T. Every economy-wide figure here rested on a round $60T for the advanced economies’ output, which is the IMF’s figure for 2022. The page now uses the IMF’s figure for 2025, $69.4T (World Economic Outlook, April 2026), the same one the new economies page breaks down country by country. Wages are 60% of it, so the pool of all wages becomes about $42T (was $36T), and the white-collar pool on the §4 slider starts at $25T (was $22T), with the slider now running $20T–$35T. §4’s jobs figure does not move, because it depends on revenue and pay, not on the pool; the share of the pool falls to 29% (was 33%). In §6, capital’s income now doubles from about $28T to about $56T (was $24T to $48T). The rates do not move: wages still keep pace at 3.2% growth, and each point of growth is still worth 7–8 points of workers’ share, because every figure in the dial scales with output. The dial’s readouts at 3% move by a point (workers’ share 49%, was 48%) because an unexplained constant is gone: capital’s fifteen-year claim had been set to $48.4T rather than exactly double $24T, and is now derived as exactly double. The BIS’s own model has no dollar figure for output, so its findings are unchanged; the 2022 base was considered and set aside, because it would need explaining everywhere it appears.

3 Oct 2026, later — §6 says its dollars are in today’s prices. The growth dial has always measured growth after inflation, and its dollar figures ($24T and $48T of capital income, $36T and $45T of wages, $93T of output) only make sense in today’s prices. The section did not say so, and a reader could take $48T as a future dollar amount, which at 2% inflation would be only about half as much again in real terms, not double. §6 now says it in the opening paragraph and beside the readouts. No figure changed.

3 Oct 2026 — what final customers must pay becomes the page’s default measure. On 30 September the page added the buyers’ bill beneath the floor and decided it should feed nothing. That decision is reversed. The floor counts what the five builders must receive; the bill counts what final customers must pay, once the firms in between, such as the AI labs, add their own costs and profit. The pools in §3 and the jobs arithmetic in §4 compare AI revenue with what customers spend, so for them the floor understated the requirement. The bill now sets the headline, §3 and §4, and a cautious one: its profit slider starts at 10%, the bottom of Goldman Sachs’s range, down from 30%, which gives $1.83T a year against the $1.30T floor (it was $2.29T). §4 now reads $7.2T of wages, 33% of the wage pool and about 90 million jobs, where it read $5.2T, 24% and 65 million. A switch at the head of §2 returns every figure to the BIS model, the floor alone, and a link with ?model=bis opens the page that way. The tracking page and the debt page stay on the floor, because they measure the builders’ own revenue and debts, and every reading recorded there remains comparable. The bill panel gains a check against Anthropic’s accounts, from its draft prospectus as reported by Reuters on 28 September, which is not an official filing and is labelled so, and a line saying what the bill leaves out: computing bought from builders outside the five, or built by the labs themselves.

2 Oct 2026, later — the jobs column charted. Tracking the bet gains the series the jobs row argues from: the change in US payrolls each month since January 2023, the date the gap chart measures revenue growth from, with the jobs gained or lost in information and manufacturing beneath. The chart says “United States only” on its face, for the reason the row gives: the bet is placed across the advanced economies, and US payrolls are where it should show first. Its figures are the BLS’s current ones, redrawn from the BLS each jobs Friday because the BLS revises earlier months, and dashed outlines keep the first estimates this site recorded on the day, from July 2026. No figure elsewhere on the site moved.

2 Oct 2026 — September payrolls, and July a loss again. September payrolls +29,000 against an average of 45,000 a month over the prior year; unemployment up to 4.2% from 4.1%; average hourly earnings +3.0% year on year. The same release revised July from +21,000 to −10,000 and August from +162,000 to +133,000, 60,000 lower between them, so the contraction the 4 Sept entry reported as revised away is back in the record. Manufacturing +9,000, up 72,000 since its December 2025 low; information −10,000 in the month and −120,000 on the year. The jobs row on Tracking the bet and its preamble are rewritten for the print. The divergence test points at the same fork, at the soft end: earnings beating against payrolls flat to weak, which is still the “tool” world on the scale §4 would need. A jobs print changes no gap-chart or strain data, so the row’s +$553B/yr, ≈$2.2T and ≈28M figures stand. Figures confirmed against the BLS release of 2 Oct and geoveu’s capture of the same morning. Next reading 6 Nov.

30 Sept 2026, later — a second measure in §2: the buyers’ bill. The reading entry added earlier today found that §3 sets the floor, which is what the builders must take in, against the software pool, which is what final customers pay, and carries nothing for the firms in between. The owner’s decision is to model that gap here, beneath the floor, and to keep the floor as the simpler, main yardstick. The new panel takes the floor from the sliders above it and asks what final customers must spend for the builders to receive it. A share of the builders’ computing is sold straight to the final customer. The rest is rented by firms that sell AI products, and those firms pass on their other costs and a profit. At the starting values (16% sold direct, computing at 75% of a product firm’s costs, a 30% margin) the bill is $2.29T a year, 1.76 times the $1.30T floor and 1.9 times the world’s business software spending. The method and the 75% and 30% are Goldman Sachs’s, from the table FT Alphaville published on 28 September. The 16% is Meta’s share of the five’s planned 2026 capital spending, because Goldman marks up every builder but Meta. Goldman’s dollar figures are not used: its table is built on two years of spending, the floor on six. The BIS models no such layer, so this is the page’s construction and the ledger classes it that way. Nothing else moved. §3’s drawing, §4’s jobs arithmetic, the tracking page’s required ramp and the debt page are still drawn from the floor, and the second measure feeds none of them. The three starting values and the software pool’s $1.2T are now registered constants, and the checker holds the panel’s figures to them. The ledger records the decision and the design that was not taken, which was to fold the mark-up into the floor itself.

30 Sept 2026 — three outside estimates of the floor. The reading page gains FT Alphaville’s piece of 28 September, which collects three sums published in the same week for the question §2 asks. Goldman Sachs: about $300B a year of AI revenue for the five to break even, and $636B for a 30% return. Stijn Van Nieuwerburgh of Columbia: $3.7T a year by 2032, for every AI data centre in the US. Jared Bernstein and Ryan Cummings: $2.4T to $3.8T over six years, for six firms. The totals are far apart because each counts a different amount of spending. For each dollar of capital they land close to §2: Van Nieuwerburgh’s method asks for 38.7 cents of revenue a year, the floor for 37. The entry also records two things the piece adds. Firms that sell AI products sit between the builders and the final buyer and need a margin of their own, so §3’s comparison of the floor with the software pool is kinder to the bet than it should be. And at today’s rents a data centre pays well, so the open question is whether prices hold as capacity grows. Van Nieuwerburgh’s paper, Bernstein and Cummings’ post and Dan Davies’s sum on the SpaceX lease were read at source. Goldman’s note is not public: its headline figures are confirmed against two reports of it, and its charts are read as the FT reproduces them. Every figure is the cited author’s claim, no constant changed, and the ledger records it that way.

22 Sept 2026 — productivity gets a row of its own. The watch list on Tracking the bet carried labour productivity inside the workers’-share row. It now has a row of its own, directly beneath it, from the 3 Sept BLS release: output per hour +1.4% annualised in Q2, from output +1.7% and hours worked +0.3%, +2.2% on the year, and 2.1% a year since the cycle began in Q4 2019 — the long-run rate, not a boom. The row exists because the measure assumes something the bet’s substitution fork removes: output per hour presupposes workers in the production, it rises whether workers produce more or fewer hours produce the same output, and at the BIS’s limit it stops meaning anything. So the row always states hours beside productivity. Written on 5 Sept and held on a branch until now; the figures are the same ones, confirmed then against the release (USDL 26-1434), the BLS API and geoveu’s own capture. No constant changed, and the row takes no slope, no gap-chart slice and no place in any total. Next reading 5 Nov.

19 Sept 2026 — a fourth clock, on the tracking page. The scoreboard’s three clocks are deadlines: when investors can lose patience, when the chips must be replaced, when the debt has to be borrowed again. Tracking the bet now carries a fourth that is a rate rather than a deadline, and the only one a reader can watch move: the five’s own announced capital spending for 2026, $840B between them, counting out at $26,636 a second while the page is open. It introduces no new figure. It is the same $840B that splits the gap chart’s required ramp, divided by the length of the year, and the constants checker now holds the counter, the three rates printed beside it and the chart to that one number, so none of them can drift from the others. What the panel says in its own words is what it is not: capital spending does not leave the door evenly, so this is the annual plan expressed per second, not money moving; it covers the BIS’s five and so leaves out Nvidia, which sells to them, and SpaceX, which builds beside them; and it is the bill being run up, not the larger bill it has to earn back, which is §2’s floor of $1.3T a year. With scripting switched off the counter is not shown at all, and the three fixed rates carry the panel on their own.

11 Sept 2026, night — will US debt burst the AI bubble? A page beside the model. Ruchir Sharma’s argument on the FT’s Economics Show of the same day (the government and the AI builders borrowing from the same savings; the ten-year Treasury yield as the number to watch, with 5% as the line) gets a page of its own, reached from the nav in the header. It sets his $800B funding gap beside the §2 floor (the same fact counted two ways), lays out his five-step route from the Treasury market to the data centre from the FT’s published transcript, puts his mechanism on a slider (the §2 return split into the Treasury yield and the premium above it, so the bond market moves the floor by about $35B a year per point at the default $3.5T), and reads the 10 Sept Oracle print as his fourth sign, over-leverage, ticking. The sliders rest at the front page’s defaults, so nothing here changed: the floor is still $1.30T at 10%. The only constants added are the yield reading (4.95% on 10 Sept, from the Treasury’s daily curve and two independent reports) and the two new sliders’ ranges, all in the ledger and held by the checker. The floor bar’s styling moved into the shared stylesheet, since two pages now draw it. Sharma’s own figures are his claims and feed nothing.

11 Sept 2026, later still — every page gets its own share card, and a wrong figure on one is corrected. A share card is the picture a chat app or social network shows when someone posts a link. Six pages had none of their own and borrowed the front page’s, so a link to the FAQ, the reading list, the tracking page, the ledger, Nvidia or Oracle all previewed as the front page. Each now has its own. Two existing cards were wrong. Meta’s carried −5% as its after-hours share move — Alphabet’s figure, not Meta’s, the same copy that was corrected on the page itself on 27 August and missed here, because the words on a card exist only inside an image and nothing on this site could read them. It now says −9.6%, matching the page. Microsoft’s capex note was too long for the box it sat in and ran over the one beside it, so the live card read “non-raiseshares +8% after hours”; it is shortened, with the same figures. Every card’s words now live as text alongside the site’s code and are re-drawn and compared before each publish, so a card that drifts from what it should say stops the publish rather than shipping. Text too wide to fit does the same instead of overlapping. No figure on this page or any other changed.

11 Sept 2026, later — the scoreboard moves to its own page. Everything that was §7 — the three clocks, the gap chart, the strain chart and the watch list — now lives at Tracking the bet, reached from the nav in the header, and this essay ends at §6 with a pointer. The owner’s reason: the argument and the record are different things, read at different times. Nothing in the scoreboard changed in the move. The one construction that could have drifted is the gap chart’s required ramp, which on this page followed the §2 sliders; on the tracking page it is drawn at the sliders’ defaults ($3.5T, five years, 10%, $250B of running costs, a floor of $1.3T a year), and the constants checker now holds those defaults equal to the registry and to this page’s sliders. The print-sync’s detector reads both pages. §7 is retired rather than reused, as §8 was, since the entries below refer to it by number.

11 Sept 2026 — plain language, every page. The prose of this page, the eight company pages, the FAQ and the reading page was rewritten so that a reader without a finance background can follow every sentence. Terms of trade are now explained where they first appear: capital spending, operating cash flow, free cash flow, the labour share, a company’s “guide” (its own guidance), “backlog” (orders signed but not yet delivered) and the like. Labels were renamed to say what they hold: the gap chart’s “Ceiling” view is now “Upper limit”, the clock chart’s “Depreciation wall” is “Chips need replacing”, the scoreboard’s third column is “What they mean for the bet”, and the nav’s “Reading beside the model” is “Further reading”. One factual slip was fixed on the way: the jobs row pointed “above” to the labour-share row, which sits below it. No figure, hedge, source, model or constant changed; the constants checker passed unaltered before and after. The 10 Sept Oracle print, folded in the same day from the trunk, got the same treatment in its row, in the §7 notes and on its page. Earlier entries below are left in their original wording, since they are the record.

10 Sept 2026 — Oracle FY27 Q1: the loudest demand reading yet, and the first equity sale. OCI +121% against +93% a quarter earlier, RPO to $664B, revenue $19.3B +30%, non-GAAP EPS $1.92 past a ~$1.74 street; FY27 revenue guide lifted to at least $90B, the FY27 capex guide held at $90–95B gross. Against it: $28.5B of capital expenditure in one quarter, more than half of all FY26 spending, free cash flow −$5.4B, and a completed $20B at-the-market share sale — the first of the BIS five to fund the build with equity. The watch-list row is rewritten and moved to the top of the table; the 10 June row it replaces survives in the companion page, refreshed for this print. Two charts moved. The gap chart takes Oracle to +$20.2B/yr above its 2023 baseline, and the strain chart takes its capex-to-operating-cash-flow ratio from 174% to 161% — a fall that is an artefact, and is now annotated as one. $11.4B of Oracle’s record $23.1B operating cash flow is customer prepayments for capacity not yet built; strip them, as Oracle itself does in its “net cash outlay” table, and the same quarter reads ≈244%. That prepayment is also the cleanest instance of §5’s circuit on the record so far: the customer financing the builder, counted as the builder’s operating cash. Both charts now carry ragged series by construction — Oracle’s August fiscal quarter lands a month before the other four report calendar Q3 — so the strain chart’s x-axis was extended a quarter, its end labels now follow each company’s own last point rather than a shared right edge, and the gap chart’s aggregate deliberately stays at the last quarter all five have reported. The aggregate actual line therefore did not move. The §1 and SpaceX capex-to-revenue comparisons are dated rather than overwritten: Oracle’s $0.86 per dollar of revenue was the June quarter, and the August quarter reached $1.47. Oracle sits outside geoveu’s capture, so every figure here is taken from the 8-K earnings exhibit filed 10 Sept, with the share move press-reported and stated as a range. Next reading ~mid-Dec.

4 Sept 2026 — August payrolls, and July’s contraction revised away. August payrolls +162,000, unemployment unchanged at 4.1%, average hourly earnings +3.1% year on year. The same release revised July from −23,000 to +21,000 and June from +20,000 to +31,000, so the “first contraction on this row’s clock” that the 7 Aug entry reported is no longer in the record. The jobs row and the watch-list preamble say so rather than silently carrying the new figure, and the two reading-page passages that quoted July’s −23,000 carry the revision inline. The divergence test still points at the same fork, more mildly: earnings beating against payrolls flat, which is the “tool” world. A jobs print changes no gap-chart or strain data, so the row’s +$553B/yr, ≈$2.2T and ≈28M figures stand. Figures confirmed against the BLS release (USDL-26-1435) and geoveu’s capture of the same morning. Next reading 2 Oct.

3 Sept 2026, later still — the labour share charted, and a channel the entry had missed. The BLS entry gains the series it argues from: all 318 quarters from 1947, with the 2013–2026 window enlarged beneath so the recent fall is legible at all. Both panels start at 52% rather than zero, and say so — the whole eighty-year range is thirteen points wide. The entry’s reasoning is also corrected. It had said the fall showed the share moving “for reasons that have nothing to do with an AI substitution shock”, which quietly treats not substitution as not AI. It leaves out the bargaining channel: what sets pay is the employer’s alternative to paying you, and a credible replacement moves that the moment it is credible — no-one need be displaced. The entry now names three channels it cannot separate, and the correction cuts against this page rather than for it. §4 has been treating displacement in the payroll data as the signal that starts the clock; on the bargaining channel the transfer starts with the threat and never reaches payrolls, so “no displacement visible” stops being reassurance. §6 already argues the other leg — a shrinking share drains labour’s leverage — so the two make a loop, not a sequence. No figure moved; what changed is what the page watches for.

3 Sept 2026, after the BLS revision — the labour share, measured. §7 gains two rows carrying the one variable this page has always modelled and never observed, and the reading page gains the entry behind them. BLS’s revised Productivity and Costs print put the US nonfarm-business labour share at 52.8%, the lowest in a series that starts in 1947, with real hourly compensation down 3.3% against productivity up 1.4%. Checked against the underlying index, the fall is recent: the share sat still through the two years to the end of 2024 and has dropped about two points since, roughly 1.3 points a year. The second row exists because the first invites a wrong reading: 52.8% is one sector of one country, while §4 and §6 run on a bloc-wide whole-economy figure, so the two cannot be set against each other. AMECO’s adjusted wage share is the measure that can, and it says two things worth having. The 60% intercept is high — the euro area reads ~57%, the EU ~56%, the US ~54% — and the blocs are moving in opposite directions, the euro area up since 2022 while the US falls, which makes this morning’s record low a US divergence rather than the developed world’s. Both rows are evidence beside the model in the standing sense: no slope, no gap-chart slice, no contribution to any total. The print is filed as evidence against the §6 dial’s boundary condition rather than for the bet: the dial is geared to the BIS’s premise that labour’s share does not move until substitution actually happens, and by §4’s own reading substitution has not started, so a share falling anyway says the premise is not safe, not that the bet is paying. Nothing in the release attributes any of the fall to AI, and no page constant changed. The intercept question the rows raise is left open on purpose — changing a load-bearing constant is the owner’s call, and the ledger now carries the row that states it.

3 Sept 2026, late — the floor, in units of a business that exists. §2 now states its floor as a multiple of Microsoft’s productivity franchise: Microsoft 365, Office, LinkedIn and Dynamics together turned over $135.3B in the twelve months to March 2026, so the floor at the default sliders is ≈9.6× that. The multiple is computed live and moves with the sliders, from 8.5× at the bottom of the capex range to 11.8× at the top. The comparator was chosen to answer the objection that the capture argument implies these companies do badly — it does not. Microsoft did extraordinarily well and still took four decades to build one franchise of that size; the floor asks for roughly ten, by 2030. The denominator is Microsoft’s own reported segment revenue (nine months FY26 plus Q4 FY25), confirmed against the 29 April press release and an independent write-up of the same print, and it is registered as a load-bearing constant with a checker test, so a stale figure now fails the deploy rather than sitting quietly in the prose.

3 Sept 2026, night — a correction to this morning’s entry. The St. Louis Fed entry called the measured productivity effect “small next to a bill that needs about 40% revenue growth a year”, and the note below said the same. That comparison was wrong twice over. It set a cumulative figure against an annual one without converting, and more importantly it compared quantities that do not share a denominator: an industry-level productivity gain measured in output per hour, against the sales growth five companies need. 2.9 percentage points cumulative over eleven quarters is roughly one point a year, which is not small — against the 2% threshold Waller uses it is substantial, if it is causal, which its authors do not claim. The entry now says what is actually true: the productivity finding is not too small to matter, it simply does not measure capture, which is the thing the bet turns on. No figure changed and no model above was affected; the error was in the comparison, not the number.

3 Sept 2026, evening — a nav, and §8 retired. The essay now ends at §7. The summary table that stood between the scoreboard and the footer was a second index of the reading page, kept in step by hand and read after the argument had already finished, so it is gone; the reading page is reached from the nav in the header instead, alongside the FAQ, the derivations ledger and RAAK. Section numbers are not reused — §8 is retired rather than refilled, and the ledger says so, since the revision entries below refer to it by number. The reading page itself is unchanged apart from stating what its order already was: newest first.

3 Sept 2026, later — Posen on why the displacement is not in the data. The reading page gains the Odd Lots interview with Adam Posen of 1 September. He splits the question the page keeps separate too: more indicative evidence on productivity than on jobs. Hiring still grows in long-haul trucking and junior coding, the two occupations most often named as first to go, which is where §4 already stands; but his endorsement of Luis Garicano’s messy-jobs argument goes further than the page does, holding that the “most exposed occupations” lists any wage-pool estimate leans on are the wrong shape. The interview’s other claim was traced to its source before it was written up: the 0.3% is not Posen’s and not the Peterson Institute’s but Epoch AI’s report of 24 August, and it concerns US GDP growth rather than the level. It moves the §6 growth dial, not the bill — Nvidia’s value-add is already on this page as capex seen from the supplier’s side, and the gap chart runs on the five’s own reported revenue, which no accounting convention changes. Bloomberg’s transcript sits behind bot protection, so Posen is paraphrased from a third-party automatic transcript and nothing is presented as a verbatim quote; the entry says so and cites both substantive claims to their own sources. As always for §8, every figure is the cited author’s claim, feeds no model above, and the ledger records it that way.

3 Sept 2026 — the productivity case, read against the capture rate. The reading page gains the St. Louis Fed’s Open Vault interview of 26 August with Alex Bick, on how productivity affects inflation, jobs and pay. It is the standard benign case, and it tests §3 and §4 rather than supporting them: in the Fed’s mechanism a productivity gain shows up as lower prices and higher wages, meaning the gain leaves the seller, while the bet needs a quarter of the displaced wage bill arriving as vendor revenue. The measured effect so far — 2.9 percentage points of extra cumulative labour productivity per 10 percentage points of AI adoption, US Q4 2022 to Q3 2025 — is about one point a year, and the same study found no industry-level employment effect either way, which matches the jobs row’s reading of the July payroll contraction. Both figures were confirmed against the interview and against the underlying research post the interview cites. As always for §8, every figure is the cited author’s claim, feeds no model above, and the ledger records it that way.

28 Aug 2026 — the financing of the bill, read beside the model. The reading page gains the FT's Big Read of 26 August on how the data-centre build-out is financed: debt from even the cash-rich, special-purpose vehicles that keep projects off tech balance sheets, synthetic risk transfers moving bank risk elsewhere, and insurance that cannot stretch to a $10bn site. It tests §7's strain chart (capex measured against a company's own operating cash flow cannot see spending financed off the books) and it adds to the patience clock: the same issuer's data-centre debt sold at 5.7% in April and over 7.2% in August. Oracle is where the credit market is already pushing back on one of the five, with a downgrade to triple B minus. The annotation label on §8 and the reading page also changed with this entry, from "where it presses" to "what it tests", because the old label did not say what the reader would get. A plain-language pass followed the same day, over the reading page and this section: em-dashes replaced where a comma, colon or full stop does the job, and two sentences reworded. No figure or hedge changed. As always for §8, every figure is the cited author's claim, feeds no model above, and the ledger records it that way.

27 Aug 2026, late — Oracle, the last of the five. A companion page for Oracle's 10 June print, written from the new blank-sourced template — the fifth BIS hyperscaler was the only one without one. Oracle sits outside geoveu's capture, so every figure was taken from its results release directly. Two things came out of that. The release's own trailing-four-quarter table independently confirms this page's strain series for Oracle, peak and all: capital spending ran to 205% of operating cash flow in the February quarter before easing to 174%. And the watch-list row's "shares −10%" did not survive checking — reports range from −5% to "as much as 10%" after hours, with the following morning at −12.6%, so the row now states the range. Free cash flow for the year was −$23.7B, which the row had described only as capex having "ate" it.

27 Aug 2026, evening — a reading page, and a correction. §8 now carries only each piece's headline and a one-line statement of where it presses; the full entries moved to a page of their own, which is also where new ones will land. A copy-edit pass went through the essay for sentence length — no figure, model or hedge changed, and the constants checker passed unaltered before and after. One real error found and fixed: meta.html reported the 29 July print as "shares −5% after hours" and carried alphabet.html's source list verbatim, both copied from that page when it was written. The move was −9.64% ($585.61 to $529.15) — as the watch-list row above had said all along — and the sources are now Meta's own. The ledger records the correction and how it was verified.

27 Aug 2026 — the supplier gets a page, and the scoreboard turns round. Nvidia was the only reported entity on the watch-list without a companion analysis, so it now has one. The material the row had no room for is the distance between the two financial statements: $59.7B of net income against $24.1B of operating cash flow, receivables up $22.3B in the quarter, and $42.4B spent on equity stakes over the half-year against $4.4B on its own plant. That is §5’s circuit with the supplier’s own capital inside it. The standing rule is now written down: sitting outside the BIS’s five governs the charts and the totals, never whether a print gets analysed, which is why Apple, SpaceX and Nvidia each have a page. Oracle is the one member of the five still missing one. The scoreboard table is also reversed to newest-first, with the two not-yet-reported rows kept at its foot, and the paragraph above it now says where the first wave actually landed rather than anticipating it.

27 Aug 2026 — reading beside the model. The page gained §8, a place for arguments that bear on the bet without feeding it. Two entries open it. The IMF's 25 August growth revision counts AI capital spending on the growth side of the ledger, which lengthens §7's patience clock without moving the revenue floor by a dollar. An FT Free Lunch piece argues that capital keeps flowing to knowledge-work augmentation because that is where the returns are, which narrows the wage pool §3 and §4 measure while leaving the bill unchanged. Everything quoted in the section is the cited author's claim rather than a page constant. The ledger's new §8 records them that way, and none enters the constants registry or any model above.

26 Aug 2026 — Nvidia Q2, and the supplier's view of the bill. The 26 August print beat its own guide and the street on every line — revenue $96.2B against a $91.0B guide, data-centre revenue $89.0B +117% yoy — and guided Q3 to $108.0B. The watch-list row is updated accordingly, and its slope is drawn as none: Nvidia sits outside the BIS's five, so under the standing roster policy it gets no sparkline against the required ramp and feeds no gap-chart slice or capex total. Two corrections travel with the print. The pre-print row described the $91B figure as a data-centre guide; it was Nvidia's guide for total revenue, and the row now says so. And the placeholder had promised to draw a slope after the print, which the roster policy adopted on 30 July does not allow for a non-BIS row — the promise, not the policy, was the error. The strain and gap charts are unchanged: Nvidia is the supplier, and its revenue is the five's spending seen from the other side of the invoice, so adding it anywhere in the model would double-count the same dollars. 27 Aug: the row now also carries the sequential base it was measured from — Q1 FY27 revenue $81.6B and data-centre revenue $75.2B — so the +18% q/q step reads as figures rather than only a percentage.

7 Aug 2026, afternoon — the other ledger reports. July payrolls −23,000, the first contraction in the watch-list's payroll column, with unemployment down to 4.1% — a pairing that points at a shrinking labour force rather than displacement at the bet's scale (the §4 arithmetic still says wage-drawn revenue at the gap chart's pace would mean ≈28M jobs-equivalent, unmissable in this column). The divergence test widens: earnings beating, payrolls now falling. Next reading 4 Sept.

7 Aug 2026 — SpaceX, and the third pool. The 4 August print — the first since June's listing — put the BIS's capex-to-revenue gauge somewhere none of the five have taken it: $2.35 of capital per dollar of revenue, $6.18 inside the AI segment. It also made the report's "circular financing" concrete enough to draw, so §5 now carries the circuit, with a reported figure on every arrow and the loop closing through asset prices rather than cash. §3 gained a third strip: the builders' own $0.84T of 2026 capex budgets, drawn to the same scale as the software and wage pools, because that budget is where a growing share of today's AI revenue actually comes from — the one source that cannot last. The strain chart gained an off-scale marker (≈820%, half-year) and §7 a watch-list row, with a full companion analysis — the ratio, the buyer, and why the $14.1B booked backlog is softer than the contracts' headline value. The model stays the BIS's five; SpaceX is tracked beside them, not inside them.

31 Jul 2026 — the growth reality check. §6 gained a decade of actual advanced-economy growth (IMF data, 2016–2026), split US versus the rest of the bloc against the dial's own reference lines. Outside the post-Covid rebound, neither half has touched the dial's happy zone in ten years; since 2022 the US carries a rest-of-bloc running ≈0.7–1.7%. The FAQ's growth answer now carries the current readings (US Q2: +1.5% annualized; euro area +0.4%).

31 Jul 2026 — the FAQ. A questions-and-answers page grown from real reader questions: advanced-economy definitions (China is not one), the US share of the bloc's growth (with its own sliders — the §6 dial's optimistic zones decompose into ≈4.3% sustained US growth), sources and inspiration, what is BIS-sourced versus page construction, and what would falsify the bet.

31 Jul 2026 — the strain chart. §7 gained the financed system's dashboard: each company's capital spending as a share of its operating cash flow, trailing twelve months. Two of the five now build past 100% — Amazon at ≈107%, Oracle at ≈174% after peaking above 200 — meaning the build-out there runs on the balance sheet, not the business. The revision-history note also became this expandable block.

31 Jul 2026 — audit pass. The gap chart gained a crediting toggle (ceiling vs trend-adjusted — the strict reading shows ≈$149B/yr of above-trend growth against $319B accrued); the white-collar wage pool moved onto a slider ($18–30T, default $22T ≈ 60% of labour income — the launch value $30T was the range's most generous end); and the §3 pools graphic was re-drawn to true scale after the audit found the software pool drawn ≈2.5× too small, overstating the floor's dominance.

31 Jul 2026 — the wave's last two analyses. Full pages for Apple, the bet's control group, and Amazon, where the bill rises with the proof. Amazon's capex-guide raise ($200B→$220B) folded into the gap chart's split, moving the five's 2026 total to ≈$840B.

30 Jul 2026, post-close — Amazon Q2 captured. Revenue $200.6B beat its guide, AWS +37% was the fastest in 18 quarters, and $53.4B of the quarter's net income is the Anthropic stake marking up. The July wave complete; all five of the BIS's hyperscalers on the board; the gap chart's aggregate moved to +$553B/yr against $319B accrued.

30 Jul 2026 — BIS-fidelity pass. Every model on the page audited against the BIS report's own constructions, deviations fixed (the trillion is two years of spend, not one; the capex window is the BIS's 2025–30; Box C's horizon is 2060, not 2065), every derivation documented in a public ledger the deploy now checks mechanically. Oracle's June print joined the scoreboard — the fifth of the BIS's five had been missing. Same day: the gap chart added (§7), and Microsoft and Meta Q2 actuals captured — Meta answered the conviction question this page poses of it, capex floor lifted to $130–145B while quarterly free cash flow fell to $0.8B.

26 Jul 2026. Street consensus and company guides added ahead of the earnings wave.

23 Jul 2026. Alphabet's capex raise — the first live test of this framework.

Launch assumptions stay visible as each earnings print revises them.