Chapter 1 · July 2026
A war, two central banks, and a market that climbed anyway
The macro story of 2026 starts in the Strait of Hormuz. The US–Iran war closed the world’s most important oil chokepoint, briefly spiking Brent crude toward $120 a barrel before a fragile June ceasefire brought it back near $73. That shock pushed US inflation to 4.2% in May, a three-year high, driven almost entirely by energy; core inflation, stripping out food and fuel, was a calmer 2.9%. Europe felt it too but is cooling faster — eurozone inflation eased to 2.8% in June, and Germany’s to 2.3%.
The central banks flipped
Both of the world’s major central banks reversed direction. The European Central Bank raised rates in June for the first time in three years, taking its deposit rate to 2.25%. The US Federal Reserve — now chaired by Kevin Warsh, who replaced Jerome Powell in May after the most partisan confirmation vote in the Fed’s history — held at 3.50–3.75% and signaled that its next move might be a hike, not a cut.
Stocks didn’t care
Despite war and sticky inflation, the S&P 500 sits near 7,490, up about 9% on the year, powered by roughly $690 billion in planned AI infrastructure spending from the big five tech firms. Whether that spending is rational is the market’s central argument, covered on the left. Meanwhile the euro trades near a one-year low around $1.14, and borrowing costs diverged sharply across the Atlantic: American mortgages run about 6.4%, German ones 3.3–3.7%, with German home prices rising again on chronic undersupply.
In the background
Two slow stories are worth watching. The US federal deficit is tracking near $1.9 trillion — about 5.9% of GDP — with interest payments alone now topping $1 trillion a year, a path the government’s own auditors call unsustainable. And crypto finally got its rulebook moving: the CLARITY Act, handing the CFTC authority over digital-commodity spot markets, advanced out of committee toward a Senate floor vote.
The open questions
The AI build-out: bubble or justified?
BubbleCapex is growing far faster than the revenue it should produce — a wider gap than the 2001 telecom bust — and the giants are starting to borrow to fund it.
Forbes JustifiedUnlike 1999, the spenders are the most profitable companies in history and can fund the build from cash flow, while cloud backlogs surge on real demand.
Guinness Global Investors The US deficit: slow-motion crisis or absorbable?
CrisisPeacetime deficits near 6% of GDP are unprecedented, interest already tops $1 trillion a year, and auditors call the path unsustainable.
CRFB AbsorbableThe US borrows in its own reserve currency, markets keep absorbing record issuance with the 10-year near 4.5%, and the likeliest path is slow drift, not rupture.
J.P. Morgan Asset Mgmt Chapter 2 · July 2026
A soft jobs report reshuffles the Fed’s summer
The third quarter opened with a labor-market wobble. U.S. employers added just 57,000 jobs in June, well under the 110,000–115,000 economists expected, even as the unemployment rate ticked down to 4.2% from 4.3% and average hourly earnings rose 0.3% on the month and 3.5% over the year. A cooling hiring pace alongside still-firm wages is exactly the mixed signal that makes the next policy move hard to call.
Stocks split down the middle
Markets took the weak report in stride, but unevenly. The Dow jumped nearly 600 points — about 1.1% — to a record to start July, while the Nasdaq slid roughly 0.8% as chip stocks extended a sell-off and Tesla fell; the S&P 500 finished close to flat. The rotation away from the priciest AI names, even on an up day for the broad market, is the kind of divergence worth watching.
A divided Fed
The Federal Reserve held rates steady in a unanimous June vote, but the committee is split on where they go next: roughly nine members saw one or two hikes as appropriate for the rest of 2026, while about as many favored holding or even a single cut. The soft jobs data gives the Fed room to stay patient and let the numbers decide rather than commit early.
The open question
Which way does the Fed lean now?
Toward patience or cutsHiring has clearly slowed, and a labor market adding only 57,000 jobs is hard to square with raising rates into it.
BLS release Still tilting to hikeWith inflation running above 4% and wages firm, half the committee still sees hikes ahead — one weak month may not be enough to change that.
Schwab Chapter 3 · July 2026
A record listing puts a face on the AI trade
Rather than use a rolling market recap to freeze a fast-moving return, this chapter follows one dated transaction in the AI trade. SK hynix’s final prospectus, filed July 10, records the price and size of its U.S. offering; Nasdaq records when trading began and the opening price. Those are fixed points a reader can check after the market has moved on.
The memory king comes to New York
SK hynix — a South Korean memory-chip maker whose prospectus describes high-bandwidth memory as a core product for AI systems — began trading on Nasdaq under SKHY on July 10. Its final SEC prospectus records 177.9 million American depositary shares offered at $149 each, for a $26.5071 billion public offering price before underwriting costs. Nasdaq says the shares opened at $170 and calls the sale the second-largest U.S. share sale at the time. Claims about oversubscription, a near-$1 trillion valuation and a sevenfold one-year rise are left out because the filing and exchange record cited here do not establish them.
Plumbing, and a pivotal week
Two official calendars set the next checkpoints as this chapter closed. BLS scheduled the June CPI release for July 14 at 8:30 a.m. Eastern, and the Federal Reserve calendar placed Chair Kevin Warsh’s semiannual monetary-policy testimony before the House Financial Services Committee later that day. On crypto market structure, Congress.gov showed the House-passed CLARITY Act still referred to the Senate Banking Committee; it did not establish a July floor vote. No rate-probability estimate is included because the draft did not identify a dated futures contract or calculation.
The open question
What does a record-sized AI-memory listing tell us?
What the record showsInvestors bought 177.9 million ADSs at $149, and Nasdaq recorded a $170 opening trade. That establishes strong demand at the offering; it does not measure future chip demand.
SEC prospectus What remains uncertainThe prospectus itself lists cyclical demand, competition, customer concentration and large capital needs among the risks. A large sale can fund growth without resolving those uncertainties.
SEC prospectus Chapter 4 · July 2026
Inflation falls fast, and the chip trade breaks
Both dated checkpoints the last chapter flagged landed on July 14, and they pointed opposite ways. The Bureau of Labor Statistics reported that consumer prices fell 0.4% in June — the largest monthly decline in more than six years — pulling annual inflation down to 3.5% from May’s 4.2%. Core prices, which strip out food and energy, were flat on the month, putting the twelve-month core rate at 2.6%. Economists had looked for a 0.2% decline and a 3.8% annual rate. Falling energy did most of the work, but the services side cooled too: shelter rose just 0.1% and transportation services fell 0.3%.
The chair declines to celebrate
Hours later Kevin Warsh gave his first congressional testimony as Fed chair, delivering the semiannual monetary policy report to the House Financial Services Committee. He would not read the morning’s data as a win. “There might be some that look at this morning’s data and say, ‘mission accomplished.’ That is not my view,” he said, adding that the committee has “no tolerance for persistently elevated inflation.” He announced five task forces — on communications, balance-sheet policy, economic data, productivity and jobs, and inflation frameworks — and, asked whether he works for the president, answered that the Fed is an independent central bank. On the build-out driving this market he was candid: “We don’t know the extent to which the economy will benefit from the AI buildout.” Traders trimmed the odds of a September hike to about 63%, from better than 75% the day before.
Chips fall into a bear market
Cooler inflation did not lift stocks, because the week’s real story was in semiconductors. The S&P 500 lost more than 1.5% over the week and the Nasdaq 2.9% — the S&P’s first losing week in three — while the Dow gave up nearly 1%. The Philadelphia Semiconductor Index entered a bear market, closing Friday at 11,673.89 after falling more than 4% on Thursday. Micron, SanDisk and Western Digital fell more than 5%; Applied Materials and Lam Research more than 6%. Three things were blamed: doubt about the return on AI spending, TSMC guiding to higher capital spending than expected, and — the crossover with this library’s AI book — the arrival of Moonshot’s Kimi K3, a Chinese model that scores near the top of the leaderboards at markedly lower cost. SK hynix, whose record Nasdaq listing filled the last chapter, sells into precisely this trade.
What comes next, and when
Oil turned back up as the Middle East reheated: Brent rose about 3% to above $86 a barrel and WTI 3.4% to above $80, on a sixth consecutive day of US–Iran fighting — well above the roughly $73 the first chapter recorded after June’s ceasefire. Corporate earnings, meanwhile, were strong: FactSet put blended second-quarter S&P 500 earnings growth at 24.7% as of July 17, with ten of eleven sectors growing. The next fixed point is the Federal Reserve’s July 28–29 meeting, with the rate decision at 2 p.m. Eastern on July 29; it is one of the four meetings a year that publishes no quarterly projections.
The open question
Is the chip selloff a wobble or a verdict?
A verdict on the spendingAn index entering a bear market on doubts about the return on AI capital expenditure is the market beginning to price the bubble case this book has tracked since its first chapter.
Yahoo Finance A rotation, not a reckoningBlended earnings growth for the quarter is running at 24.7% with ten of eleven sectors expanding, and buyers stepped back in before Friday’s close. One bad week in one sector is not demand disappearing.
FactSet Chapter 5 · July 2026
Record earnings, and a market that sells its biggest spenders
The checkpoint the last chapter left open was earnings season, and its first full week produced a strange picture. With about a quarter of the S&P 500 reported by July 24, roughly 86% had beaten profit estimates and 80% had beaten on revenue, and FactSet put the blended second-quarter growth rate at 37.9% — which would be the fastest since 2021. Almost none of that jump was broad strength. Strip out a single company — Alphabet, whose quarterly results carried a $98 billion gain — and the blended growth rate falls to 25.9%; FactSet called Alphabet’s surprise the single biggest reason the index figure rose. The record was largely one accounting line wide.
The market sold the spenders
The reaction told the real story. Strong results should have been good news, but investors fixed on what the largest companies are spending to produce that growth, and punished the two names spending the most on AI. Alphabet’s shares fell even with that gain on the books, as attention turned to its climbing capital-expenditure plans, and Tesla fell after its own report — dragging the communication-services and consumer-discretionary sectors down roughly 6% on the week. Tesla dropped 14.52% on Thursday, July 23, its worst single-day reaction to an earnings report on record. It was the same argument the chip index delivered in the last chapter, now aimed at the hyperscalers footing the bill.
Broadening underneath, and the Fed straight ahead
Beneath the megacap wobble the rest of the market was, if anything, widening out: the share of S&P 500 stocks trading above their 200-day average reached its highest since December 2024. And the fixed point Chapter 4 flagged now sits directly ahead — the Federal Reserve’s July 28–29 meeting, with the rate decision at 2 p.m. Eastern on July 29 and, as one of the four meetings a year without them, no fresh quarterly projections. After June’s soft jobs report and the sharp drop in June inflation, the market goes into the decision still unsure whether Chair Kevin Warsh’s next move is a hike or a hold.
The open question
Is the market turning against AI spending, or just spreading out?
A verdict on the spendersThe two companies investing most heavily in AI infrastructure were sold specifically on that spending — the capex-doubt verdict the chip index delivered last chapter, now reaching the hyperscalers themselves.
Financial Sense A healthy broadeningMost reporters beat, and money rotated into the rest of the market rather than leaving it — breadth is the widest since December 2024, which looks like leadership spreading out, not demand disappearing.
CNBC Chapter 6 · August 2026
The market pays for proof
The fixed point the last chapter left ahead arrived on July 29. The Federal Open Market Committee held the federal funds rate at 3½–3¾% — 3.50 to 3.75 percent — on a 9–3 vote. Beth Hammack of Cleveland, Neel Kashkari of Minneapolis and Lorie Logan of Dallas dissented, each preferring a quarter-point increase. The statement itself is worth reading for its brevity: under Chair Kevin Warsh the committee has stripped out forward guidance entirely, and the whole thing now runs to four short paragraphs, ending “The Committee will deliver price stability.” Warsh told reporters afterwards: “I asked for a good family fight and I got one.”
A slower economy that isn’t
The next morning the Bureau of Economic Analysis put second-quarter growth at an annual rate of 1.5%, down from 2.1% in the first quarter and below the roughly 2% forecasters expected. The headline understates the private economy. Real final sales to private domestic purchasers — consumer spending plus private fixed investment, the cleanest read on underlying demand — accelerated to 3.9% from 1.7%. What pulled the top line down was government: federal nondefense spending fell, driven largely by sales from the Strategic Petroleum Reserve, which the national accounts deduct from government consumption. Prices were the uncomfortable part. The price index for gross domestic purchases rose 5.7% in the quarter, up from 3.6%, and the PCE price index rose 5.1%; but core PCE prices, which strip out food and energy, decelerated to 3.4% from 4.4%. A quarter that looks weak on growth and hot on prices turns out, underneath, to be strong private demand plus an energy shock working through the numbers.
Five companies in one week
Then the largest technology companies reported almost on top of each other, and the market answered the question the last chapter left open. Microsoft’s results, out after the close on July 29, showed Azure revenue growing 43% in the quarter and passing $100 billion for the fiscal year; commercial remaining performance obligations, a measure of contracted future revenue, rose 84% to $678 billion; Microsoft 365 Copilot passed 30 million paid seats. Quarterly revenue was $90.0 billion, up 18%, against a $87.6 billion consensus, and the shares rose roughly 7–9%. Amazon followed on July 30 with $200.6 billion of quarterly revenue — the first time any company has reported more than $200 billion in a quarter — AWS growing about 37%, its fastest in eighteen quarters, operating income up 43% to $27.5 billion, and a raised full-year capital-expenditure plan of about $220 billion, up from about $200 billion, which the company attributed partly to higher memory costs. Apple reported record revenue of $109.4 billion, up 16%, and fell anyway, on a services miss and softer Greater China sales.
The same spending, the opposite verdict
Read against the last chapter, that is a reversal. A week earlier investors sold Alphabet and Tesla specifically for what they were spending. This week they bought Microsoft and Amazon for spending more. The distinction the market drew was not between thrift and excess but between capital expenditure with a demand number attached and capital expenditure without one: a $678 billion contracted backlog and a cloud business accelerating are answers to the question “who is paying for this?”, and a raised capex line on its own is not. The clearest evidence came from Seoul. On July 31 the KOSPI closed up 17.91%, adding 1,001.89 points to finish at 6,595.45, after SK hynix hit its daily limit at close to +30% and Samsung Electronics rose about 27%. Korea JoongAng Daily and KED Global both record it as the index’s largest single-day gain, in points and in percentage terms alike. SK hynix, whose record Nasdaq listing filled Chapter 3 of this book and whose sector fell into a bear market in Chapter 4, sells precisely the memory that Microsoft and Amazon had just told the market they cannot get enough of.
The next fixed points
Two dates are already on the calendar. The Bureau of Labor Statistics publishes the July employment situation on August 7 at 8:30 a.m. Eastern — the first read on hiring since the 57,000-job June report that opened Chapter 2. The BEA’s second estimate of second-quarter GDP, together with corporate profits, is scheduled for August 26 at 8:30 a.m. Eastern, and will revise the 1.5% figure above.
The open question
Did the AI trade just get proved, or just better dressed?
The demand is now visibleAzure growing 43% with $678 billion of contracted future revenue behind it, and AWS at its fastest in eighteen quarters, are demand figures rather than narratives — the spending is being underwritten by a backlog customers have already signed.
Microsoft Amazon It is still a promiseRemaining performance obligations are contracted future revenue, not revenue; Amazon raised its capital-expenditure plan partly because memory now costs more, not only because demand rose; and a market that swung from punishing spenders to rewarding them inside eight days is repricing sentiment at least as much as fundamentals.
CNBC CNBC (last chapter) Chapter 7 · August 2026
The jobs number arrives, and reaches back into this book
The first of the two dates the last chapter put on the calendar arrived on Friday, August 7 at 8:30 a.m. Eastern. Nonfarm payroll employment fell by 23,000 in July. The Bureau of Labor Statistics describes that as little changed, and sets it against an average monthly gain of 34,000 over the prior twelve months. The losses were concentrated and identifiable rather than general. Local government education shed 50,000 jobs after little net change over the previous year. Retail trade lost 19,000, with warehouse clubs, supercenters and other general merchandise retailers down 21,000 and gasoline stations and fuel dealers down 5,000, partly offset by 10,000 added in sporting goods, hobby, musical instrument, book and miscellaneous retail. Financial activities lost another 14,000 and is now down 121,000 since a recent peak in May 2025. Health care kept growing, at 22,000 — but against a prior twelve-month average of 36,000, which is the same direction of travel in a sector that has been carrying the number.
The revision reaches back to Chapter 2
The more consequential number was not July’s. BLS revised May down by 66,000, from 129,000 to 63,000, and June down by 37,000, from 57,000 to 20,000, leaving the two months combined 103,000 lower than previously reported. The June figure is the one this book built a chapter on. Chapter 2 opened on a soft June jobs report of 57,000 and what it meant for a Federal Reserve already divided about its next move; that report now reads as 20,000. Nothing here is deleted, so the earlier chapter stands as written and this one records what happened to its central figure. The agency’s own explanation is routine and worth keeping in view: revisions come from additional reports received from businesses and government agencies since the last published estimates, and from the recalculation of seasonal factors. A monthly payroll figure is an estimate that gets more accurate for two months after it moves markets.
Two surveys, two readings
The release carries two different surveys, and in July they pointed different ways. The establishment survey, which asks employers, produced the fall in payrolls. The household survey, which asks people, put the unemployment rate at 4.1 percent, with 6.9 million people unemployed, and BLS calls that little changed on the month and over the year alike. Underneath that, the labour force participation rate was 61.4 percent and the employment-population ratio 58.9 percent; since January participation has fallen 0.7 percentage point and the employment-population ratio 0.5. An unemployment rate can fall because more people found work or because fewer are counted as looking. The number of people on temporary layoff rose by 153,000 to 921,000. The long-term unemployed, jobless 27 weeks or more, numbered 1.8 million and made up 25.5 percent of all unemployed people. Pay barely moved: average hourly earnings for all private employees were $37.62, up two cents on the month and 3.2 percent over the year, with the average workweek unchanged at 34.3 hours.
The next fixed points
Two more dates are already set, and the first is the larger one. At 10:00 a.m. Eastern on August 28, BLS publishes the preliminary estimate of the annual benchmark revision to the establishment survey, alongside first-quarter 2026 data from the Quarterly Census of Employment and Wages. The benchmark rebases the survey against state unemployment insurance tax records that nearly all employers are required to file — a count rather than a sample. Official estimates are not updated on the preliminary figure; the final benchmark arrives with the January 2027 release in February 2027. Then on September 4 at 8:30 a.m. Eastern comes the August employment situation. A survey that just moved two months by 103,000 will publish its own estimate of how far it is off for the year, a week before the next monthly print.
The open question
Is the labour market breaking, or cooling?
What the release says about itselfBLS characterises both the payroll change and the unemployment rate as little changed in July, and as little changed over the year as well. Health care continued its upward trend, and the declines sit in two identifiable places: local government education and retail trade.
BLS BLS summary table A What the composition saysPayrolls fell in a month the prior year averaged 34,000 of gains, two earlier months were revised down by 103,000 combined, temporary layoffs rose by 153,000 to 921,000, and both the participation rate and the employment-population ratio are lower than in January.
BLS BLS summary table B Chapter 8 · August 2026
The price data go quiet, and spending turns down
Three of the fixed points on this book’s calendar arrived in a single week. On Wednesday, August 12 at 8:30 a.m. Eastern the Bureau of Labor Statistics reported that consumer prices rose 0.1 percent in July, seasonally adjusted, after falling 0.4 percent in June. The twelve-month rate eased to 3.4 percent from 3.5. Underneath, the calm was broader: the index for all items less food and energy rose 0.2 percent on the month and 2.5 percent on the year, down from 2.6, and shelter’s 0.1-percent rise accounted for roughly two-thirds of the whole monthly increase. What remains of the spring price shock sits in the twelve-month columns: energy up 14.7 percent on the year and gasoline up 24.6, even as the energy index fell 1.5 percent in the month itself. For a reader checking later, the dated fixed point: the CPI-U stood at 333.918 (1982–84 = 100) in July, unchanged on the month before seasonal adjustment.
Producer prices, published the next morning, told the same two-part story. The index for final demand was unchanged in July, after edging down 0.1 percent in June — a 0.2-percent rise in services and a 2.2-percent advance in construction offsetting a 0.7-percent fall in goods — while the twelve-month rate stood at 4.7 percent. The measure the release singles out with food, energy and trade services stripped out rose 0.4 percent on the month and 4.7 percent on the year. A quiet month, sitting on a year that was anything but.
Friday’s number
Then on Friday, August 14 the Census Bureau reported that retail and food services sales fell 0.6 percent (±0.4) in July, to $763.6 billion — still up 5.0 percent on July 2025, with the May-through-July period up 6.3 percent on a year earlier. The declines were concentrated where the dollars are biggest: motor vehicle and parts dealers fell 1.8 percent on the month and nonstore retailers — the online category — fell 2.2, while clothing stores rose 1.9. One caveat comes from the release itself: the series is adjusted for seasons and trading days “but not for price changes.” That matters in this particular year. Gasoline-station sales were up 16.2 percent on the year in dollars — in the same twelve months in which the CPI put pump prices up 24.6 percent — so a dollar figure at the pump says more about prices than about how much anyone bought.
The next fixed points
The calendar this book keeps is now crowded at the end of the month. The preliminary benchmark revision to the payroll survey — the count-against-tax-records exercise the last chapter explained — lands Friday, August 28 at 10:00 a.m. Eastern. The Kansas City Fed’s Jackson Hole symposium, the annual gathering of central bankers, policymakers and economists it hosts in Wyoming, runs August 27–29; this year’s theme is “Financial Innovation: Implications for Payments and Policy.” Then September 4 brings the August jobs report, September 11 the August CPI, and September 16 the August retail figures.
The open question
Last chapter asked whether the labour market was breaking or cooling. This week’s version: is demand following it down?
Still growing, just slowerThe BLS describes July payrolls and unemployment alike as little changed. Retail sales remain 5.0 percent higher than a year ago, and the May-through-July total runs 6.3 percent above the same months last year. Core inflation has slowed to 2.5 percent on the year.
BLS Census Bureau BLS CPI Turning down togetherPayrolls fell in July and the two prior months were revised down by a combined 103,000. In the same month, spending fell 0.6 percent, led by cars and online — and July’s 0.1-percent headline came with energy prices falling 1.5 percent in the month.
BLS Census Bureau BLS CPI