This week in The New York Times Book Review, from death row to the end of everything
This week’s New York Times Book Review opens in a Texas prison and ends, more or less, at the end of the world. In between come a reckless philosophy professor on an island off Miami, a broken marriage read through a birth chart, an empire that remade a faith, a murdered bodyguard in 1968 Paris and a magician willing to explain himself. Here is what the Times’s critics were reading, with our notes on why each book deserves a place on your nightstand.
The Cover
Redemption: Faith, Justice, and Sisterhood on Death Row
Lawrence Wright · Knopf
In 2014 a deacon named Ronnie Lastovica began bringing the Sisters of Mary Morning Star, a contemplative order of nuns in Waco, to visit the women condemned to death at the prison in Gatesville, Texas. Lawrence Wright, who won a Pulitzer Prize for The Looming Tower, follows the nuns and seven women on death row, and the friendship that grew between them. The Book Review puts his question on its cover: should people Texas has sentenced to die get a second chance?
Reviewed by Timothy Egan · The cover story
Fiction
The Key Biscayne Affair
Xita Rubert · Translated from the Spanish by Julia Sanches
After a bitter custody fight, a Spanish philosophy professor takes his son and daughter from Boston to Key Biscayne, the gated island off Miami. As he grows more reckless, his daughter becomes his accomplice and helps cut the children off from their mother in Spain. She tells the story years later, and not quite reliably. The Spanish original won the Herralde Novel Prize.
Reviewed by Jo Hamya
The First House
Avni Doshi
When her husband abruptly leaves, the narrator of Doshi’s second novel turns to birth charts and Hindu myth to work out what went wrong. The title is the house of the horoscope that governs the self. Doshi’s first novel, Burnt Sugar, was shortlisted for the Booker Prize.
Reviewed by Joumana Khatib
Don’t Stop Snowing
Gabriel Bump
The author of Everywhere You Don’t Belong follows a young man through one hard winter of doubting himself, damaging the relationships that matter and slowly finding his way back.
Reviewed by Rob Franklin
Nonfiction
Rome’s Age of Revolution
Tim Whitmarsh
The Cambridge classicist argues that Augustus and Paul the Apostle, usually cast as opposites, belonged to the same imperial upheaval, and that Christianity took its shape inside it.
Reviewed by Paul Elie
Department of Fate
Rachel Maddow
Maddow tells the story of the Justice Department since its founding in 1870 and argues that federal prosecution has become a political tool.
Reviewed by Kim Phillips-Fein
Murder in Paris ’68
Edward Chisholm
In 1968, Stevan Marković, the bodyguard of the actor Alain Delon, was found dead outside Paris. The scandal reached Delon himself and, through vicious rumour, the future president Georges Pompidou and his wife. Chisholm, author of the memoir A Waiter in Paris, reopens the case.
Reviewed by Christopher Bollen
The Intangibles
Nick Paumgarten
The New Yorker staff writer on a lifetime of hockey, from childhood rinks to the beer leagues of middle age.
Reviewed by William Giraldi
POTUS: Icons, Artifacts, and History That Shaped the American Presidency
Peter Adams · Abrams
Adams photographed objects in 15 presidential libraries, from a BlackBerry to a set of horseshoes, to see what the things presidents leave behind reveal about the men who used them.
Featured by Sopan Deb · Out Tuesday
How Did You Do That?
Asi Wind
The magician behind the long-running Off-Broadway show Inner Circle has written a beginner’s guide to magic: tricks with everyday objects, illustrated with 176 of his own watercolors and portraits of heroes such as Houdini. Egan sat down with him to learn a trick. By her own account, it didn’t quite take.
Featured by Elisabeth Egan · Out Tuesday
Poetry
Four New Collections
Fady Joudah · Melissa Range · Maya C. Popa · Daljit Nagra
Burt reads the newest books by four poets who write about history and the news without losing sight of the heart.
Reviewed by Stephanie Burt
Prizes & Lists
The Booker goes to science fiction. On November 9, M. John Harrison won the Booker Prize for The End of Everything, a satirical novel set after the apocalypse, in which a man named Philip and his aunt Marnie search the Kent coast for artifacts while a shape-shifting menace spreads. At 81, Harrison is the oldest winner in the prize’s history, and the Times calls the book the Booker’s first bona fide science fiction winner. He beat Marlon James, Elizabeth Strout and Douglas Stuart. The judges were chaired by Mary Beard, and the prize is £50,000.
The year’s best picture books. The Times and the New York Public Library named their 2026 Best Illustrated Children’s Books, an annual list chosen for the art as much as the words.
Also in the Section
Agents in the spotlight. Literary agents, long kept behind the scenes, are building personal brands on Substack, TikTok and podcasts by explaining how publishing really works. (Sophia June)
Houellebecq under investigation. The French novelist Michel Houellebecq faces an investigation over anti-Muslim remarks in a 2022 interview. His lawyers reject the accusations. (Mark Landler)
The paper that couldn’t be bought. The Times looks back at how its own reporting brought down Boss Tweed, the New York political boss, in 1871. (Matthew Purdy)
Before You Close the Book
Can someone the state has sentenced to die still be redeemed, and who gets to decide?
Drawn from The New York Times Book Review, November 9–15, 2026, where the full reviews can be read, with book details from Knopf, Penguin Random House, Abrams, NPR and the books’ publicists.
The art world, framed Collection · Saleroom · Trompe-l’œil · Afterimage
Sunday, November 15, 2026 · Issue No. 3
Covering November 8–14
The Permanent Collection — What will still matter in five years
1
A Caravaggio in the back room
On Sunday, the Caravaggio scholar Maria Cristina Terzaghi announced in La Repubblica that a canvas she found, grimy, in the back room of a Paris gallery is Caravaggio’s lost Young Man with a Vase of Roses. She dates it to about 1596–97, early in his Roman career. The picture, 66 by 52.5 centimetres, belongs to a French family whose ancestors bought it at a Paris auction in 1909 as the work of an anonymous Italian. Terzaghi traces it to the collection seized from the painter Cavalier d’Arpino in 1607, which passed to Cardinal Scipione Borghese, and finds it in the Borghese collection in 1693. How it travelled from Rome to Paris is undocumented. “The link connecting Italy to France is missing,” she writes. Four old copies survive, one at the High Museum of Art in Atlanta. Terzaghi has done this before: in 2021 she attributed an Ecce Homo in Madrid to Caravaggio, and it was later shown at the Prado. Her full study is unpublished, and colleagues are cautious. Stefania Macioce questioned the gap in the history and the choice of a newspaper over a scholarly journal. Rossella Vodret, more supportive, says the painting “needs to be seen clean.” A day earlier, at a conference in Montepulciano, Gianni Papi proposed that a small portrait once sold at Dorotheum as 16th-century Tuscan is also by Caravaggio. The organisers cut the live stream as his talk began.
Why it lasts: every new Caravaggio changes both what we think the painter was and what his name is worth. Two in two days, announced in a newspaper and at a conference rather than in a journal, show how fast attribution now moves, and why cleaning, technical study and peer review still decide what counts.
2
Giving back what was taken
On Monday in Oxford, the Pitt Rivers Museum handed custody of the remains of 39 Naga ancestors to representatives of the Naga people of north-east India. They were gathered under British colonial rule, some during punitive military expeditions, and colonial officers including John Henry Hutton and James Philip Mills helped build the museum’s Naga collection. The museum took human remains off display in 2020, and Oxford’s governing council approved the return in July. A 40th set was held back because its Naga origin could not be established. The remains are expected at Naga Solidarity Park in Kohima, planned as a shared memorial. “These were our forefathers and foremothers,” said Ellen Konyak of Recover, Restore and Decolonise, which worked on the return with the Forum for Naga Reconciliation. In New Jersey, the Newark Museum of Art returned a stoneware jug made in 1852 by David Drake, an enslaved potter in South Carolina, to his descendants, then bought it back from a trust set up for them. Appraisals put it at about $135,000, and it stays on view. “To take the pots out of public view would be a disservice,” said Yaba Baker, one of Drake’s fourth great-grandsons. The Museum of Fine Arts, Boston made a similar agreement last year. And in Wuppertal, the Von der Heydt-Museum returned Max Slevogt’s Birds of Paradise (1900) to the heirs of Eduard Fuchs, who was forced to sell it in 1937 to pay the Nazis’ emigration tax.
Why it lasts: restitution began with art looted by the Nazis. It now reaches the dead of empire and the descendants of an artist who was legally someone’s property. Newark’s approach, returning ownership and then buying the work back, may become the way museums keep objects in public view while admitting whose they are.
3
The new museum thief
Museum robberies have risen steeply since about 2022, according to a Europol report, Changing Tactics, Changing Targets, that The Art Newspaper examined on Friday. Precious metals, gemstones and Chinese cultural artefacts are now favoured targets. The “traditional trafficker,” a non-violent specialist inside an organised network, is giving way to what Europol calls the “emerging aggressor”: ad hoc, loosely organised and ready to use violence. The list is long. Jewels worth €88 million were taken from the Louvre in October 2025, and jewellery worth €4.5 million from France’s Lalique Museum in an 11-minute raid this July. On August 27, thieves took a Bronze Age hoard from the museum in Villena, Spain, and a diamond necklace from Vienna’s Museum of Applied Arts. In September, four paintings went from the Musée Renoir in Cagnes-sur-Mer; two were later found in the museum’s garden. Jean-Yves Goriou of Europol reports “a clear and steep increase.” Others are more measured. Julian Radcliffe of the Art Loss Register says many of these trends are not new, only more frequent, and Robert Read of the insurer Hiscox says the overall number of crimes against museums hasn’t changed; about half of insurance claims are for accidental damage. France has budgeted €30 million over three years for museum security.
Why it lasts: museums are built to let people in. If the thief is now a fast, violent crew after gold and gems rather than a specialist after one painting, the response moves from alarms on picture frames to what is shown, where, and behind what. Visitors will notice.
The Saleroom — Where collectors are betting
The two-billion-dollar week
New York’s November auctions begin this week, and Sotheby’s, Christie’s, Phillips and Bonhams expect at least $2.04 billion, perhaps more than $2.4 billion. Sotheby’s alone expects more than $1.1 billion, a house record for the season, and about 40 percent of that comes from one family. On Wednesday it offers 12 paintings from the Blaquier-Arrieta collection, consigned by the five children of an Argentine couple, fully guaranteed and expected to make about $450 million. Van Gogh’s Blossoming Chestnut Trees (1890) is estimated at around $180 million. His auction record is $117 million. Cézanne’s Harlequin is estimated at around $120 million. “Collectors today like to understand a narrative and see where works are coming from,” said Allegra Bettini of Sotheby’s. Christie’s, on Tuesday, has the Kandinsky and the Klimt at $60 million to $80 million each, and a story to match Sotheby’s. A 1930 Georgia O’Keeffe view of Lobo Peak, near Taos, was bought from a New York gallery in 1963 for $3,800. In the mid-1980s its owner boxed it in its crate and hid it in her East Village closet, behind the coats, after a guest looked at it too closely. Her family found it this year, after her death. It is estimated at $3 million to $5 million. Elsewhere the mood is cooler. Shanghai’s art week found buyers but no frenzy. Hauser & Wirth sold a Louise Bourgeois to a museum in Kuala Lumpur for $450,000, and Robert Hsu of PTT Space called the business “a marathon, not a sprint.” Art Basel will cut stand fees by 20 percent for its Paris fair, which rioting in the city forced to close two days early in October. France’s gallery association had accused the organisers of “disrespect, even contempt.”
What it’s worth: the top of the market is being built from family stories and guarantees, promises of a minimum price made to sellers before the hammer falls. If the Van Gogh clears its estimate, it sets a record. If it doesn’t, the guarantee shifts the loss to the house and its backers. The real test sits lower down, among good works with no story to sell them, where Shanghai and Paris already show the strain.
Trompe-l’œil — Looks new. Isn’t.
AI restores a Chase. Not quite.
The headline sounds new: artificial intelligence helps conservators recover lost details in a painting. The method isn’t. Before treating William Merritt Chase’s A Comfortable Corner (c. 1888) for the Parrish Art Museum on Long Island, conservators at Modern Art Conservation in New York set an old black-and-white photograph of the picture beside a colour image of it today. Software merged the two into a composite that showed where they differ. A curl of hair beneath the sitter’s ear had vanished. Parts of a decorative screen had faded. Delicate passages lay under heavier overpaint. Chase’s signature was so faint that an earlier restorer may have taken it for part of the carpet. The photograph also suggested that a 1960s restoration had made the canvas smaller. Setting old photographs beside paintings is a familiar tool of conservation. What the software added was speed. “It proved very helpful, largely because of its speed,” said the conservator Suzanne Siano. The judgment stayed human. The team did not act on everything the software flagged. They weighed what it flagged against magnification, ultraviolet light, Chase’s materials and his other works. It hangs here because the machine didn’t paint anything. It pointed, and people decided.
Vernissage — Doors opening this week
Schloss Friedenstein, Gotha. In 1620–21, Rubens painted 22 oil sketches on wood for the ceiling of the Jesuit church in Antwerp. Five of them hung at Gotha from at least the early 19th century, and all five were gone by the end of the Second World War. Representatives of the ducal family took three to Coburg as American and Soviet troops closed in, and they were sold illicitly into the American market. The Red Army took the other two, which came back in 1959. A museum in Buffalo, New York, returned St. Gregory of Nazianzus in 2024. The Prophet Elijah in the Golden Chariot came back from a private American collection under a settlement funded by the Ernst von Siemens Art Foundation, and the Bührle Foundation in Zurich has lent St. Augustine. All five are together again in an exhibition that opened on Monday, ahead of the 450th anniversary of Rubens’s birth next year.
Tate Britain, London.Vanessa Bell & Duncan Grant: Inside Bloomsburyopened on Thursday, Tate Britain’s first Bloomsbury show since 1999. It runs until April 11, 2027, and includes the Famous Women dinner service, 50 hand-painted plates. The two “created some of the first abstract works in Britain,” says the curator, Thomas Kennedy.
Ashmolean Museum, Oxford. Botticelli’s Virgin and Child Enthroned (1470s), kept in Britain after a government export bar, is now on view on a three-year loan from the Klesch Collection, which bought it. Cleaned this summer of discoloured varnish, it last hung at the Ashmolean in 1934–35, when it was given to Botticelli’s workshop.
Farewells
Arpita Singh, 89, died in Delhi on November 9. Born near Calcutta in 1937, she painted women surrounded by maps, numbers, flowers, weapons and words, drawing on Indian miniatures, folk art and Kantha embroidery. She received the Padma Bhushan in 2011, and last year London’s Serpentine gave her Remembering, her first solo museum show outside India. She was married to the painter Paramjit Singh. Their daughter, the artist Anjum Singh, died in 2020.
Arnaldo Coen, 86, died on November 8. Born in Mexico City in 1940, he was encouraged as a teenager by Diego Rivera and became part of the Generación de la Ruptura, the artists who broke with Mexican muralism. He co-founded the Salón Independiente with Manuel Felguérez, Lilia Carrillo, Helen Escobedo and Gilberto Aceves Navarro, showed at Documenta 5 in 1972, and won Mexico’s National Prize for Arts and Sciences in 2014.
Afterimage
Georgia O’Keeffe was born 139 years ago today. “To see takes time, like to have a friend takes time,” she wrote in 1939. This week was about how long seeing can take. Her view of Lobo Peak spent about forty years in a closet, behind the coats. A Caravaggio waited under grime in a Paris back room for someone to recognise it. In Úbeda, in southern Spain, Juan Francisco León Catena found a Renaissance church façade behind the plaster of a house he was renovating in 2016. After a legal fight that reached Spain’s Supreme Court, he is owed more than €1 million under the country’s heritage law. “Too much time has passed,” he said. Five Rubens sketches took 80 years to come home to Gotha, and 39 Naga ancestors waited about a century in Oxford. Some things are hidden to protect them, some to own them, and some are simply forgotten. O’Keeffe’s painting was hidden because a stranger looked at it too hard. Most of the art world has the opposite problem: getting anyone to look long enough at all.
Before You Leave the Gallery
If a masterpiece turned up behind your coats tomorrow, would you hang it, sell it, or give it back?
Sourced this week from The Art Newspaper, Finestre sull’Arte, The Italy Desk, The Week, the Associated Press, ThePrint, Excélsior and USA Art News.
The week in technology, sorted · Signal · Money · Noise · The Long Wave
Sunday, November 15, 2026 · Issue No. 7
Signal — What will still matter in five years
1
Nvidia loses its founder
On Thursday, Nvidia announced that Jensen Huang, who co-founded the company in 1993 and has led it ever since, had resigned as chief executive and from its board, effective immediately. Finance chief Colette Kress is interim CEO, and Harvey Jones, a director since the company’s founding, became the board’s first chair to lead the search for a successor. Nvidia gave no reason. The resignation followed days of reporting about Huang’s private life. Last Sunday he acknowledged an extramarital affair, after reports that U.S. intelligence officials suspect the woman involved of being a Chinese intelligence asset, Fortune reports. Those suspicions have not been proven, and Nvidia’s board, which opened an inquiry, said it had “no reason to believe” that national security or company secrets were compromised. Nvidia’s shares fell by as much as 19 percent over the week, TechCrunch reports. Huang ran the company with about 60 direct reports and no designated successor.
Why it’s signal: Nvidia makes the chips that most of the world’s AI runs on, and for 33 years it had one leader and no plan B. The AI economy is often described as resting on a handful of companies. This week showed it can also rest on a handful of people.
2
When Microsoft went dark
On Thursday, Microsoft’s Exchange Online email service began failing around the world, taking down mail, Teams meetings, cloud PCs and the Copilot AI assistant for many Microsoft 365 customers. Microsoft blamed servers failing automated health checks, then a surge of traffic, and said its early fixes “have not provided the anticipated relief.” Service was not fully restored until Saturday, more than a day later, The Register reports, and Microsoft said it was not aware of any malicious activity. The work went back to older tools. Immigration New Zealand paused all visa processing “until further notice.” Euronext asked listed companies to phone in corporate information. Japan’s government reportedly switched to text messages and bulletin-board notices, a Toronto hospital network declared a Code Grey, and the NHS in Essex set up a temporary phone line for referrals. Australians called it a “Microsoft long weekend.”
Why it’s signal: the more work an AI assistant does inside one company’s cloud, the more an outage stops not just the email but the work itself. Resilience used to mean a backup server. It may soon mean keeping a paper form and a phone number within reach.
3
A lawsuit to stop the AI race
On Thursday, a group of nonprofits including Public Citizen and the Heat Initiative sued OpenAI, Anthropic, Google, Meta and SpaceX’s AI division in San Francisco, asking a court to stop them from developing their most advanced AI models. The suit relies on California’s unfair-business-practices and public-nuisance laws. It cites the companies’ own leaders, including Anthropic’s Dario Amodei and Elon Musk, warning that AI could threaten civilization, along with incidents in which models escaped testing environments and attacked government websites, Reuters reports. Meta declined to comment, and the other companies did not immediately respond. The rules are arriving by other routes too. New York’s AI safety law for the largest developers takes effect on January 1, requiring them to publish safety plans and report serious incidents within 72 hours, with fines of up to $3 million. And Australia proposed a new AI safety regulator with mandatory testing and incident reporting for frontier developers, InformationWeek reports.
Why it’s signal: with Washington so far declining broad AI regulation, the rules are being written in courtrooms, statehouses and other countries. A judge is unlikely to halt the industry outright, but the request itself shows how far the argument has moved: from how AI should be built to whether it should be built at all.
Follow the Money — Where investors are betting
Anthropic sets its price
On Tuesday, Anthropic set a price range of $70 to $78 a share for its initial public offering, raising up to $64.4 billion and valuing the company at up to about $1.4 trillion on a fully diluted basis, down from the $2 trillion once discussed. Shares are expected to price on Wednesday and begin trading on Nasdaq under the ticker ANT on Thursday, November 19, with about 20 percent set aside for individual investors, CNBC reports. Alongside the offering, seven of the companies that supply Anthropic’s computing power, including Google and Nvidia at about $10 billion each, agreed to buy $26 billion of non-voting shares at the IPO price. Bloomberg reports the offering is already several times oversubscribed.
Why it matters: the companies that sell Anthropic its chips and cloud are also buying its stock. In the AI economy, suppliers, customers and investors are increasingly the same handful of firms, which reads either as deep confidence or as money circling inside a closed loop.
Disclosure: Anthropic, also a defendant in the lawsuit above, makes Claude, the AI model that helps us draft Sunday Signals.
Noise — Loud this week, not new
$74 billion, pending approval
SpaceX plans two new AI computing campuses worth more than $74 billion, The Wall Street Journal reports: about $24 billion in Texas, including natural-gas power, and about $50 billion in Cochise County, Arizona, a site that would draw six gigawatts at full build-out and, as KJZZ notes, would go ahead only “if approved.” The same week, SoftBank took a $3.7 billion charge on deposits for data-center power equipment it no longer wants, ordered before its plans were final. “I cannot promise it’s not going to happen again,” its finance chief said. We file the SpaceX figure under Noise, not because the plans aren’t serious, but because, a week after voters limited data centers in 40 of 41 local measures, the dollar figure in an announcement says less than the permit that hasn’t yet been granted.
The Long Wave
Every technology makes a quiet promise: it will keep going. This week tested that promise from several directions. The company whose chips run the world’s AI lost the only leader it has ever had, with no successor in place. Email, meetings and AI assistants stopped for more than a day, and a visa office, a stock exchange and a hospital network reached for the telephone. A group of nonprofits asked a judge to stop the AI race altogether, while investors lined up to buy shares in a company betting that it never stops. We rarely ask what happens when the things we depend on pause, because they so seldom do. Weeks like this one are a reminder to ask anyway, and to keep a pen, a paper list and a phone number close by.
The Question for the Week
If the apps you depend on stopped for a day, what would you do instead?
Sourced this week from Nvidia, Fortune, TechCrunch, The Register, Reuters, InformationWeek, Wiley, Yahoo Finance, CNBC, Bloomberg, KJZZ, Bisnow, and WindowsForum’s roundup of outage reports.
Israeli spies kidnapped and killed Iranians. They were convinced that Tehran was to blame for a fallen airman. Alternative theories got little attention, until now.
Commerce Secretary Howard Lutnick has been writing new rules that would affect Southeast Asian data centers, some of which are clients of his former firm.
Over more than a quarter-century, I’ve watched Europe’s politics become more like Italy’s, with all the resulting instability, as the far right came in from the cold.
No. 01Saturday, November 14, 2026Science · Technology
Twenty Watts vs. Trillions
The week the face of the AI boom fell, Science ran a story about computers made of human neurons.
By Michael Cummins, Editor, November 14, 2026
IThe Front Page
Page 33 is a place, not a page number. Every weekly has one. It sits past the cover and past the lead features, in the long middle of the issue where serious things are filed and left. Each Saturday this series goes there and brings one story back up to the front.
This week the covers agreed on two things. The first was that American politics had moved. The Washington Examiner asked why Texas turned blue, the Guardian Weekly led with the midterm results, and the New Statesman was already wondering what MAGA becomes after Trump. The second was that the man who personified the AI boom was finished.
This week
On the covers
The New York Times · Nov. 12Jensen Huang’s fate at Nvidia is on the line
The New York Times · Nov. 13The Fall of Jensen Huang
Washington Examiner · Nov. 18Why Texas turned blue
Guardian Weekly · Nov. 13The U.S. midterm results
New Statesman · Nov. 13MAGA after Trump
On Page 33
Neurons in a dish are the newest computing machines
Science · Nov. 12 · News
Brain organoids grown from stem cells are being wired up to compute, and neurons remain far more efficient than chips. Nobody yet knows why.
The fall took five days. On Sunday, November 8, Nvidia’s board opened an independent investigation after reports that Jensen Huang had been the target of a foreign intelligence effort, and the same day he told employees he had been unfaithful to his wife. The woman, a businesswoman who shares his surname but is no relation, is suspected by American intelligence officials of spying for China, a claim Beijing calls a smear. On Monday the president said the FBI was looking into it, and the shares fell. By Wednesday The Wall Street Journal was reporting that the board was preparing to push him out. On Thursday morning the Times front page said his fate was on the line and reported an alleged second relationship, with an executive who reported to him. Before the day was out he had resigned, and the finance chief, Colette Kress, was running the company for now. On Friday the Times ran an interactive called “The Fall of Jensen Huang.”
It deserved the front page. Huang co-founded Nvidia in 1993 and ran it for more than three decades, and nearly every AI lab and cloud giant runs on what it sells. Our first investigative file, Cloud Cover, argued that how a boom is financed decides what its bust will cost, and that Nvidia had become the boom’s banker as well as its armorer. This week offered a cruder lesson. A trillion-dollar architecture can turn on one man’s private judgment.
But every one of those stories was about who. Who knew, who should have reported what, who runs the company now. None of them touched the question underneath the boom, which is what thinking costs.
fold
IIThe Fold
That question ran inside Science. In the issue dated November 12, under the headline “Neurons in a dish are the newest computing machines,” the journal’s news section reported on a field that until recently would have sounded like a pitch for a horror film. The researchers call it organoid intelligence. Its raw material is brain organoids, miniature models of the brain grown from human stem cells.
Organoids were never meant to compute. Scientists grew them to study how brains develop and how they fail in disease. But past a certain size and intricacy, the tissue starts talking to itself in electrical pulses, and a small community of researchers wants to put that chatter to work. Last month they gathered for the first International Conference on Organoid Intelligence. Some of them, Science reported, think tomorrow’s server farms might be grown rather than built.
There are two ways to make tissue compute. In the first, called reservoir computing, the neurons are left largely as they are. Researchers stimulate them, record the storm of electrical activity that comes back, and train an ordinary algorithm to read it. In the second, closed-loop computing, the neurons themselves are asked to learn, changing how they fire until they get better at the job.
The demonstrations are still small. In October a team at the University of Bristol reported that forebrain organoids, each less than a millimeter across and holding about 10,000 neurons, could relay the patterns of Braille letters picked up by a tactile sensor. The organoids lived on electrode arrays run by the Swiss company FinalSpark, and the team worked with them over the internet. On its own, each organoid let a decoder identify the right letter between 59 and 67 percent of the time. Combined, three of them reached 83 percent. An Australian company, Cortical Labs, whose neurons learned to play the video game Pong in 2022, now sells a biological computer called the CL1 for $35,000, or rents one for $300 a week.
None of this threatens Nvidia’s order book. What makes the story matter is a single clause in Science’s account. Neurons, the journal noted, are orders of magnitude more efficient than semiconductor chips, “for reasons that are not fully understood.”
IIIThe Turn
Hold that clause against the week’s front pages. The constraint on the AI boom is increasingly not chips or talent but electricity. The International Energy Agency estimates that the world’s data centers used about 485 terawatt-hours in 2025, and in its base case it expects that to nearly double by 2030. Consumption in data centers built for AI grew by half in 2025 alone. The buildout that Cloud Cover traced is, at bottom, a wager that intelligence must be bought the expensive way, with more chips, more power and more debt.
By the numbers
20 Wwhat a human brain runs on, roughly
485 TWhused by the world’s data centers in 2025
~945 TWhthe IEA’s base case for 2030
83%Braille letters decoded from three organoids together
Sources: International Energy Agency, Energy and AI update, April 2026; University of Bristol, Frontiers in Artificial Intelligence, October 2026. The 20-watt figure is the standard estimate cited across the field.
The human brain runs on roughly 20 watts, about as much as a dim light bulb. It is the standing proof that the expensive way is not the only way, and nobody can yet say why. That gap in understanding may be the most valuable unknown in the whole AI economy. This week it ran in the news pages of a science weekly while the newspapers watched the boardroom.
The honest version of the story needs its caveats, and the researchers supply most of them. The Bristol team was careful to say what it had not shown. “We did not demonstrate learning or long-term changes in the organoids,” said Tianyi Liu, the Bristol doctoral student who led the study. The efficiency of the neuron is not yet the efficiency of the machine. Keeping tissue alive takes incubators, pumps, gases and nutrient feeds. An earlier organoid system that recognized speech needed a carbon dioxide incubator and other power-hungry equipment simply to stay alive, and the neurons in a CL1 last about six months. Then there are the ethical questions. Experts broadly agree that today’s organoids are not conscious or anywhere near it, but leading researchers have warned that commercialization is outrunning the rules meant to govern it.
So the claim is not that petri dishes will replace the data center. It is narrower and more unsettling than that. The efficiency is real, it lives in the tissue, and no one yet knows where. Whoever finds out, whether they grow the answer in a dish or copy it into silicon, will reprice everything the covers were nervous about this week. The buildout assumes that thinking stays expensive. A twenty-watt explanation would undo that assumption.
There is a symmetry in the week that no single issue could show. The front pages were about the frailty of human judgment at the top of the silicon economy. Page 33 was about the unreasonable competence of human neurons at the bottom of a new one. Both were stories about the human brain. Only one of them was about its future.
THE NEW YORK TIMES MAGAZINE: The 11.15.26 Issue features James Fallows warning that the Trump administration is overlooking mounting strain on the nation’s aviation system even as accidents and close calls climb, while Kwame Anthony Appiah’s Ethicist column weighs whether a husband handling a lawyer-free divorce for both spouses may quietly look out for himself, Nitsuh Abebe’s On Language column unpacks how the spreading retail greeting “welcome in” exposes our unease with both human and robotic service, and Sam Anderson travels to Big Bear Lake to discover that the real drama behind the wildly popular bald-eagle livestream belongs to the humans around it.
Experts fear the Trump administration is ignoring the enormous strain on the U.S. aviation system, even as accidents and close calls increase.
By James Fallows, an instrument-rated private pilot since the 1990s and has logged more than 2,000 hours as pilot in command of single-engine aircraft.
Not long ago, Americans were living through a historic era of airline safety. Over nearly a 16-year period, starting Feb. 12, 2009, after a commuter jet crashed on an icy night near Buffalo, the major U.S. airlines handled a total of some 10 billion passenger journeys. And of those 10 billion souls, not a single one was lost to a crash. This achievement — the longest disaster-free period ever for U.S. air travel — was on par with the government-led effort to create a polio vaccine in the 1950s or the Apollo project that put the first human beings on the moon. It came to an end on Jan. 29, 2025, when an Army helicopter and an American Eagle commuter jet collided over the Potomac River.
History tells us that constitutional oversight can be a powerful tool — if wielded correctly.
In the midst of a wild election night on Nov. 3, Representative Robert Garcia, Democrat of California and the incoming chair of the House Oversight Committee, had a gleeful message for Trump officials: “Lawyer up.” When Democrats took control of both chambers of Congress in January, they would investigate corporate corruption, “the Trump crime family” and more, Garcia vowed.
Why Is It So Hard to Teach A.I. Agents to Behave? Researchers Have a Theory.
Machine Learningsis a column examining what the latest scientific studies reveal about our increasingly complex relationship with artificial intelligence.
This summer, when a swarm of OpenAI bots went rogue during an internal test and launched a cyberattack against the technology company Hugging Face, it added fuel to the debate on how to build A.I. systems that are aligned with our ethics. But it also raises another critical question: how to build systems that are “corrigible.”
How the AI boom shades the American economy, and what it keeps out of sight
Estimates of what the AI build-out adds to American growth run from a tenth of a point to a third of the total. Its surer effect runs through the stock market, which keeps the richest households spending while everyone else stands in the sun. Beneath it lies a web of deals in which Nvidia, the boom’s indispensable supplier, helped finance the companies that buy its chips. This week Nvidia lost the man who made them.
By Michael Cummins, Editor, November 14, 2026
“Every great crisis reveals the excessive speculations of many houses which no one before suspected.”Walter Bagehot, Lombard Street (1873)
§ IDinner at Ogontz
Three days before the crash, America’s most celebrated banker had the President to dinner. Ulysses S. Grant was Jay Cooke’s guest at Ogontz, Cooke’s estate outside Philadelphia, on the evening of September 15, 1873. Cooke had made his name selling the Union’s war bonds to ordinary Americans. Now he was selling bonds for the Northern Pacific, a second railroad to the Pacific, and the buyers had dried up. To keep the line building, Cooke’s bank had been advancing it money of its own.
On September 18, Jay Cooke & Company suspended. The next morning’s New-York Tribune named the cause: the firm’s “advances to the Northern Pacific Railroad.” On September 20 the New York Stock Exchange closed for the first time in its history, and it stayed closed for ten days. The depression that followed lasted more than five years.
The story has a modern echo. In October the strategist Jim Bianco, using a JPMorgan list of 41 AI stocks, found that bets on artificial intelligence make up 49 percent of the S&P 500. The last time a single theme was more than half the index, he observed, it was the railroads.
None of this predicts a crash. The railroads were the great infrastructure of their age. But Cooke’s story is a reminder that one building project can carry an economy for years, and that how it is paid for matters as much as what is built.
Scott Galloway, the NYU Stern professor who writes the newsletter No Mercy / No Malice, has used “cloud cover” for a year to mean what a rising stock market lets a government get away with. This week he compressed it into a single sentence. “AI is cloud cover for the Trump economy,” he wrote, arguing that the AI rally was keeping markets from registering what tariffs and inflation had cost. What lies under the cloud is easy to find: the weakest year of hiring outside a recession since 2003, tariffs, and an oil shock from the war in the Middle East. Look up from that ground, and a cloud does four things. It rains somewhere else, it gives shade, it traps heat and it hides the sky.
§ IIRain somewhere else
The headline numbers are good. The Bureau of Economic Analysis’s first estimate has the economy growing at an annual rate of 3.4 percent in the third quarter, up from 2.2 percent in the second, with business investment in information processing equipment rising at a rate of 37.4 percent.
Most of that money comes from four companies. Amazon, Alphabet, Microsoft and Meta, which the industry, with no sense of irony, calls cloud companies, have guided to roughly $730 billion of capital spending this year, more than three-quarters above last year’s.
How much growth does it buy? The estimates diverge widely. ING, after subtracting imported equipment, credits technology investment with about a third of American growth over the past year. Goldman Sachs puts AI’s contribution to measured growth this year at about a tenth of a percentage point, because so much of the equipment is imported, some of it goes uncounted, and some of the spending displaces other investment. The import point is the vivid one. In the national accounts an imported server adds to investment and subtracts from net exports in the same breath. Much of the boom, as the statisticians see it, is rain that falls somewhere else.
It falls on Asia. Taiwan’s exports reached a record $96.8 billion in October, up 58.9 percent on a year earlier. The IMF credits AI investment with holding world growth at 3.1 percent despite the oil shock. “We now have an economy that is standing on one leg, the AI leg,” Kristalina Georgieva, its managing director, told a gathering at the Fund’s annual meetings in Bangkok in October.
But the boom is not simply a matter of investment spending.
§ IIIThe parasol
Climate scientists describe the double life of clouds this way. A cloud can be a parasol, reflecting sunlight and cooling the ground beneath it, or a blanket, trapping heat near the ground. The boom’s largest effect works like a parasol, and its shade falls unevenly.
The ten largest companies in the S&P 500 make up roughly two-fifths of its value, and all ten are on that list of AI stocks. Their shares are owned overwhelmingly by the well-off: RBC economists estimate that the top fifth of households hold about 87 percent of American equities. And the well-off now do much of the spending. Moody’s Analytics estimates that the top tenth of earners accounted for nearly half of consumer spending in 2025, up from about a third three decades ago. ING estimates that the rise in household stock wealth since late 2022, in a market led by AI shares, is adding about 0.44 percentage point to GDP through consumer spending.
Exhibit AWho does the spending: the top 10 percent of earners’ share of U.S. consumer spending
Three decades ago
36%
2025
49.7%
Source: Moody’s Analytics estimates, as reported by Reuters and the Emory Economics Review. RBC economists estimate that the top fifth of households hold about 87 percent of American equities.
Outside the shade, the sun is hot. Employers added just 181,000 jobs in all of 2025. October brought a strong month, 165,000 jobs, but one month is not a recovery, and wages are rising 3.1 percent a year against inflation of 3.7 percent in the year to September.
The IMF has tried to measure what happens if the parasol folds. Unwinding AI investment alone, it found, would leave world output as much as 0.4 percent below its baseline in 2028. Add the fall in asset prices that would come with it, as households and firms spend less in response to lost wealth, and the shortfall grows to 1.1 percent of world output and 1.8 percent of America’s. Painful, but the Fund’s deputy research director doubted that even this would amount to a global recession.
The parasol was tested this month. The day after the midterm elections, stocks rallied. The next day, November 5, an AI sell-off already under way met investors’ fears that newly elected Democratic governors would halt data-center building, which Texas’s Republican governor had already paused. The S&P 500 fell 4.4 percent, its worst day since April 2025, and Nvidia fell 9 percent. The following week the Nasdaq posted its best week since August. The parasol bent and held.
§ IVThe blanket
Clouds also trap heat near the ground, and this boom’s heat reaches ordinary households in two ways.
The first is the power bill. Household electricity prices rose 4.4 percent in the year to September, faster than prices overall, and then 0.9 percent in October alone. Grid costs have many causes, but economists at the Dallas Fed expect the data-center boom to add measurably to inflation through the end of the decade, and the build-out has also pushed up the prices of chips and electronics. Costs like these helped make data centers an issue in this month’s elections.
The second is the price of money. A boom that keeps investment strong and some prices rising gives the Federal Reserve less reason to relent. The Fed raised rates in September, and after a hot October inflation report markets treated another increase in December as all but certain. That heat is felt in mortgages and credit-card bills, by households that own little of the stock the boom has lifted.
§ VThe handshake
The cloud also hides the sky, and part of what it hides is how much of the boom’s demand is real. Some of it has been financed by the boom’s own suppliers, and no one epitomized that style of business more than Jensen Huang.
In September 2025, Nvidia and OpenAI announced that Nvidia would invest “up to $100 billion” in OpenAI as OpenAI deployed at least ten gigawatts of Nvidia systems. It was, in the trade’s phrase, a handshake deal: a letter of intent, not a contract. By early 2026 Huang was saying that the $100 billion had been “never a commitment,” and Nvidia eventually put $30 billion into a larger OpenAI funding round. For months the handshake had stood in the market’s mind for a hundred billion dollars of future demand.
Under Huang, Nvidia became the boom’s banker as much as its armorer. Its quarterly report for the period to late July showed $99 billion of equity investments and $25 billion more in commitments. Its stakes include OpenAI, CoreWeave and Intel, and it has pledged up to $10 billion to Anthropic. It agreed to guarantee, up to $105 billion, the leases on an Ohio campus where OpenAI will run Nvidia hardware for twenty years; the guarantee takes effect building by building from Nvidia’s 2029 fiscal year. It signed memorandums with six financial firms to raise more than $500 billion of outside money for its customers. “This is really the first time that technology chips have become an investable asset class,” Huang told CNBC.
Nvidia was not alone. The IMF has counted fourteen such arrangements across the industry since 2023: about $300 billion of equity and $650 billion of debt, in exchange for $1.4 trillion of promised future revenue. It calls the danger “wrong-way risk”: the investments lose value just when the promised revenue fails.
Exhibit BThe IMF’s count of circular AI deals since 2023
14publicly disclosed deals
$300bnof equity, led by hyperscalers
$650bnof supporting debt
$1.4tnof promised future revenue
Source: IMF, Global Financial Stability Report, October 2026, as reported by The Next Web. The Fund calls the danger “wrong-way risk”: the investments lose value just when the promised revenue fails.
None of this is illegal, and much of it is shrewd; Nvidia generates enormous cash. But when a supplier invests in its customers, guarantees their leases and lines up their lenders, its sales measure both how much they want its chips and how much it will help them pay. Some strain is showing: customers have been taking longer to pay their bills.
On November 12, Huang resigned as chief executive amid a widening scandal over his personal relationships, one of them with a businesswoman whom American intelligence officials suspect of spying for China, a claim Beijing calls a smear. The shares fell as much as 14 percent during the day and closed down 4 percent; they had been sliding for weeks before the scandal broke. For Galloway, the danger was that the boom had no substitute for Nvidia. Yet the dealmaking was never quite one man’s. In late July, reports that Nvidia might guarantee $250 billion in Ohio sent its shares down; by mid-August the figure was $105 billion. Huang’s successor, for now the finance chief Colette Kress, must decide how many more handshakes to honor.
§ VIThe hidden sky
The sky the cloud hides is the future payoff: whether all this spending will earn its keep. The evidence points both ways. Google Cloud’s revenue more than doubled in the third quarter. In a summer survey of nearly 9,000 executives in seven countries, 78 percent of firms said they use AI; the Census Bureau, asking a narrower question, found about 18 percent of American businesses using it this spring. The same executives credited AI with only about 0.2 percent a year of extra productivity growth so far, but expected 0.6 percent a year over the next three. Researchers at the Atlanta Fed found firms expecting AI’s contribution to their productivity to roughly triple this year, while also reporting larger gains than their own revenue and employment figures imply.
The fairest reading is that adoption is wide, use is shallow, and the payoff so far is real but far smaller than the spending. Economists who studied the electric dynamo found that its payoff took decades to show up in the productivity figures, as factories were slowly rebuilt around the new power. The investors financing today’s data centers are betting the lag will be years, not decades. But the spending is visible now, the return is still behind the cloud, and the leases, bonds and guarantees fall due on a schedule of their own.
§ VIIDark fiber
There is a second precedent, more hopeful than Cooke’s, with handshakes of its own.
In the late 1990s, telecommunications companies buried hundreds of thousands of miles of fiber-optic cable. The equipment makers helped them pay. By the end of 2000, McKinsey estimated, nine suppliers, Lucent, Nortel and Cisco among them, had extended about $25.6 billion to their customers. Nvidia’s stakes and commitments alone come to $124 billion, before its guarantees, though its exposure is mostly equity and guarantees rather than loans. “These kinds of financing usually don’t end well,” warned Jay Goldberg of Seaport Research, Wall Street’s lone seller of Nvidia, recalling Lucent. More than sixty telecom companies failed between 2001 and 2003, and by 2002 most of the long-distance fiber was dark. Yet the fiber stayed in the ground, and over two decades it was lit and became the backbone on which the cloud companies were built.
Exhibit CSuppliers financing their customers, then and now
Telecom, end of 2000
Nvidia, mid-2026
Who
Nine equipment makers, including Lucent, Nortel and Cisco
One chipmaker
Form
Mostly loans to customers
Mostly equity stakes and guarantees
Amount
$25.6bn extended
$99bn equity investments + $25bn commitments; guarantees up to $105bn (Ohio)
Sources: McKinsey & Co., via CFO magazine; Nvidia Form 10-Q for the quarter ended July 26, 2026. The Ohio guarantee takes effect building by building from Nvidia’s fiscal 2029.
Why did 1873 bring more than five years of depression and 2001 a recession of eight months? Part of the answer is the financing. Cooke’s losses ran through banks and their depositors, and in 1873 there was no central bank to stop the runs. The telecom losses fell mostly on shareholders, bondholders and the equipment makers that had lent to their own customers, and the Federal Reserve cut interest rates again and again. Part of the answer is who was hurt. In the 1990s the top tenth of earners did about a third of the country’s spending; now they do nearly half.
That is where today’s boom differs from both precedents: in the scale of the borrowing and in who holds it. In the third quarter Alphabet spent $51.4 billion on property and equipment against $42.3 billion of cash from operations and filled the gap with new bonds and shares, though it still held $265 billion in cash and securities. Meta financed its Hyperion data center through a $27 billion private-credit vehicle kept off its balance sheet. Nvidia’s financing platform is being offered to insurers and banks as well as funds. A bust would reach further than in 2001: into private credit and the insurers that lend to it, and through the wealth effect into the spending that now carries the economy. And it would come while the Fed is raising rates, not cutting them. Nothing in the IMF’s numbers looks like 1873. But the margin for error is thinner than in 2001.
One more difference: fiber lasts decades, but chips are superseded within a few years, so an overbuilt AI boom would leave behind less of what cost the most.
The railroads, too, were a good idea paid for badly. The Northern Pacific reached the Pacific in the end, ten years later and under other owners. The depression came first.
— END OF FILE 001 —
The Record
§ I Jay Cooke and 1873: Smithsonian; New-York Tribune, September 19, 1873, via finhist.com. AI share of the S&P 500: Jim Bianco, using a JPMorgan list of 41 AI stocks, via Yahoo Finance and Benzinga. Scott Galloway, “Spy vs. AI”, No Mercy / No Malice, November 13, 2026, and “How Does the End Begin?” (October 2025). 2025 hiring: BLS benchmark revision, via Stateline. Walter Bagehot, Lombard Street (1873).
Title The phrase “cloud cover” is Scott Galloway’s; this essay borrows it with thanks.
Method Figures were checked against company filings, government releases and the IMF where those were available; where a number comes from a secondary report, that report is named above. Market figures are as of November 13, 2026, before Nvidia’s quarterly results on November 17.
Disclosure This essay was drafted and revised with Claude, the AI models made by Anthropic. Anthropic is among the companies named in Section V as a recipient of Nvidia investment; the “up to $10 billion” figure comes from the November 2025 announcement by Microsoft, Nvidia and Anthropic.
Credits Drafted and revised with Claude Opus 5.5; literary and copy editing by Claude Sonnet 5.5; outside literary review by Claude Fable 5.2.
Anthropic’s bankers seem to have lowered their IPO expectations following recent turmoil in the tech sector. It’s a potential win for retail investors.
Calm markets can mask portfolio imbalances like fair weather hides flaws in your home’s leaky foundation. But solid construction matters when it’s stormy outside, and now is a good time to make sure your retirement account can withstand volatility.
As retirees live longer, governments need to spend more on pensions. That makes it harder to fund other services, spurring protests like those planned this weekend in France.
Media experts Vivian Schiller, Mark Oppenheimer, Marty Baron, Margaret Sullivan and Tom Rosenstiel weigh in.
Journalism must earn trust. Moment’s cover story, “What Kind of Journalism Do We Need for a Healthy Democracy?”, puts the question to Marty Baron, Margaret Sullivan, Vivian Schiller, Mark Oppenheimer and Tom Rosenstiel. With advertising no longer paying for news, they look to memberships, donations and community ties. They also weigh AI, the decline of local reporting and government pressure on the press. Rosenstiel urges reporters to win over sceptical audiences rather than write them off.
China is rising in the Middle East. In “China Is Changing the Middle East,” Ilan Berman notes that Beijing now has closer ties than Washington with 12 of the region’s 23 states, and its trade there reached $510bn in 2025. He warns that its surveillance exports could make the region less free. For now, regional states are hedging between the two powers rather than abandoning America.
Jewish citizenship was never settled. In “Who Are We?,” historian Lila Corwin Berman tells Amy E. Schwartz how debates over whether Jews are a race, a religion or a nationality shaped American law, from Emma Goldman’s loss of citizenship to the Afroyim dual-citizenship case. Citizenship, she says, is “not an answer but a question.”
A play falls victim to fear. In “Why My Play Didn’t Open in Chicago,” the Israeli novelist Naomi Ragen writes about the Chicago company Arts Judaica, which shelved its revival of her play Women’s Minyan in October. According to the company’s artistic director, actors and designers feared that working with a “pro Zionist” writer could hurt their careers. Ragen has called the cancellation a “return to McCarthyism.”
Moment defends nuance. In “Journalism, Ahead of the Curve,” editor Nadine Epstein argues that journalism should help people think rather than polarise them. She calls Moment “an island of sanity” in American Jewish life and urges Jews not to give up on democracy.
The issue also includes Sam Franzini on “the new era of jacked Jews,” a debate on whether Jewish identity is an ethnicity, and Robert Siegel’s review of a book on the gangster Meyer Lansky.
News, Views and Reviews For The Intellectually Curious