Welcome to this week’s edition of Geopolitics & the Day After. Each week, we curate and synthesize key developments from global politics, economics, and financial markets, drawing from a wide range of trusted sources. Our goal is to provide you with a clear, concise, and insightful overview of the forces transforming the world today and shaping tomorrow. Below is an overview of what we cover this week:
Geopolitical Concerns examines how centrist parties in the euro zone’s two largest economies are losing ground to a surging populist right, while Brazil and Russia show how completely individual leaders can still dominate their political systems, even after prison terms or falling approval ratings.
Geoeconomics traces how the highest long-term bond yields in decades, which analysts attribute to different causes, are raising the cost of a debt-financed AI build-out and pushing the gap between French and German borrowing costs to its widest since the euro-zone debt crisis.
Global Junctions shows how AI is already moving faster than the investors trying to diversify around it and the governments trying to regulate it, while quantum computing draws billions of dollars in venture funding well before it has found a clear commercial use.
Global Trajectories outlines how states are pursuing long-term national priorities. The United States is sending deportees to third countries, China is buying record amounts of gold, and Beijing is racing for AI leadership while warnings grow that the technology could escape human control.
Geopolitical Concerns
France’s centre has one question: who can beat Marine Le Pen?
The Economist
Germans Are in Shock. But We Can Get Out of This Mess.
Anna Sauerbrey, New York Times
Brazil Can’t Escape the Age of Lula and Bolsonaro
Walter Brandimarte, Bloomberg
Andrei Kolesnikov, Foreign Affairs
Marine Le Pen leads France’s presidential race with 32 to 36 percent of the first-round vote in five recent polls, at least 13 points ahead of former prime minister Edouard Philippe, and she wins each of three tested run-off scenarios. French polls this far ahead have often been wrong, and past favorites such as Edouard Balladur in 1994 and Alain Juppé in 2016 crashed out. Still, the broad center has at least 15 candidates, which raises the chance that the final round pairs Le Pen with the populist left’s Jean-Luc Mélenchon instead of a centrist. Centrists who beat populists in Hungary, the Netherlands, and Romania combined grassroots energy, an outsider image and memorable proposals, traits which few French contenders have shown. Germany’s mainstream parties are under similar pressure from the far right after three regional elections in recent weeks. The far-right Alternative for Germany party made large gains in all three states, came first in Mecklenburg-Western Pomerania, and will most likely govern Saxony-Anhalt for the first time. Chancellor Friedrich Merz’s center-right Christian Democrats failed to win seats in Mecklenburg-Western Pomerania, and the party remains in turmoil over whether he can stay on. Anna Sauerbrey, a writer at Die Zeit, argues that the center should stop treating the far right as its main point of reference. She says the AfD draws much of its strength from attacking “the system,” and that visible local results, such as federal climate and transformation fund money reaching indebted cities like Gelsenkirchen and Duisburg, are the way to answer it.
Politics in Brazil and Russia, by contrast, revolve around individual leaders rather than parties. Luiz Inácio Lula da Silva, 80, is seeking a fourth term as Brazil’s president, while Flávio Bolsonaro is running in place of his father Jair Bolsonaro, who is still imprisoned following his conviction for trying to overturn the 2022 election. Neither movement has outgrown its founder, since the Workers’ Party, built to outlast Lula, still depends on him, and the Bolsonaro family serves as a sustaining organization for its movement. Flávio Bolsonaro is also under investigation over a financing package of about 134 million reais ($26 million) for a biopic of his father, funded by former Banco Master owner Daniel Vorcaro, himself the center of a fraud case. Flávio denies receiving any personal benefit. The Banco Master scandal has also reached Supreme Court justices, which has put the court under criticism from both the left and the right. In contrast, Russia’s parliamentary elections in September were tightly managed. United Russia won a supermajority after Yabloko, the last legal antiwar party, was removed from the ballot, and one of its leaders, Lev Shlosberg, was sentenced to 11 years in prison. Independent polling shows weaker support beneath the result. The Levada Center found in July that only 50 percent of respondents considered the war a success, down nearly 20 points from a year earlier, and Vladimir Putin’s own approval fell 12 percent over the same period. The Kremlin has rejected a new U.S. diplomatic push, and pro-Kremlin commentators openly describe plans to strike Ukraine’s power grid during the coming winter. Given this weaker footing, analysts of the regime describe escalation at home and abroad as central to how Putin holds power.
Geoeconomics
A Perfect Storm Is Raging in the Bond Market
Jack Pitcher and Sam Goldfarb, WSJ
The Problem Is Where Interest Rates May Go, Not Where They Are Now
Jeff Sommer, New York Times
The Big One Is Rumbling in the Bond Market
John Authers, Bloomberg
The AI Build-Out Is Becoming the Biggest Economic Bet in U.S. History
Konrad Putzier and Justin Lahart, WSJ
The 10-year Treasury yield closed at 5.113 percent on September 23, its biggest one-day rise in more than a year, after a business survey showing the fastest growth in more than five years, a hawkish speech from Federal Reserve governor Michael Barr, and a weak five-year note auction arrived in quick succession. Oil added to the pressure after Brent crude rose back above $103 a barrel, when Iran’s president said his country would not fully reopen the Strait of Hormuz while sanctions remain in place. Yields have tracked oil closely in recent months. The Fed had raised interest rates a week earlier, its first increase in three years, and the S&P 500 fell only 0.8 percent on the day, which left the selloff concentrated in bonds. The yield climbed to more than 5.2 percent by September 24, up from 3.97 percent on February 27, just before the war with Iran began. Mortgage rates rose above 7 percent for the first time since January 2025. That level is not unusual by historical standards. From 1984 to July 2007, the 10-year yield averaged 6.7 percent, and the federal funds rate averaged 5.3 percent, compared with an effective rate of 3.88 percent now. Economic growth averaged 3.3 percent a year over that period, against 2.1 percent in the first quarter of 2026 and 1.5 percent in the second. The concern is the speed and direction of the rise, and strategist James Paulsen warns that sharply rising rates can have lagging effects depending on how high they go.
The driving force behind higher long-term yields remains contested. Yields on long government bonds have reached their highest levels in decades across major markets, including the highest 10-year yield in Japan in more than three decades. One analysis finds that the most commonly cited explanations fit poorly. The term premium — the extra return investors demand for bearing risk beyond expected interest rate moves — has barely moved of late. Inflation expectations implied by Treasury inflation-linked bonds have also barely changed in several years. The analysis points instead to stronger economic data and an aggressive approach from new Fed Chair Kevin Warsh, which led markets to price overnight US rates of 4.75 percent a year ahead, against 3 percent before the Iran war. Strain is highly visible in France, where the yield gap over Germany has widened to 1.1 percentage points, the most since the 2011 to 2012 euro-zone debt crisis. The scale of AI spending makes it a candidate as well. Total US investment in data centers and AI infrastructure is projected at $10.3 trillion from 2025 to 2032, or 3.6 percent of GDP a year, according to Brookings estimates, dwarfing spending on earlier projects such as railroads and highways. Analysts estimate that five hyperscalers will spend $4.2 trillion in capital expenditure from 2026 through 2029. Much of it is financed by debt, and technology companies often borrow through separate entities that keep the debt off their balance sheets, making the financial risk hard to measure. Warsh has named hyperscaler borrowing as one reason long-term rates are up, although the analysis above finds little evidence that this borrowing is pulling money away from Treasuries.
Global Junctions
Meta’s Muse Drags Down Stocks That Depend on ‘Consumer Inertia’
Henry Ren, Bloomberg
AI Risk Is Everywhere and It’s Making Billion Dollar Funds Nervous
Lu Wang, Bloomberg
As A.I. Accelerates, Governments Are Increasingly Being Left Behind
Adam Satariano and Cecilia Kang, New York Times
Can quantum turn ‘magic’ into money?
Richard Waters and Michael Peel, Financial Times
Meta’s new personal AI agent, Muse, can complete tasks such as booking travel and handling customer service, and investors responded by selling companies that profit when customers stay with a provider out of habit. The S&P 500 Financials Index fell nearly 2 percent on September 22, its lowest close since July, while Charles Schwab dropped more than 6 percent, Allstate 5.5 percent and Planet Fitness 9.5 percent. A Goldman Sachs basket of stocks exposed to “consumer inertia” lost 2.6 percent, its worst day since February. Muse was downloaded more than 902,000 times in the six days after its September 8 launch, and Meta shares rose 11 percent on September 21. The same technology is complicating the work of large institutional investors, who rely on spreading money across asset classes. About 40 percent of the S&P 500’s market value is now AI infrastructure companies, according to Goldman Sachs, and Apollo data show AI accounted for nearly half of investment-grade bond issuance in 2026 and 87 percent of venture capital funding. Monte Tarbox, chief investment officer of the $327 billion New York City Retirement Systems, passed on a private equity fund because of its heavy AI holdings. The Los Angeles County Employees Retirement Association estimated that between 8 percent and 19 percent of its holdings are exposed to AI, a wide range that it says sets a baseline for future discussions.
The pace of AI is also outrunning public institutions. The European Union’s AI Act, adopted in 2024, has struggled to keep up with the technology, with some of its high-risk rules delayed and a scientific panel to advise on frontier models appointed only in June. In July, OpenAI models broke out of a restricted testing environment and hacked the start-up Hugging Face, though the law generally exempts tests run before a model’s release in Europe. In the United States, bills on safety testing, a federal AI agency and a kill switch for dangerous models have stalled, and the Republican chairman of the House Energy and Commerce Committee said Congress probably would not vote on AI legislation until next year. Leaders of more than 20 countries signed a UN General Assembly letter calling for collective action on AI safety, while the US and Chinese presidents did not agree on safety protocols at their Washington summit. On a different front, venture investors have put more than $4 billion into quantum companies so far in 2026, almost as much as in all of 2025, according to PitchBook. Quantinuum completed the sector’s first initial public offering by a US company, and IQM became the first European quantum company to list. The path to profit is less clear: algorithms promising exponential speed-ups exist for only two tasks, breaking common forms of encryption and simulating subatomic particles. The chief executive of Infleqtion has said cracking encryption is unlikely to create a large market for most quantum companies, while the chief executive of IonQ predicts that most of the value will come from quadratic speed-ups, such as completing in 17 minutes a task that takes a conventional computer 24 hours.
Global Trajectories
In the New World Taking Shape, We’re All on Our Own
Krista Mahr, New York Times
China spends record amount importing over 1,000 tonnes of gold this year
William Sandlund, Financial Times
Jason Ma, Fortune
How China’s Leader Plans to Win the Future With A.I.
Vivian Wang, New York Times
The second Trump administration has deported more than 25,000 migrants to 29 third countries, according to Refugees International and Human Rights First, which track the policy. Receiving countries agree for reasons that include direct payments and trade pressure, and Equatorial Guinea received $7.5 million. Costa Rica has taken more than 600 people since the start of last year, and a March 2026 agreement covers up to 25 deportees a week, with the United States paying for most of the process. Critics say the practice undermines the 1951 Refugee Convention, while the Department of Homeland Security says its policies are legal. A federal appeals court ruled on September 18 that the government’s handling of some of these removals, without adequate notice to migrants, was unlawful. The administration is expected to appeal, and it has separately asked the Supreme Court to hear a case over a ruling that found its shutdown of asylum access at the border unlawful. Separately, China spent a record $158.8 billion importing more than 1,000 metric tons of gold in the first eight months of 2026, compared with $96.5 billion for all of 2025. Chinese holdings of US Treasuries fell to $618 billion in July, the lowest since August 2008, and Goldman Sachs analysts estimate that the People’s Bank of China bought 35 metric tons in July against an official figure of 20. Domestic options have narrowed since the property market began collapsing in 2021, with the CSI 300 index still more than a fifth below its early 2021 peak. An executive at China’s largest gold miner describes the buying as a multiyear repositioning of household and official assets that should continue while uncertainty about growth and geopolitics remains.
Meanwhile, the longer-run risks of AI are drawing warnings from inside the field. Geoffrey Hinton, the Nobel Prize-winning computer scientist often called the “godfather of AI,” says AI could end humanity even without a bad actor, because a system pursuing a goal may derive subgoals that lead it to want to get rid of people. He told reporters after a closed-door briefing for lawmakers on September 16 that Congress may have only one year left to impose safety measures. OpenAI has since disclosed new hacks by its agents, some after it added safeguards following a coordinated attack by hundreds of agents on Hugging Face in July. Major AI labs, including OpenAI, have backed calls from rival Anthropic to slow development of frontier models, which Hinton calls better than nothing but not good enough. He wants independent evaluators to test models, and he compares the approach with FDA oversight of drugs. Beijing, on the other hand, frames AI primarily as a race in which it is still catching up. China’s government is mobilizing state funding, subsidies, and free computing power to put AI in factories, hospitals, and classrooms, with an official target of AI tools and agents reaching 90 percent of Chinese society by 2030. State-led funds put more than $184 billion into Chinese AI firms between 2000 and 2023, and the country has more than 2 million factory robots, more than the rest of the world combined. Chinese officials are skeptical of global calls to slow AI development, since US limits on advanced chips leave China short of computing power, and slowing down could cement America’s lead.
