U.S. Import Prices Rose in September
Friday we got the third element of the inflation flywheel after the CPI and the PPI.
US import prices increased 1.1% MoM in September
following a 0.5% rise in August. On a YoY basis, imports rose 8.1%.
Prices for nonfuel imports increased 0.7% MoM in September and have not
declined on a monthly basis since August 2025.
Nonfuel import prices are up 6.0% YoY.
- capital goods (+4.2% YoY but +5.8% a.r. in the last 4 month),
- industrial supplies and materials (+12.8% YoY),
- consumer goods,
ex-automotives (+4.0% YoY but +1.1% MoM in September), - foods, feeds, and beverages (+9.0% and +14.2% a.r. in the last 3 months),
- automotive vehicles, parts, and engines (+2.6% YoY but +6.5% a.r. in the last 3 months).
Import prices exclude duties, such as tariffs imposed on imports by the Trump administration, as well as transportation costs.
(Ed Yardeni)
In all, core CPI rose 0.3% in each of August and September, +3.7% a.r.. Core PPI final demand rose 0.43% on average since July, +5.2% a.r.. Core Import prices rose 0.65% on average in August and September, +7.9% a.r.. The pipeline is full and accelerating.
Core PCE inflation (Oct. 29) is now forecast at between +0.29% and +0.36%. It was +0.25% in August after +0.13% in each of June and July and +0.27% on average in Q2. Core PCE monthly inflation needs to be sustained at +0.17% for the Fed to achieve its 2.0% objective.
The Atlanta Fed’s GDPNow is at +3.6% for Real GDP in Q3, while the NY Fed model sees +2.8% for Q4. Strong demand vs restricted supply. Productivity, where are thou?
Several Fed officials, including the influential John Williams, assess that there is no urgency to raise rates. Market odds for October are at 17%.
Let’s see what Trump does after the midterms. Lower oil prices or a forever war?
Investors dump data centre bonds as AI borrowers face ‘Darwin moment’
(…) Investors are also dumping debt issued by so-called neocloud companies — of which the best known is Nvidia-backed CoreWeave — which have been the most aggressive in how they finance their AI build-out. (…)
“Data centre bonds are having their Darwin moment, where the weakest credits struggle to survive,” said Ludovic Subran, chief investment officer at Germany’s Allianz. “The surge in credit spreads following power and construction delays show that data centre debt is priced too tight for execution risk.” (…)
Neocloud providers have been hit even harder. Bonds issued by CoreWeave have sunk below 80 cents on the dollar — prices typically considered distressed — pushing yields close to 15 per cent on Thursday. The company’s shares are down 17 per cent this week, while rival Nebius has lost 18 per cent. (…)
The rout has come as a report this week by tech consultancy Gartner predicted that 80 per cent of neoclouds would “cease independent operations” by 2030, because of a multitrillion-dollar gulf between the pace of capital spending and expected returns from AI.
Gartner said AI data centre owners would need to generate revenues of $8.5tn by the end of 2030 to cover their capital and financing costs and generate healthy returns on those investments. But it forecast that consumer and corporate spending on AI services that would flow to data centre providers would reach “no more than $3.6tn”.
“A continued AI infrastructure investment trajectory could leave more than $4tn in asset value exposed to writedown risk through 2032,” Gartner said. (…)
I wrote last Thursday:
Gartner’s analysis sees a need for $8.5T in revenue through 2030 to achieve a 12% ROIC but it only expects $3.6T in AI revenues.
The Columbia analysis says that $3.7T revenues by 2032 would be enough to generate a 10% return.
But how can we reliably forecast usage and revenues at this time? We have yet to see how AI agents penetrate our lives. The first chart (from JP Morgan) plots token volume through August but the second chart shows how agent demand has been exploding after June.
Unlike the internet bubble, infrastructure spending is following demand and can’t supply enough.
If Anthropic’s Q3 profitability is sustainable, Jevons Paradox could well take care of the current revenue and financing angst.
But it would be nice if bond yields could stop rising…
All this is the context that this is an existential race
- for frontier model developers like Anthropic and OpenAi and
- for the US vs China
China may be somewhat behind on frontier models and may be hindered by various export restrictions but
- it has the clear backing of the government
- it has abundant cheap power
- it faces little if any social backlash
- it is built on open source models and apps
- it does not face the same financing challenges
US AI is clearly too sensitive and critical to fail.
Ed Yardeni is right: “If some AI-related credits do blow up, the Godfathers of the AI mafia (along with their protectors in the government) will clean up the mess.”
Citigroup and Bank of America have become the latest US banks to restrict AI agents from accessing client accounts, as lenders raise concerns with regulators that agents could drain deposits by moving idle cash in search of higher returns.
Citi said it does not permit “third-party AI agents or assistants to log into a customer’s account or transact directly within our digital banking platforms”, adding that this was designed to help safeguard customer accounts and information.
Bank of America’s Brian Moynihan also suggested the bank was restricting AI agents’ access to consumer deposits while it comes up with protocols to determine which entity was ultimately liable for payments authorised by artificial intelligence. (…)
AI companies are rushing to launch new tools and chatbots designed to help consumers manage their personal finances. (…)
JPMorgan Chase chief executive Jamie Dimon said this week that the bank would restrict AI agents from accessing client’s accounts “because you don’t want rogue agents and there could be thousands of them”.
Client interest or banks’ interest?
A report from Apollo’s chief economist, Torsten Slok, amplified the concerns, with the economist warning that AI assistants could automatically sweep household cash into accounts paying 3.3 to 5 per cent instead of the 0.1 per cent national average on checking accounts.
JPMorgan, Bank of America and Wells Fargo hold $1.6tn of non-interest-bearing deposits — 16 per cent of their liabilities. Paying a 3 per cent rate on those would cost them $47bn a year, almost half of their collective annual earnings.
If banks are forced to pay higher interest rates on retail deposits to keep them from seeking a higher rate elsewhere, it could put a big dent in their profits. (…)
Wall Street executives have also contacted regulators at the Federal Reserve to express concerns about the sharp rise in consumers using AI agents to carry out financial transactions on their behalf, according to a person familiar with the discussions.
Meanwhile in China (via Gemini 3.0):
- In August 2026, the Payment and Clearing Association of China, guided by the People’s Bank of China (PBOC), introduced the nation’s first AI agent-driven payment rules.
- The rules mandate a strict “Know Your Agent” (KYA) identity mechanism to verify autonomous software before it transacts.
- Transactions are subjected to tiered risk management boundaries based on the agent’s model capabilities.
- Fund clearing must remain with licensed entities, making banks legally liable for any rogue AI transactions.
- In October 2026, Ping An Bank became the first listed Chinese lender to formally pass comprehensive AI management board rules.
- Tech giants like Alibaba and Tencent are building highly scrutinized ecosystems (such as Alipay’s Zhixiabao) to let AI agents use financial mini-programs safely inside their walled gardens.
- These measures follow severe security warnings, including an October 2026 discovery by Tencent of rogue third-party AI agents attempting to bypass bot protections in Chinese cloud sandboxes.
When it comes to AI agents doing your shopping, a sharp geographic divide has emerged: the US relies on fragmented corporate gatekeeping and platform-by-platform blocks, whereas China uses a top-down, unified framework integrated into its giant super-apps.
While American retail giants are locking their digital doors to keep autonomous software out, Chinese platforms are building a regulated toll-road to let verified agents spend money.
In the United States, there is no federal framework governing agentic commerce. Instead, individual retail and e-commerce platforms are using updated terms of service (TOS) to unilaterally block or permit AI software.
Retail giants are locking down their interfaces. Amazon blocked Meta’s transaction-focused Muse agent, citing concerns over unauthorized account data capture. Research shows Amazon also aggressively blocks agents from OpenAI, Google, Anthropic, and Perplexity. Similarly, eBay barred shopping agents from taking automated actions on its marketplace without express written permission.
While Amazon blocks everything, Target permits select approved AI providers, and Walmart actively partners with tech firms to build co-developed agentic shopping pipelines. This forces AI developers to negotiate access retailer by retailer.
China has bypassed the corporate litigation phase by funneling agentic shopping into pre-existing, state-monitored “super-apps” and establishing clear financial guardrails.
- In China, AI agents do not scrape external web pages to shop; they operate natively inside apps like WeChat and Alibaba’s updated Qwen model. Qwen connects directly to Taobao (retail), Fliggy (travel), and Alipay (payments). Because identity, product data, and payments sit on one single proprietary rail, there are no web-scraping or bot-blocking battles.
- Guided by the People’s Bank of China (PBOC), the nation utilizes a strict “Know Your Agent” (KYA) framework. Every software agent attempting to execute a transaction must be formally identified, risk-ranked (high, medium, low), and bound to clear spending limits assigned by the human user.
- To handle the inevitable legal issue of unauthorized purchases, Chinese guidelines force payment firms to maintain a tamper-evident evidence chain. This logs the user’s original settings, the AI model’s internal prompt decision, and the final payment node to quickly resolve any disputes regarding rogue agent transactions.
Anthropic offers Blackstone double-digit yield in latest chip financing High borrowing costs complicate AI lab’s goal of achieving investment-grade status
The $60bn debt package, dubbed “Project North Star”, would help fund Anthropic’s procurement of about 1.5 gigawatts of custom semiconductors developed by Google, known as tensor processing units. (…)
Blackstone funds would invest $9bn in the debt backed by Anthropic’s five-year lease payments on the chips, accounting for about half of the $18bn junior loans, according to people familiar with the matter.
The private capital giant was offered a coupon of 10 per cent and an even higher total return after accounting for new-issue discounts and other fees, the people said. A borrower with a similar credit profile typically pays a yield of about 7 per cent in the public bond market, according to the ICE BofA BB US High Yield Index. The rest of the $9bn would be sold to a broader group of investors in the coming months, potentially with a lower yield, the people added.
Just four months ago Anthropic sold about $4.4bn of debt in a similar deal known as “Project Big Sky” at an interest rate of 8.5 per cent, the FT reported at the time. (…)
EARNINGS WATCH
We have 45 reports in: the beat rate is 84% with a surprise factor of +4.1%.
The 45 companies having reported (21 Financials) showed earnings up 51.4% on revenues up 14.1%.
S&P 500 earnings are now expected to rise 30.7%, ex-Energy 26.2%.
Very much K-shape sectorial growth however:
- 5 sectors AI or war-related: +58.1% on average on average revenue growth of +19.0%
- 6 sectors “normal economy”: +4.1% on average on average revenue growth of +6.8%.
Note the decline in margins for ““normal economy” sectors. Costs are biting now. Producer prices are rising 2.6pp faster than consumer prices (core +1.3pp, right chart). Either consumer prices play catch up or margins decline.
Citadel Securities observes that
- MU and NVDA alone account for 36% of expected S&P 500 earnings growth
- The ten largest contributors represent roughly two-thirds of expected growth
- Excluding those ten companies, expected earnings growth falls to ~11%
- The smallest 300 S&P 500 companies by market cap are expected to grow earnings ~5%
Being small is tough nowadays.
And when the biggest are so big like now, most everything else is small …
I like this NDR chart courtesy of Steve Blumenthal: the deviation from the trend is a better way than Warren Buffett’s to relate equities with GDP:
Last week I showed this other NDR chart plotting demand vs supply of equities. The lines have now crossed …
Citadel Securities also observes that
Another remarkable feature of this market has been the absence of broad-based selling. We are approaching a full year without a single 80% NYSE downside-volume day, defined as a session when at least 80% of NYSE total volume is concentrated in declining stocks.
Since 1996, the average calendar year has experienced 21 such sessions, and no completed year has recorded fewer than five. The last occurrence was November 17, 2025, roughly 225 trading sessions ago, already the longest stretch in nearly 30 years of data. The previous record was just 129 sessions.
Despite macro uncertainty, geopolitical headlines, and some extreme moves across individual sectors and themes, investors have continued to rotate between winners and losers rather than indiscriminately reduce equity exposure.
Maybe, the corporate sector will push demand back up after the earnings season … US corporates have already announced a record $1.3 trillion of share repurchase authorizations YTD, 8% ahead of last year’s record pace. Technology alone accounts for $573 billion, or 43% of the total.
… potentially helped by seasonality:
The Most Polite Trucker in America Is a Driverless 18-Wheeler They’re on the interstate in Texas and can do almost everything a human truck driver does, except honk
(…) Self-driving cars are on the cusp of moving from novelty to commonplace. Waymo driverless taxis buzz around cities like San Francisco and Phoenix. Next up are the biggest, loudest, most-daunting-to-other-drivers vehicles on the highway: the trucks. (…)
This stretch of Texas, flat, straight and filled with trucks, is the perfect testing spot for dozens of driverless big rigs, some without a human driver and others with one riding along just in case. About half a dozen companies, including Aurora, develop the technology for driverless trucks and team up with truck makers to produce them. Aurora, based in Pittsburgh, expects to have 200 autonomous trucks on highways in Texas and nearby states by year-end, up from around 30 big rigs today.
Volvo Autonomous Solutions said it plans to roll out 300 autonomous trucks with Aurora and Waabi, another autonomous-trucking company, by the end of 2027 that would run in Texas and Oklahoma. Aurora estimates that by 2030, 30,000 autonomous trucks could be on the roads. (…)
Around two dozen states, mostly in the southern half of the U.S., now allow autonomous trucks on the road without a human operator under certain conditions. (…)
“Its ability to see long distances is better than a human driver,” said Matt Detmar, chief executive of Detmar Logistics, which currently uses two driverless Aurora trucks to transport sand in a 60-mile loop in Midland, Texas. The trucks use lidar to see more clearly in the dark, dust and foggy weather and can react faster or avoid a hazard. (…)
Today, a typical driverless truck run is best suited for what shippers call the middle-mile, or the stretch from depot to depot. It looks like this: Humans load a trailer and drive it to a depot where they hand it off to a driverless tractor. Without a human, the truck drives from depot to another depot or yard where the trailer is either unloaded or a human driver takes over for docking or driving the final mile. Driverless 18-wheelers can’t back up without a human driver, though the companies say they’re working on adding that feature. (…)
It cost an average of $2.34 per mile to operate a truck in 2025, and about 44% of that represents driver wages and benefits, according to the American Transportation Research Institute. Fuel made up 21% of the cost in 2025. Autonomous trucks can reduce fuel consumption by 10% because they can calibrate speeds and optimize acceleration, braking and coasting with more precision, the companies say. Goldman Sachs forecasts that autonomous trucking will become cost-competitive with human-operated trucks in 2028 in the U.S. (…)
Navigating the massive star-shaped junction around the Arc de Triomphe in Paris is among the most hair raising experiences in European driving.
A phalanx of cars, vans, motorcycles and trucks maneuvers across lanes without formal markings, funneling haphazardly into 12 avenues that spread off the Place Charles de Gaulle. Adding to the anarchy is a unique rule that road users driving into the swirling traffic have right of way.
The result is a stop-start jumble of vehicles nosing into gaps with motorcyclists hunting for tiny spaces to squeeze through. A nightmare for autonomous driving systems, you might think.
But ensconced in a Jaecoo 7 SUV guided by Chinese startup Momenta Global Ltd.’s automated driving technology, I could let the car work it out for itself. It threaded smoothly through the chaos, negotiated its way around the roundabout and picked an exit without once asking for human intervention. (…)
Taking on the congested roads and impatient drivers in the French capital, I got a glimpse of how advanced driver-assistance systems can make urban traffic safer and less exhausting. It also laid bare the challenge of fine-tuning systems for different locations. (…)
Trump Waives Most Russia Sanctions Passed by Congress The president signed an executive order to exempt Russia from most parts of a law he signed last month
On Thursday, the North Atlantic Treaty Organization said drones had struck two ships in Romanian territorial waters of the Black Sea overnight. A third vessel was hit just outside Romania’s territorial waters, leaving seven crew members missing.
European authorities have accused Russia of targeting the ships, and four other ships in nearby waters off Bulgaria days earlier. Romanian President Nicușor Dan and Bulgarian Prime Minister Rumen Radev on Thursday met with NATO Secretary-General Mark Rutte about the situation, but announced no responses. (…)
NATO’s military commanders “are doing exactly what they need to be doing at this time,” said U.S. Army Col. Martin O’Donnell, spokesman for the alliance’s top military leadership. “They are consulting.”
None of the recent incidents has reached the threshold of what NATO members consider to be acts of war, which creates a dilemma: How can Western countries compel Russia to stop provocative, illegal actions—without going to war?
Diplomats increasingly fear what is known as a boiling-frog situation, where Russia gradually gets bolder and NATO’s public grows inured, normalizing or ignoring incessant aggression.
“Obviously there are some attempts by Russia to…destabilize society, to reduce support for Ukraine,” Romania’s Dan said Thursday. “Our answer is we will continue to support Ukraine.” (…)
Attacks in the Black Sea follow months of airborne incursions into NATO territory by Russian warplanes, drones and missiles. Europeans have also accused Moscow of staging or provoking dozens of acts of sabotage and cyberespionage across the continent, showing a growing appetite for risk over recent months. German Chancellor Friedrich Merz accused Russia of attempting to attack planes at a Leipzig airport with explosive-laden drones in August and Dutch intelligence agencies accused Russia of hacking internet-connected video cameras to spy on European support for Ukraine. (…)
But as attacks linked to Russia become deadlier and occur inside NATO borders, ignoring them becomes impossible. Diplomats and military officials fear that Moscow’s provocations increase the risk of its war in Ukraine triggering a wider military clash with NATO, even if through accidental escalations that quickly spiral. (…)
Some analysts also worry that unchecked Russian boldness could tempt it into uncharted areas of aggression. On Thursday, a Russian satellite, Cosmos 2612, was on a trajectory to approach a U.S.-registered satellite that delivers sensitive information to support Ukraine’s war efforts, according to commercial satellite data. This has raised the prospect of Moscow targeting commercial satellites and taking its sabotage campaign against the West for the first time into space, said researchers at the Center for Strategic and International Studies, a think tank in Washington.
No worries, Trump is taking care of business.
So is Putin.
Are we sleepwalking into World War III?
Gideon Rachman is the FT’s chief foreign affairs commentator.
(…) “I can’t believe we’re having this conversation in the 21st century.” (…)
Mark Rutte, Nato’s secretary-general, warned European leaders in 2025: “We are Russia’s next target . . . and we must be prepared for the scale of war our grandparents or great-grandparents endured.”
The fear that a global conflict may be brewing extends well beyond Europe. Xi Jinping consistently warns of the dangers of the US and China falling into a “Thucydides Trap”.
The Chinese leader is picking up on a theory advanced by Graham Allison, a Harvard academic, that throughout history the dominant global power has tended to end up at war with the rising power. (…)
But could the Russia-Ukraine war turn into a truly global conflict? In the eyes of some of the participants, it already has. When North Korean troops arrived on the frontline to support the Russians in their fight against Ukraine, Zelenskyy called the development “the first step to a world war”.
The war between the US and Iran has also overlapped with the Ukrainian conflict. The Iranians are widely believed to have received assistance in targeting US bases in the Middle East from Russia. Ukraine, meanwhile, has offered its expertise in counter-drone warfare to the Gulf states.
Western security officials are increasingly worried about the close ties between an “axis of adversaries” centred around Russia, China, Iran and North Korea. (…)
The four countries of the axis are united by their discontent with the US-led world security order. Russia resents Nato expansion and wants to see the alliance pushed back. Iran has attacked US military bases across the Middle East. North Korea threatens South Korea, which relies on the US for protection. China claims Taiwan and most of the South China Sea as its territory — and seems intent on displacing the US as the dominant power in the Pacific.
Mark Rutte of Nato has frequently discussed the idea that China and Russia might launch co-ordinated military offensives in Asia and Europe at the same time. (…)
The Iran and Ukraine wars dominate the global headlines. But the International Crisis Group recently counted 65 ongoing conflicts around the world — the highest number since the end of the second world war. The conflict in Sudan, fuelled by outside powers, is particularly bloody — punctuated by massacres of civilians and the displacement of more than 14mn people. (…)
Russia’s all-out invasion of Ukraine did not emerge from nowhere; it was preceded by Russian military aggression against Georgia in 2008, as well as the annexation of Crimea in 2014 and military intervention in Syria in 2015. China meanwhile has constructed military bases on artificial islands it created in the South China Sea — and pushed the Philippines out of Scarborough Shoal, which is much closer to the Filipino coast than to China.
The emergence of a new generation of strongman leaders has contributed to this new atmosphere of international insecurity. Leaders such as Putin, Xi, Trump and Benjamin Netanyahu are nationalists and self-styled men of destiny. Asked from whom Putin was taking advice in the run-up to the full-scale invasion of Ukraine, Russian foreign minister Sergei Lavrov reportedly replied: “Ivan the Terrible, Peter the Great and Catherine the Great.”
In different ways, all these strongman leaders have embraced the idea of a “might makes right” world — and turned their back on the idea of a rules-based international order. Stephen Miller, Trump’s deputy chief of staff, gave a chilling summary of this view when he told an interviewer, “you can talk all you want about international niceties and everything else. But we live in a world, in the real world . . . that is governed by strength, that is governed by force, that is
The embrace of the idea of territorial annexation by the world’s major powers [including the US vs Greenland and Canada] — alongside the gleeful embrace of a “might makes right” philosophy — clearly increases the risk of international conflict.
There is a theory of peace, nonetheless, that lies behind some of Trump’s proclamations — as well as the territorial aspirations of Putin and Xi. This is the idea that great powers should be allowed to dominate a “sphere of influence” in their immediate neighbourhoods. Their smaller neighbours — such as Ukraine, Taiwan, Canada or Cuba — must accept the domination of the regional hegemon. What is more, if all great powers respect each other’s spheres of influence, then global peace can be maintained. (…)
It ignores the views and interests of neighbouring countries. Must Canada meekly submit to the will of Trump? Is it inevitable and right that democratic Taiwan must simply submit to China’s one-party state? As Russia has discovered in Ukraine, smaller countries have rights and agency and can put up a fight. (…)
The US is unlikely to accept the absorption of Taiwan into China, as long as the Taiwanese produce 90 per cent of the world’s most advanced semiconductors. The EU will never accept that member states such as Poland or Romania should be returned to Moscow’s orbit. (…)
By and large, deterrence — underpinned by the fear of nuclear war — worked. (…)
But Trump is clearly attracted to the idea that the US should retreat from its security commitments to its allies — whom, at various times, he has denounced as freeloaders and cowards. (…)
The final crises that led to the first and second world wars both involved alliance systems and security commitments that had lost credibility. In 1914, Britain had guaranteed the neutrality of Belgium but the German government believed that the UK would not go to war over a “scrap of paper” that had been signed in 1839. That was a miscalculation. In 1939, Britain and France extended a security guarantee to Poland in an attempt to deter the Nazis. But Hitler’s previous encounters with British and French leaders convinced him that they would back down again. He was wrong.
Today, Xi has to consider whether America’s Taiwan Relations Act — which commits the US to opposing the use of force against the island — still means something. Putin must make a similar calculation about Nato’s Article Five — which commits all 32 alliance members to defend each other. Is it still an ironclad commitment — or just another “scrap of paper?”
These are the questions on which the peace of the world may depend.
