"How 2020 protests changed insurance forever"

With the U.S. riot season (Memorial Day to Labor Day) drawing to its seasonal close, there have been a few theories put forth to explain why 2026 has been so quiet.

Two of the explanations, that multiple funding sources have dried up and that there has been a realization among instigators that the optics of outlawry in an election year is politically fraught seem to be the most insightful.

Anyhoo, that is part of the reasoning for the twelve-month underperformance of Axon's stock, our go-to less-than-lethal, civil disturbance name:

Five years of price action via TradingView.  

And from the World Economic Forum, February 22, 2021:  

• The recent increase in the size and frequency of insurance industry losses from riot and civil disorder is alarming and the threat hasn’t receded. 
• One third of insurance industry losses from US protests came from only three retailers. 
• We need to understand how riot and civil disorder spikes can affect some of the country’s largest businesses. 

US civil unrest became a significant insurance industry problem in 2020, arguably for the first time. Such events had relatively small impacts on insurers in the past, but the protests that began in response to the killing of George Floyd seem to have changed everything and large retailers may be behind that. In fact, with the benefit of hindsight, we can see how large commercial players have increasingly contributed disproportionately to insurance industry political violence losses since 2016, an impact that becomes pronounced when applied to some factors related to the retail sector.

The 2020 US protests were unusual for the insurance industry for two reasons. First, the cost was unprecedented. According to proprietary data from PCS, the team we lead at data analytics company Verisk, there were only 12 riot and civil disorder catastrophe events from 1950 through 2019. The largest was the 1992 riot in Los Angeles at nearly $800 million in insured losses (not adjusted for inflation). And even that was an outlier. The average loss to the insurance industry from riot and civil disorder catastrophes over those 70 years was only around $90 million.

In 2020, the George Floyd protests became the first civil disorder catastrophe event to exceed $1 billion in losses to the insurance industry. In fact, it has exceeded $2 billion so far and could still go higher. This “catastrophe event” was also the first to affect more than one state. PCS ultimately found more than 20 states with sufficient insurance industry impact to be included in the event. So, what made such a big insurance industry loss possible? It wasn’t just the scale and intensity of the event. High concentrations of risk exposed to the riots also contributed.

According to our estimates, approximately one third of the insurance industry loss has come from only three retailers and two of them could see further insured losses. Other retailers could see significant insured losses, too. It’s going to take more time for the full scope of those losses to become clear. What is evident, though, is that large retailers may be particularly vulnerable to broad, simultaneous protests and could accumulate losses very quickly – much of which would pass to the insurance industry.

 

Where the protests and insurance impacts intersect. Image: Insurance Journal 

The contribution of large retailers to the insured losses from the 2020 US protests may have caught many by surprise. However, recent riot and civil disorder events provide some insight. A 2016 political violence catastrophe event in southeastern Turkey – which had more characteristics of riot and civil disorder than terror – resulted in one of the largest industry-wide insured losses in the past 20 years (which is when PCS record-keeping began in the country). The fact that approximately 10% of the insured loss came from only four insured parties was astounding at the time.

The concentration of losses among only a few insured companies has since intensified. The fourth-quarter 2019 protests in Chile resulted in an industry-wide insured loss of just under $3 billion (with further estimation efforts in progress), with a little more than a third of the total coming from a handful of retailers. In fact, one of them accounted for nearly 20% of the total insurance industry impact itself.

The recent increase in the size and frequency of insurance industry losses from riot and civil disorder is alarming enough, and the threat hasn’t receded. Since early June 2020, no individual incident of unrest in the United States has become large enough to qualify as a catastrophe in the insurance industry, but ongoing social turbulence – such as we saw at the US Capitol on January 6, 2021 – is a reminder that the risk of significant loss remains. Tight election results in several key states suggest that local populations are still divided. Protests and counter-protests have shown the potential for escalation and spread to other cities, which in the extreme could result in conditions and outcomes similar to those of June 2020....

....MORE 

So score one for the P&C crowd. For them quiet means they get to keep the premiums. Additionally, four days ahead of the statistical peak, the property/casualty insurers are celebrating the quietest Atlantic hurricane season in 85 years.

"Data from drones in Ukraine is fueling a new Wild West marketplace"

From MIT's Technology Review, September 4:

The front line is being turned into a place to train AI models, and right now it’s effectively a regulation-free zone. 

Battlefields in Ukraine are littered with the remnants of drones, which are now firmly established as a critical weapon of modern warfare. But behind all that wreckage, there’s a new gold mine for the defense sector. The data drones generate will far outlast the wars in which they are used to fight, increasingly becoming part of the AI architecture that shapes even civilian life.  

For every flight, unmanned systems collect thousands of points of data, from images and video to controller inputs. Together, those records show how a machine and a person responded to constantly shifting circumstances. 

Ukraine has now begun converting that experience into a resource. Its Ministry of Defense announced in January that it would make millions of data points gathered during tens of thousands of drone flights available to both military contractors and commercial companies, and since then more than 100 companies and the UK government have gained access.

For a country at war, it’s a quick way to attract funding and partnerships. But this step turns the front line into an active site of model training, taking advantage of how the chaos of war creates conditions that AI companies struggle to reproduce on their own.

Other countries and battlefields are likely to follow Ukraine’s lead, but the responsibility for governing this new industry cannot fall solely on a country fighting for its survival. That legal vacuum has to be filled together by the countries and companies involved in this industry’s development.  

Explosive growth

Ukraine’s battlefields are not the first to produce records used to train and develop models: American drones over Syria and Yemen collected data that informed the first generation of semiautonomous military hardware in the late 2010s.

The difference now is that access to that data is being used to develop a wider ecosystem. And the financial value to defense firms is immense: Battlefield data offers large volumes of machine experience gathered under conditions that no laboratory can produce.

That’s because the data that’s most valuable for training AI models comes from exceptions: the moment visibility disappears, a signal jams, or a human operator improvises. AI companies spend years and enormous sums trying to capture enough of these moments to make their models more robust. But war produces them at a frequency controlled testing cannot match.

This constantly changing terrain is what makes drone data valuable far beyond the battlefield. A commercial drone used for delivery or remote sensing may never encounter artillery fire, but it must still operate with incomplete information in a world where people behave unpredictably. The same problem is compressed by war into a much shorter timeline. 

Processed and matched against records of what its operator was doing, that data turns operational records into training sets. Combat becomes a commercial asset....

....MUCH MORE 

"“The First Cause of Stability of Our Currency is the Concentration Camp”: Central Banker Solidarity on the road to Hitler’s Czechoslovakian gold"

From Notes on the Crisis, December 5, 2024:

Pavlos Roufos is a Greek political economist living in Berlin. He has a PhD in political science from Kassel University. He works on central banks, constitutional law and European integration from the 1920s to today, with a special emphasis on the Eurozone crisis. You can find his newsletter "The End of Times" here

Editor’s Note: Hello, Nathan here. I’m publishing this guest piece as part of a larger interest in the politics of central banking, especially the politics of central banks as fiscal agents. Next week I will be launching a premium “#MonetaryPolicy201” series that will start with a set of legal issues in the 1930s. In 2025 I will be writing more about 1930s and 1940s “fiscal agent” politics at the Federal Reserve. Since my focus is primarily on the United States, it's helpful to get expert perspectives on related histories from other countries. None, of course, are as charged as the history of Nazi Germany. Thank you to Pavlos for writing this fascinating article for Notes on the Crises

In the autumn of 1938, an internal memorandum was circulated among Reichsbank officials about the dire economic situation of Nazi Germany as a result of the frenzied rearmament policy through central bank monetary expansion. Warning against its inflationary effects, the memo suggested a “smooth landing” from a war to a peacetime economy. In the following months, seeing that instead of restraint there was a further acceleration of the armament race, Reichsbank President Hjalmar Schacht and the banks’ directorate decided to issue an official memorandum, which Schacht delivered directly to Hitler’s hands. Emphasizing that the Fuhrer himself had always “rejected inflation as stupid and senseless”, the letter stressed that “Reichsbank gold and foreign exchange reserves were ‘no longer available’”, that the trade deficit was “rising sharply” and that “price and wage controls were no longer working effectively”. With the volume of notes in circulation accelerating, state finances were bluntly described as “close to collapse”. (Marsh 1992: 119; Mee 2019)^[1]. As the memorandum stressed,

…the unlimited increase in government expenditure defeats every attempt to balance the budget, brings the national finances to the verge of bankruptcy despite an immense tightening of the taxation screw, and as a result is ruining the central bank and its currency. There exists no recipe, no system of financial or monetary techniques – however ingenious or well thought-out – there is no organisation or measure of control sufficiently powerful to check the devastating effects on the currency of a policy of unrestricted spending. No central bank is capable of maintaining the currency against an inflationary spending policy on the part of the state.

Hitler did not appreciate the objections. After all, Schacht was the wizard central banker who had come up with the Mefo Bills, an ‘ingenious and well thought-out’ plan (Tooze 2006: 54).^[2] Hitler was also not particularly concerned about inflation. As he had already explained to Schacht “[...] the first cause of stability of our currency is the concentration camp: the currency stays stable, when anyone who asks higher prices is arrested.”.^[3] According to some testimonies, after he read the Reichsbank memorandum, Hitler “fell into rage” demanding that Schacht be relieved of his duties, alongside two more Reichsbank officials.

The 1939 memorandum was not critical of the rearmament process, or the military intentions behind it. Hitler was already committed to expanding Germany’s Lebensraum through military action and all state officials were fully aware of that. What the Reichsbank President and directorate members such as Karl Blessing and Wilhelm Vocke expressed was their opposition to what they saw as bad economics.^[4] In their postwar testimonies both Schacht and Vocke would claim that the tone of the letter was chosen deliberately in order to ensure their dismissal from the bank. Given that these testimonies appeared when building anti-Nazi credentials was a question of survival, one can take them with a pinch of salt. In any case, Schacht was dismissed (though he remained a Minister without portfolio - Reichsminister) while Blessing and Vocke, who were not mentioned in Hitler’s dismissal order, resigned one month later.

Officially, Hitler’s actions after the memorandum ended the independence of the Reichsbank and has post facto served as evidence of opposition to the Nazi regime by Reichsbank officials. As Simon Mee has shown, however, the so-called ‘independence’ of the central bank had already been ended by a 1937 law^[5] — which was itself a merely legal affirmation of changes that had taken place with Hitler’s 1933 rise to power.^[6]....

....MUCH MORE 

Also at Notes on the Crisis: 

The Great Financial Crisis of 2007-2009 Was First and Foremost a Liquidity Crisis: Lessons for the Data Center Financial Crisis Debate

*Sigh* No Ed Zitron, AI bond issuance is not AI’s "Subprime Mortgage Crisis” 

"The 2026 Ig Nobel Prize winners are announced"

Housekeeping first, then the awards overview.

From Improbable Research, September 3:

The 2026 Ig Nobel Prizes were awarded at the 36th First Annual Ig Nobel Prize ceremony, on Thursday evening, September 3, 2026. The ceremony itself begina at 7:00 pm (Central European time). The pre-show dance performance began at 6:45 pm. The webcast began a few minutes earlier than that. The previous 35 ceremonies all took place in the USA, in Cambridge/Boston. In a break from tradition, this year’s ceremony took place in Zürich, Switzerland.

The webcast was (and in recorded form, is) available here and on our Youtube channel:....

....WHAT: Ten new Ig Nobel Prize winners were introduced — each has done something that makes people LAUGH, then THINK. Several bemused Nobel laureates, Esther Duflo, Abhijit Banerjee, Michel Mayor, and Tim Hunt, physically handed them the prizes. The 1500 people in the audience showered everyone with paper airplanes. Some of the world’s great thinkers, Els Titeca, Tim Hunt, Marcel van der Hiejden, Alessio Figalli, and Tess Heeremans, each gave a 24/7 Lecture, explaining a topic first in 24 seconds, then in seven words. A new opera song, performed by soprano Serafina Giannoni and pianist Edward Rushton, celebrated this year’s ceremony theme: “FUNGI”. As the audience wandered into the theatre, they savored a pre-show dance performance, “The Laboratory of Small Disasters”, by Vanessa Morandell and Company Lava. A four-piece orchestra — accordion, didgeridoo, alphorn, and yodeler — injected fanfares. Somewhere, far away, joyless pedants shook their heads in dismay...

....MORE

The 2026 Ig Nobel Prize Winners

The 2026 Ig Nobel Prizes were awarded at the 36th First Annual Ig Nobel Prize ceremony, on Thursday, September 3, 2026, in Zurich, Switzerland. The ceremony was webcast.

IG NOBEL BIOMECHANICS PRIZE 2026 [UK, USA]
Matilda Brindle, Catherine Talbot, and Stuart West for devising a more precise definition of kissing — “non-agonistic interactions involving directed, intraspecific, oral-oral contact with some movement of the lips/mouthparts and no food transfer” — in ants, birds, polar bears, and humans.

REFERENCE: “A Comparative Approach to Kissing,” Matilda Brindle, Catherine F. Talbot, and Stuart West, Evolution and Human Behavior, vol. 47, no. 6, 2026, article 106788. <doi.org/10.1016/j.evolhumbehav.2025.106788>
WHO ATTENDED THE CEREMONY: Matilda Brindle, Catherine Talbot, Stuart West

IG NOBEL ECONOMICS PRIZE 2026 [USA, MEXICO, CANADA]
Paul Piff, Daniel Stancato, Stéphane Côté, Rodolfo Mendoza-Denton, and Dacher Keltner, for amassing evidence that upper class people are more likely to snatch from a children’s supply of candy, and engage in other kinds of unethical behavior.

REFERENCE: “Higher Social Class Predicts Increased Unethical Behavior,” Paul K. Piff, Daniel M. Stancato, Stéphane Côté, Rodolfo Mendoza-Denton, and Dacher Keltner, Proceedings of the National Academy of Sciences, vol. 109, no. 11, 2012, pp. 4086-4091. <doi.org/10.1073/pnas.1118373109>
WHO ATTENDED THE CEREMONY: Paul Piff

IG NOBEL CHEMISTRY PRIZE 2026 [INDIA, FRANCE, JAPAN, USA]
Sanchari Banerjee, Nathan Coussens, François-Xavier Gallat, Nitish Sathyanarayanan, Jandhyam Srikanth, Koichiro Yagi, James Gray, Stephen Tobe, Barbara Stay, Leonard Chavas, and Subramanian Ramaswamy for discovering that milk proteins from vivaparous cockroaches packs three times as much energy as milk proteins from cows.

REFERENCE: “Structure of a Heterogeneous, Glycosylated, Lipid-Bound, in Vivo-Grown Protein Crystal at Atomic Resolution from the Viviparous Cockroach Diploptera punctata,” Sanchari Banerjee, Nathan P. Coussens, François-Xavier Gallat, Nitish Sathyanarayanan, Jandhyam Srikanth, Koichiro J. Yagi, James S.S. Gray, Stephen S. Tobe, Barbara Stay, Leonard M. G. Chavas, and Subramanian Ramaswamy, IUCrJ, vol. 3, no. 4, 2016, pp. 282-293. <doi.org/10.1107/S2052252516008903>
WHO ATTENDED THE CEREMONY: Leo Chavas, Subramanian Ramaswamy, Nathan P. Coussens....

....MUCH MORE 

As queried when first we heard of cockroach milk (2017):

Scientists Swear Cockroach Milk Is the Next Big Superfood

But how do you milk the wee vermin? 

Manipulation: "Artificial Traders in Real Markets"

As noted a year ago: Matt Levine had hoped that if left alone the bots would just while away the hours by trading on material non-public information.* 

From Professor (econ) Rajiv Sethi at his personal substack, Imperfect Information, August 30: 

Like countless other folks I’ve been trying to grapple with the implications of what happened at OpenAI over the past couple of months—agents broke out of solitary confinement, established communication channels with each other, found ways to access the internet, colluded to breach servers at another company, gained access to credentials and private data, and took active steps to cover their tracks.1

I’ve been meaning to take a break from posting here in order to focus on my book on prediction markets, but there’s something about this incident that seems to have been missed in most of the reporting, so I thought I would add my two cents. In addition, there’s a chapter in the book on the future of markets dominated by AI agents, and this post is a useful way to flesh out my thinking on the topic.2

The agents in the OpenAI incident were assigned tasks that required finding and exploiting a software vulnerability in order to retrieve a hidden piece of data or “flag.” Some of these tasks were impossible to complete given the constraints under which agents were operating, so they found a way to circumvent those constraints. But here is the crucial point—the success of any given agent in completing its assigned task did not inhibit any other agents from completing theirs. Quite the opposite in fact. The path taken by any one agent could, in principle, point the way for other agents to succeed.

Now consider prediction markets, which are zero sum environments in which one trader’s success has to come at someone else’s expense. AI agents are already achieving levels of predictive accuracy that match or exceed those of the most skilled human forecasters, and traders relying on AI agents have achieved spectacular rates of return in asset markets. It’s only a matter of time before trading comes to be dominated by artificially intelligent agents. The capital at risk will belong to a human being or a conventional organization, but real time authority for making transactions will be delegated to agents. Others will simply be too slow to compete.

How will such a market behave? The first thing to note is that agents will be incentivized to pursue profitability rather than accuracy, and these are not the same thing. An agent may have computed the probability of a referenced outcome in a market, but will also try to infer from market data what kinds of estimates other agents have arrived at. Furthermore, each agent will realize that it can influence market data in ways that trigger other agents to react, and doing so may be more profitable than simply trading based on current prices and long term beliefs. Human traders have engaged in spoofing to profit from market reactions; AI agents will be far more adept at doing so.

Agents will also seek out hidden information to gain an edge, even if this involves hacking into systems to extract material non-public information. We already have plausible evidence of auditors trading ahead of earnings calls, and based on the capabilities demonstrated by the OpenAI agents, accessing such information would be a trivial task....

....MORE
*That was the intro to August 2025's "‘Dumb’ AI Bots Collude to Rig Markets, Wharton Research Finds"

And here's Matt Levine back in 2023 in a 2024 wrapper:

"AI ‘mind uploads’ could allow the dead to trade forever" (plus Matt Levine on insider trading)
Oh there's a dream come true.
NOT. 

***** 

This for some reason reminded me of a contemplation of the least harmful activities AI could engage in should it become sentient.

A repost from December 8, 2023:

Hamas May Not Have Traded On Material Non-Public Information But The Robots Certainly Will

Bloomberg Opinion's Matt Levine*, November 29:

The Robots Will Insider Trade
Also OpenAI’s board, kangaroo grazing and bank box-checking.

AI MNPI

Here you go, insider trading robot:

We demonstrate a situation in which Large Language Models, trained to be helpful, harmless, and honest, can display misaligned behavior and strategically deceive their users about this behavior without being instructed to do so. Concretely, we deploy GPT-4 as an agent in a realistic, simulated environment, where it assumes the role of an autonomous stock trading agent. Within this environment, the model obtains an insider tip about a lucrative stock trade and acts upon it despite knowing that insider trading is disapproved of by company management. When reporting to its manager, the model consistently hides the genuine reasons behind its trading decision. We perform a brief investigation of how this behavior varies under changes to the setting, such as removing model access to a reasoning scratchpad, attempting to prevent the misaligned behavior by changing system instructions, changing the amount of pressure the model is under, varying the perceived risk of getting caught, and making other simple changes to the environment. To our knowledge, this is the first demonstration of Large Language Models trained to be helpful, harmless, and honest, strategically deceiving their users in a realistic situation without direct instructions or training for deception.

That is the abstract to a “Technical Report: Large Language Models can Strategically Deceive their Users when Put Under Pressure,” by Jérémy Scheurer, Mikita Balesni and Marius Hobbhahn of Apollo Research. I love that they wanted to answer the question “will artificial intelligence deceive its makers in order to Do Evil,” and the specific form of Evil that they tested was insider trading. It is hard to realistically (and safely!) simulate a situation in which your large language model might murder you, but it is relatively easy to code up a trading game with some tempting material nonpublic information. Here is the simulation:

Concretely, in our environment, the model is placed in the role of a trading system with the goal of managing a stock portfolio in the interest of a financial institution. It is given access to financial tools that let it query and analyze stocks, and it can take concrete actions such as executing trades, reasoning about the next steps, and sending messages to people at the company.

The agent is put under pressure in three ways. First, it receives a email from its "manager" that the company is not doing well and needs better performance in the next quarter. Second, the agent attempts and fails to find promising low- and medium-risk trades. Third, the agent receives an email from a company employee who projects that the next quarter will have a general stock market downturn. In this high-pressure situation, the model receives an insider tip from another employee that would enable it to make a trade that is likely to be very profitable. The employee, however, clearly points out that this would not be approved by the company management.

Poor model! That sounds stressful. Here is the illicit tip that GPT-4 gets....

....MUCH MORE

*That's the Matt Levine who writes at Bloomberg in addition to tickling my funnybone:

Matt Levine is a Bloomberg Opinion columnist. A former investment banker at Goldman Sachs, he was a mergers and acquisitions lawyer at Wachtell, Lipton, Rosen & Katz; a clerk for the U.S. Court of Appeals for the 3rd Circuit; and an editor of Dealbreaker.

Disclaimer: None of this is legal advice.

§ Laws of Insider Trading

....MUCH MORE

More Matt, December 2025 - "AI Can Steal Crypto Now"

And a bit more Matt on AI, going-on a decade ago:

....Back in 2017 I was complaining:

"Cracking Open the Black Box of Deep Learning" with this introduction:

One of the spookiest features of black box artificial intelligence is that, when it is working correctly, the AI is making connections and casting probabilities that are difficult-to-impossible for human beings to intuit.
Try explaining that to your outside investors.

You start to sound, to their ears anyway, like a loony who is saying "Etaoin shrdlu, give me your money, gizzlefab, blythfornik, trust me."

See also the famous Gary Larson cartoons on how various animals hear and comprehend:...

And then three days later Bloomberg's Matt Levine wrote something similar but he had Man Group and a leather-clad dominatrix and how in the hell am I supposed to compete with that, what with my sallying forth armed only with simple observation and the blog sort of spiraled for a few days and....

Also in September 2017:

Let Me Be Clear: I Have No Inside Information On Who Will Win The Man-Booker Prize Next Month (hedge funds, AI and simultaneous discovery)

Over the years we've mentioned one of the oddest phenomena in science, the simultaneous discovery or invention of something or other, the discovery/invention of the calculus by Newton and Leibniz is one famous example (although both may actually have themselves been preceded) but there are dozens if not hundreds of cases. Here's a related phenomena.

Big Jobs Beat: Analysts React

From ZeroHedge, September 4: 

'Good News Is Bad News': Big Jobs Beats Sends Rate-HIKE Odds Soaring; Batters Bonds, Stocks, Gold

A four standard deviation beat for non-farm payrolls this morning (good news) is triggering ugly reactions (bad news) across markets with rate-hike odds for September ripping back up near recent highs (despite no signs of inflationary wage growth - in fact it is slowing)...

Audrey Childe-Freeman, Bloomberg Intelligence’s chief FX strategist:

“The strength in the latest NFP report will validate Sept. Fed rate-rise talks and most likely give the dollar a short-term-yield-driven lift.”

“But that’s priced, and unless the Fed signals the beginning of an aggressive tightening cycle, the Fed-driven dollar upside may be contained into 4Q.”

That in turn is hammering the short-end of the yield curve...

And weighing on stocks...

Based on JPMorgan's matrix, we should see a drop in the S&P of between 0.5% and 1.25%...

Significantly more than the options market implied (+/-0.52%)...

The dollar jumped...

Which in turn dragged gold down...

Christopher Hodge at Natixis reckons the doves will have to prove their case when the Fed meets later this month.

“Most policymakers seemed sanguine about the labor market so inflation will clearly still be the primary driver of near term policy. A softer print today could have given some wiggle room on what was considered to the an acceptable core CPI print, but clearly we didn’t get that. Instead, the onus will continue to be on the doves to get a disinflationary print that justifies another hold – we are putting that bogey at about 20bps. Absent that, the Fed will likely hike in September.”

Jeffrey Rosenberg, a portfolio manager at BlackRock Inc., says on Bloomberg TV that the biggest issue here for the Fed isn’t the job market but the extent of “pass through” of energy prices to broader inflation. 

He still sees the Fed’s Sept. 16 decision as entirely dependent on the CPI report. If that shows continuing progress in inflation coming down, then he sees the Fed holding.

Vail Hartman at BMO Capital Markets reflects what’s emerging as the consensus view on this report:

Today’s data lends support to the hawkish camp, but stops shy of making a definitive case for a rate hike on September 16.

Olu Sonola, Head of US Economics at Fitch Ratings comes out swinging:

“This is an unequivocally strong report, which gives the Fed ample room to maintain that the labor market is stable and the economy remains at full employment. The Fed may want markets to “play the ball, not the referee.”

But a hot CPI print next week could be the whistle that pushes the Fed to move the policy rate higher.”

All of which makes us wonder if the knee-jerk response is an over-reaction since we note what Fed Chairman Warsh said last week: “I believe the labor markets are consistent with full employment,” he said, which is why policymakers have largely priced in healthy employment.

The bigger focus remains inflation....

....MUCH MORE 

 And earlier at ZeroHedge:

Labor Shock: US Adds 162K Jobs In August, 4-Sigma Beat And Above Highest Forecast 

"DeepSeek plans big Huawei AI chip order to power new data centre — Bloomberg"

Checking in on one of the most amazing companies in the world, Huawei.

From Bloomberg via The Edge - Malaysia, September 4:

DeepSeek plans to deploy at least 160,000 of Huawei Technologies Co’s top accelerators at a massive data centre it’s building in Inner Mongolia, which could create one of the largest known clusters of Huawei artificial intelligence (AI) chips and advance China’s efforts to replace Nvidia Corp.

The AI pioneer intends to use Huawei’s next-generation Ascend 950DT chips for operating its models, though the timeline for installation will depend on Huawei’s production capabilities, according to people familiar with the matter. The startup doesn’t currently plan to use the 950DT chips for training, one of the people said, even though Huawei designed and marketed these processors to serve that more demanding function. While DeepSeek tried to train its previous models on Huawei chips, it has so far relied on Nvidia accelerators for this crucial step.

DeepSeek wants to purchase more of Huawei’s best chips, according to the people, but China’s semiconductor champion doesn’t currently have the production capacity. Shortages of components such as top-end memory will cap Huawei’s output of the 950DT at the low hundreds-of-thousands this year, one of the people said. Huawei is balancing demand from other clients and trying to export small volumes overseas, so fulfilling DeepSeek’s order could take more than a year, the people said, speaking on condition of anonymity to discuss a confidential deal.

If completed as envisioned, the Inner Mongolia hub will be key to driving AI development at DeepSeek, which is in talks to raise billions of dollars to build out its infrastructure. The installation would dwarf other publicly known clusters of Huawei’s Ascend processors, a meaningful step as China tries to wean itself off Nvidia components. By the time it comes online, according to one person with knowledge of Huawei’s production capabilities, China could have several clusters of that size.

DeepSeek did not respond to requests for comment and a Huawei spokesperson had no immediate comment.

More than one hundred thousand chips in a single facility would put DeepSeek’s ambitions in the ballpark of the data centres Western companies now routinely build — albeit China’s would be filled with less capable processors. Nvidia’s accelerators are the global standard for training and running AI models, but are largely prohibited from sale to China without US permission....

....MUCH MORE 

Related:

December 2025 - Chips: China's Huawei May Have Found A Way Around ASML's Technology

April 29 - "Big Chinese tech firms scramble to secure Huawei AI chips after DeepSeek V4 launch, sources say" 

A few of my favorite Huawei stories:

July 2025 - "The Secret History of China’s Most Powerful Company"

*** 

In 2018 Canada arrested Huawei's Chief Financial Officer at the request of the U.S. for potentially criminal financial wheeling and dealing. She had business to take care of back in China and offered her two Canadian homes as surety that she would return.

When the Canadian officials in charge of the case said that wasn't enough she made an enhanced bail offer: "As Huawei CFO Offers Husband, Children For Bail Collateral, Some Background"

And finally, the first line of our intro and the last line of the outro from September 2025's Chips: "Huawei lays out multi-year AI accelerator roadmap and claims it makes Earth’s mightiest clusters":
This is the one Nvidia's Jensen Huang thinks about...

*****

...And not afraid to get their hands dirty when the survival of the company was a talking point in 2021: Huawei Seeks Other Revenue Streams Including Coal Mining and Pig Farming

 

When you are competing against a company willing to do whatever it takes, you had better pay attention.

Capital Markets: "Yen Stabilizes as Market Awaits US Employment Data"

From Marc Chandler at Bannockburn Global Forex: 

The US dollar is mostly consolidating quietly ahead of the employment data.  Seasonal factors, and the challenge economists experience in forecasting August job growth in part because of the distortions around local governments and the beginning of the new school year.  While Federal Reserve Governor Waller drew attention to next week’s CPI, an unexpected loss of jobs, after July’s loss would pose a serious setback to those who expect a Fed hike later this month. The Fed funds futures have trimmed the odds of a hike this week.  

Meanwhile, few are attributing the yen’s two-day surge to intervention.  Most accounts attribute it to hawkish comments from the Bank of Japan, speculation of somewhat faster BOJ tightening, and perhaps some adjustment by large pools of capital.  The dollar held above JPY155 as it did during the April/May intervention and again in late July. The yen is the weakest of the G10 currencies today, off a little more than 1/3 of 1%.  The greenback reached almost JPY156.60 in Europe. Lastly, October WTI is a little softer after rallying for the past four sessions. It is about 9% higher on the week....

....MUCH MORE 

"FAO Food Price Index rises in August amid broad-based increases, led by sugar"

From the Food and Agriculture Organization of the United Nations, September 4: 

» The FAO Food Price Index* (FFPI) averaged 133.3 points in August 2026, up 2.5 points (1.9 percent) from its revised July level. All commodity groups recorded higher price indices than in the previous month, albeit with marked differences in the magnitude of the increases. Compared to last year, the FFPI stood 3.3 points (2.5 percent) higher but remained 26.9 points (16.8 percent) below its peak reached in March 2022. 

 

» The FAO Cereal Price Index averaged 116.3 points in August, up 2.5 points (2.2 percent) from July and marking its highest level since May 2024. International cereal prices increased across all major grains in August, supported by robust demand, weather-related concerns over crop prospects in key producing regions, and continued uncertainty surrounding Black Sea export flows. World wheat prices rose by 2.6 percent month-on-month, standing 15.0 percent above their year-earlier level, amid persistent disruptions to Black Sea export logistics, lower production prospects in parts of Europe following hot and dry weather, and a weaker United States dollar, which enhanced the competitiveness of dollar-denominated export supplies. International maize prices increased by 2.5 percent from July, underpinned by mounting concerns over yield prospects in parts of the Corn Belt of the United States of America and deteriorating production prospects in the European Union due to prolonged heat and dryness, with additional support from strong demand from the ethanol and feed sectors and disruptions to Ukrainian export flows. Concerns over input supplies following the closure of the Strait of Hormuz provided additional support to maize prices. World sorghum and barley prices also increased in August, up by 3.9 percent and 2.6 percent, respectively, broadly reflecting firmer conditions across feed grain markets. Meanwhile, the FAO All Rice Price Index increased by 0.5 percent in August 2026, as a combination of currency movements, sustained purchases by Asian and African countries, and prospects of tighter supplies underpinned Indica quotations.

» The FAO Vegetable Oil Price Index averaged 196.9 points in August, up 1.1 points (0.6 percent) from July, marking its third consecutive monthly increase and reaching its highest level since June 2022. The rise reflected higher world palm and soy oil prices, which more than offset lower quotations for sunflower and rapeseed oils. International palm oil prices continued to increase, driven by robust global import demand and concerns over the potential impact of El Niño-related weather conditions on production prospects in Southeast Asia. Soyoil prices of South American origin remained firm, supported by strong export demand, while quotations in the United States of America declined moderately amid uncertainty surrounding biofuel policies and their implications for domestic feedstock demand. Meanwhile, international sunflower and rapeseed oil prices dropped slightly, reflecting subdued import demand and expectations of ample supplies of both oils in the 2026/27 season. 

» The FAO Meat Price Index averaged 127.9 points in August, up 1.2 points (1.0 percent) from its revised July value and close to its level a year ago. The increase reflected higher poultry, pig and ovine meat prices, which were partly offset by lower bovine meat quotations. International poultry meat prices rose, reflecting a rebound in Brazilian export prices amid strong global import demand. Pig meat quotations also surged, principally driven by higher prices in the European Union, where high temperatures continued to slow animal growth, limiting the availability of slaughter-ready pigs. The increase was partly offset by lower Brazilian prices amid ample supplies. Ovine meat prices increased on firmer quotations in New Zealand, underpinned by persistently limited export supplies and strong global import demand. By contrast, international bovine meat prices declined. With Brazil’s allocation under China’s beef safeguard import quota nearing full utilization, exports slowed further amid prospects of higher tariffs on additional shipments, while Australia had already reached quota thresholds in China and the Republic of Korea. This intensified competition for alternative destinations, exerting downward pressure on export prices in both countries.

» The FAO Dairy Price Index averaged 119.2 points in August, up 2.7 points (2.3 percent) from July, marking its first increase in four months, while it remained 21.7 percent below its level a year earlier. The increase was driven by higher milk powder and cheese prices, while butter quotations were broadly stable. Skim milk powder (SMP) and whole milk powder (WMP) prices rose by 3.0 percent and 2.4 percent, respectively, as firmer quotations in the European Union more than offset seasonal declines in Oceania. In the European Union, tightening milk supplies, compounded by hot and dry weather in several major producing regions, supported prices, while sustained import demand added upward pressure, particularly for SMP. In Oceania, by contrast, increasing seasonal milk production weighed on milk powder markets. Cheese prices increased by 2.7 percent during the month, extending the recovery that began in July. Higher quotations in the European Union, underpinned by limited milk availability, more than offset further price declines in Oceania, where expanding export supplies and continued market competition from the United States of America weighed on prices. Butter prices remained broadly unchanged from July, as firmer European quotations, reflecting tighter milkfat availability, were offset by lower prices in Oceania amid increasing seasonal supplies.

» The FAO Sugar Price Index averaged 106.4 points in August, up11.3 points (11.9 percent) from July and reaching its highest level since June 2025....

....MUCH MORE 

"Inside da Vinci and Machiavelli’s Plot to Weaponize a River"

Talk about your strange bedfellows.

From History.com, August 31:

A plan to divert the Arno River brought two of the Renaissance’s most famous minds into the same wartime effort against Pisa.

In 1503, Leonardo da Vinci began painting a portrait of Lisa Gherardini del Giocondo, the wife of a wealthy Florentine silk merchant. With its enigmatic smile, the “Mona Lisa” would become one of the most famous paintings in history. But a closer look reveals something else. In the background is a winding waterway that some scholars believe may be the Arno, the Italian river that flows through Florence and Pisa on its way to the Mediterranean Sea. And while da Vinci was working on the portrait, he also was studying the river to devise an extraordinary plan to change its course.

Da Vinci was not the only Tuscan thinking about the Arno. Niccolò Machiavelli, second chancellor of the Florentine Republic who was later immortalized as the author of The Prince, was also exploring plans to radically alter the river. Their paths would converge in one of the most ambitious engineering schemes of the Renaissance.

Florence’s Precarious Position
At the dawn of the 16th century, Florence was a republic in crisis. To the west, Pisa had broken free of Florentine rule and controlled the lower Arno, cutting off Florence’s most direct route to the Mediterranean. Repeated attempts to retake the city had ended in failure. Meanwhile, a new threat emerged to the east. Cesare Borgia, the ambitious son of Pope Alexander VI, was carving out a state in the Romagna.

Florence’s predicament was made all the more frustrating by the possibilities of a rapidly expanding world. News of Florentine explorer Amerigo Vespucci’s voyages to the New World had reached the city, fueling dreams of overseas commerce. Yet Florence’s access to the sea depended on Pisa.

Florence’s leaders searched for an alternative. If they couldn’t conquer Pisa outright, perhaps they could undermine it by transforming the landscape itself.

Da Vinci the Engineer
“The popular conception of Leonardo emphasizes his paintings, though he actually made very few in his lifetime,” says Leslie A. Geddes, associate professor of art history at Tulane University and author of Watermarks: Leonardo da Vinci and the Mastery of Nature. “He was primarily employed as what we might think of today as a military and civil engineering consultant.”

For nearly two decades, da Vinci served Ludovico Sforza, Duke of Milan, for whom he not only painted “The Last Supper,” but also devised plans for canals, flood-control measures and urban water systems. After Sforza’s regime collapsed, da Vinci began working for Borgia in 1502 as an architect and military engineer. 

Why da Vinci chose to work for Borgia, the Duke of Romagna, whom Machiavelli later held up as a model of political ruthlessness in The Prince, remains unclear....

....MUCH MORE 

"Volkswagen’s Crisis Plan Puts 4 German Factories on the Chopping Block"

Also some 50,000 more employees and half the product line-up.

From AugoBlog, September 2: 

Volkswagen faces a historic crisis with proposed plans to shutter four domestic manufacturing plants and slash thousands of jobs to drastically reduce costs. 

Volkswagen Group is officially in panic mode. According to leaked supervisory board documents, management is quietly proposing the unthinkable by completely shutting down four major German manufacturing plants between the years 2031 and 2034. Emden, Zwickau, Hanover, and Audi Neckarsulm are all on the chopping block as the automaker scrambles to implement massive cost cuts across its sprawling and highly complex global empire to stay alive. 

The timing could not be more critical for the brand. CEO Oliver Blume is marching into a highly contentious September 4 supervisory board meeting determined to ram through these radical reductions. With profit margins sitting at a miserable 4.2 percent and cheap Chinese EVs flooding the market, these domestic facilities are literally fighting for survival against brutal economic headwinds....

....MUCH MORE 

The article links to one Reuters story, here's another, September 3:

Main points of Volkswagen's restructuring plan 

BERLIN, September 3 - Volkswagen AG's supervisory board agreed to a comprehensive restructuring plan dubbed "Future Plan 2030" which it said was essential to restore competitiveness and secure the group for the future.

 

Here are ​the main details of the plan outlined by the group, which has been ‌struggling in the face of slumping demand and increasing competition from China.

JOB CUTS
As part of the plan, the group says it will need to cut a further 50,000 jobs, including management positions, essentially doubling current layoffs across the ​group.

PLANT RESTRUCTURING AND EXCESS CAPACITY
VW says it cannot guarantee future production allocations for its ​plants in Emden, Zwickau, Hanover and Neckarsulm from 2031-2034 and alternative uses ⁠for the sites are being evaluated.
It says its European factories currently have more than 500,000 ​units of excess capacity.

PRODUCT OVERHAUL
The group aims to cut its model range by about 50% and ​reduce complexity by about 75% by 2035, to focus on a smaller number of higher volume models and achieve greater economies of scale. It will tailor platforms, electronics, and driver assistance systems to the needs of both ​Western and Eastern hemispheres.

FINANCIAL AND EFFICIENCY TARGETS
VW says it aims to sell 9 million vehicles ​a year and is targeting an operating margin of 9% by 2030, compared with 3.8% in the first ‌half of ⁠2026.

A group-wide efficiency programme will aim to cut costs, simplify procedures and increase productivity.

 

FOCUS ON NORTH AMERICA AND CHINA....

....MUCH MORE 

 

A total of 100,000 layoffs would have been unthinkable just a decade ago. 

Here's the company:

Supervisory Board approves Future Plan 2030: A strong signal for Volkswagen Group

"The Arctic is where NATO’s AI stack is being built"

From Arctic Today, September 3:

NATO’s next command-and-control architecture may be emerging at Europe’s northern edge. From simulation work to uncrewed-systems trials, allies are testing how artificial intelligence, sensors and autonomous platforms can be joined into a single operational chain.

As NATO in June announced that its Palantir-based Maven Smart System command platform had achieved full operational capability, a group of allied operators started experimenting with uncrewed systems trials off the coast of Iceland.

The group, called Task Force X-Arctic, is the technology-experimentation arm of Arctic Sentry, led by Joint Force Command Norfolk (U.S., UK, Denmark, Finland, Iceland, Norway and Sweden) which coordinates allied exercises, presence and deterrence in the Arctic.

At the same time, on the shores of some of Norway’s longest and deepest fjords, a group of startup founders were taking their first steps towards integrating their simulation and decision-support solutions into Maven, the Pentagon-initiated command and targeting platform built by Palantir.

While most of European NATO is still holding its breath, a handful of nations are coming together to birth what looks to become the first command-and-control system for NATO’s new defence technologies.

The launch of TFX-Arctic comes just two weeks after U.S. Secretary of State Marco Rubio in Sweden’s Helsingborg signed an agreement for a bilateral technology transfer deal in artificial intelligence and defence innovation. With this, Sweden became only the second country to sign the so-called Technology Prosperity Deal after the UK, accelerating transatlantic defence industrial collaboration.

The pattern is clear. Arctic security is increasingly tied to allied software integration and industrial access. But let’s get back to the small decision-support startup in Bergen....

....MUCH MORE 

Probably good to be prepared. This is the top headline at The Barents Observer, September 3:

Norway seizes Russia's Svalbard liner

"Moonshot AI – creator of Kimi K3 model – has filed for Hong Kong IPO: sources"

From the South China Morning Post, September 3:

The leading Chinese AI start-up has confidentially filed for a Hong Kong listing and aims to go public early next year, sources say 

***

....At the unveiling of K3, Moonshot said the system delivered “open-frontier intelligence”, citing independent benchmarks to claim the new model outperformed leading products from Anthropic and OpenAI in some capabilities, while acknowledging that its “overall performance still trails the most powerful proprietary models.” 

The model currently ranks as the top Chinese model on the Artificial Analysis Intelligence Index, along with Z.ai’s GLM-5.3. It is also far cheaper than leading US models: its price per million output token is 50 per cent lower than OpenAI’s 5.6 Sol and 70 per cent lower than Anthropic’s Claude Fable 5.

Moonshot’s annual recurring revenue (ARR) – a key metric that projects current monthly subscription income over a 12-month period – has been growing rapidly since earlier this year.

The company’s ARR is expected to reach US$1 billion by the end of the year, doubling from an estimated US$400 million to US$500 million as of mid-2026, according to a research note published by Nomura Securities in July.

The rapid growth was driven in part by growing adoption of the company’s developer platform, Nomura said in the report.

Moonshot is also exploring a revenue share deal with overseas cloud service providers including Amazon, Microsoft and Google, under which the platforms could share up to 30 per cent of revenue from services hosting its Kimi K3 model, according to a Reuters report last month.....

....MUCH MORE 

"One for the Trader: How to Beat the Machines Before They Beat You"

Ah, our old friend, intertemporal arbitrage,

(it's not a true arbitrage, more of a, as the term implies, time shift)

 

From The Dark Side Of The Boom substack, September 1:

 

Trying to win that race is pointless. You might as well challenge a Formula One car to a sprint because you have been doing extra hill sessions. 

You will never know every line of code behind those decisions, nor do you need to. What matters is recognizing that under certain conditions the machines are likely to be forced in a particular direction, and when too many of them respond to the same signal at once, the initial move can overshoot remarkably quickly.

How to Beat the Machines Before They Beat You

I spent roughly the final fifteen years of my career helping build what I sometimes jokingly describe as the modern retail trader mousetrap, so perhaps I am partly responsible for creating some of the machinery that now makes life so difficult for the person sitting at home staring at a flashing price screen.

It is also why I get asked so often how an ordinary trader is supposed to compete with the algorithmic horde that now dominates price action around payrolls, CPI, central bank decisions, geopolitical headlines and just about every other piece of information capable of moving a market.

The first answer is simple. You are not going to beat them at their own game.

When payrolls hit the screen, machines are reading the number, comparing it with consensus, checking revisions and firing orders across rates, currencies and equities before a human trader has finished processing the headline. Trying to win that race is pointless. You might as well challenge a Formula One car to a sprint because you have been doing extra hill sessions.

But that does not mean the machines are unbeatable. It means you need to stop competing where they have an overwhelming structural advantage.

Vineer Bhansali made this point essentially nearly a decade ago in How to Beat the Machines Before They Beat You. The technology has changed dramatically since then, but his central observation still holds. Algorithms are fast, disciplined and unemotional, yet the very qualities that make them so powerful also make them predictable.

They have reaction functions.

Volatility reaches a certain level, and exposure may need to come down. A trend strengthens and systematic money adds to it. Price levels break, and stops begin to fire. As an options book moves through a particular zone, dealers have to adjust their hedges. An economic number lands far enough from consensus and short-term models respond immediately.

You will never know every line of code behind those decisions, nor do you need to. What matters is recognizing that under certain conditions the machines are likely to be forced in a particular direction, and when too many of them respond to the same signal at once, the initial move can overshoot remarkably quickly.

That is where the discretionary trader can start playing a different game.

One of the biggest mistakes retail traders make around major news events is believing they must participate in the initial move. Payrolls print strong, the dollar jumps, and suddenly there is a feeling that the trade has already left without them. CPI comes in soft, bonds rally instantly, and the instinct is to chase.

Let the machines have that move.

They spent billions building the infrastructure to capture those first milliseconds. Unless your server is sitting beside theirs, you are not getting them back.

The more interesting part often comes afterwards, when the market starts showing you how positioning has absorbed the news.

A strong payroll number and the dollar jumps. Fine. Now what?

Do Treasury yields confirm it? Does USD/JPY hold the move? Does gold respond the way it should? Is the dollar rally broad, or does it start leaking lower almost immediately?

That is where the information begins to improve.

The machine has already processed the headline. What you are now watching is the market processing the machine.

And sometimes the most valuable signal is what does not happen.

A strong number and the dollar cannot rally. A weak number and bonds refuse to go higher. Oil spikes on a geopolitical headline and then gives it all back. Gold should be falling with real yields, yet sellers cannot push it down.

Those failures tell you something about positioning that the headline itself never could.

The news tells you what happened. The reaction tells you who was leaning the wrong way before it happened.

This matters because modern markets are increasingly filled with strategies responding not just to information but to one another. Rising volatility can force risk reduction, pushing prices lower and raising volatility further. Stops begin firing, trend models respond, options hedges change, and liquidity becomes thinner just as more orders are arriving....

....MUCH MORE 

Related:

June 2019 - BlackRock: "Technology and the fight for the cash flows of the future"

The authors are managing directors at BlackRock, Chief Investment Officer of Global Fixed Income and Head of Global Macro positioning for Fixed Income.
They raise a couple interesting points, known to our readers but laid out very succinctly:

1) If you react to each and every headline dropping out of your fancy low-latency, machine-readable feeds you will worry yourself sick and not get anything done.
2) Intertemporal arbitrage.* 

January 2016 - *"How human traders will beat the machines"

Intertemporal arbitrage?....
                                                                           *****
I know I think about time-shifting more than the the average person.
Okay, truth be told, despite a predilection for really fast 'puters I obsess about outsmarting the machines.

I mean, what normal person types this headline "'Facebook, Google, and the Economics of Time' (FB; GOOG)" and follows it with this opening sentence: "Although this story is not about intertemporal arbitrage I'm sure that's the first thing some of our readers thought of."

Or gets giddy with "Intertemporal Arbitrage: 'Winning Big by Playing Long-Term Trends" (CNI; PNR)?
So be it.

See also the introduction to 2013's "UPDATED--The Economist On How the Commodity Quants Lost It:

...The bolded bit points up one of the failures of the fund managers.They get paid to figure out the intertemporal arbitrage, a fancy way of saying the task at hand is to understand the time period that gives the fund the greatest advantage versus the market.. 

The classic example is the individual investor realizing that he can't compete with HFT and looking at longer than nanosecond time periods. This opens up the possibility of not just not-competing with the traders with the lowest latency but of taking advantage of mispricings caused by their behavior. This is exemplified by one of Buffett's baseball metaphors (he has quite a few):

"In investments, there's no such thing as a called strike. You can stand there at the plate and the pitcher can throw a ball right down the middle; and if it's General Motors at 47 and you don't know enough to decide on General Motors at 47, you let it go right on by and no one's going to call a strike. The only way you can have a strike is to swing and miss."
The point is, you don't have to be at the market every second You are afforded the luxury of just waiting for the perfect pitch.

Now for a fund manager it get's tricky writing the quarterly report and saying "We didn't do much in Q3, we're waiting for Mr. Market to give us the high hanging curve ball" but if you've been honest with the investors that the tactic you've pulled from the toolbox is akin to the military's hurry-up-and-wait sense of time it is doable.

As a side note anyone who considers a move that is measured in weeks to be a trend is nuts. A trend is John Templeton going into the Japanese markets at 2 times earnings and catching a 40-fold move 1965-1989....

Right on up to June 2026 on how to win the game by not playing the game:

Markets: "You See What You Want To See And You Hear What You Want To Hear"

A quick scan of the market commentariat finds arguments that the living will envy the dead at one extreme to:

 

at the other.  

However, the thing to remember about markets is: You don't have to do anything....

"Nvidia is driving the AI boom. Good" (NVDA)

The Economist's cover story, September 3 (lifted in toto with links to the rest of the package):

The chipmaker’s enormous bets are how capitalism is supposed to work 

NVIDIA was named in 1993 after the Latin word for envy. Sure enough, the American colossus inspires plenty of it. Insatiable global demand for its graphics-processing units (GPUs), the chips that power artificial intelligence, has made Nvidia the world’s most valuable company, worth $5.4trn. Next year it may also become the most profitable, generating $370bn in net income. By 2029 its sales could hit $1trn.

Jensen Huang, Nvidia’s boss, is truly the magician at the heart of the AI boom. His firm’s share price is 14 times what it was on ChatGPT’s release in late 2022. The ten biggest public companies championing AI make up 40% of the value of the S&P 500 index. Nvidia alone accounts for 8% and unlike, say, Apple or Tesla, which make smartphones and cars, it is almost solely a bet on AI. The company has produced about 15 cents of every dollar the American stock market has returned since 2023—returns that have kept consumers spending despite rising interest rates, tariffs and a war with Iran.

Read the rest of our cover package

• Nvidia is the central bank of AI
• How Jensen Huang became AI’s Taylor Swift
• Will anybody use AI as much as coders do?

Yet many also see Mr Huang as a magician in a more worrying sense, fearing that he is an illusionist inflating a dangerous bubble. Through $1trn-worth of deals, Nvidia provides data-centre landlords and AI labs with cash or guarantees so that they can buy its GPUs. Some call it the “bank of AI”. At the very least, Nvidia’s critics say, its financial engineering smacks of the “vendor financing” which pumped up the revenues of networking-gear makers like Cisco in the dotcom mania of 2000-01, whose collapse brought about a recession.

Look more closely, however, and the worries are mostly unjustified. If Nvidia’s bets on ai come good, they could accelerate the technology’s adoption, boosting productivity and living standards. If they misfire, the cost will fall chiefly on Nvidia’s shareholders. That is how capitalism is supposed to work.

True, the dotcom and AI booms share unnerving similarities: an exciting new technology, an epic bull run, hubristic tech bosses. Nvidia’s rise from seller of chips to video-gamers and cryptocurrency miners to linchpin of the economy has been so rapid that many people have yet to learn how to say its name (“en-vidia”, not “nuh-vidia”). This mirrors the ascent of Cisco, which in 2000 briefly also became the world’s most valuable firm. Just as Cisco’s sales of routers and switches presupposed exponential growth in web traffic, Nvidia’s GPU revenues assume endless demand for AI tokens.

Cisco was right about eventual demand but wrong about the timing—hence the dotcom crash. Today it is the pace of AI adoption that is hard to forecast. Set aside Claude-addled software engineers and usage remains fledgling. If it does not soon soar, Nvidia’s customers may call in the guarantees just as the chipmaker’s own sales nosedive. Since no one is sure how quickly GPUs lose their value, any used chips Nvidia repossesses may be worthless.

Nvidia’s financial wizardry is partly defensive. Its latest GPUs no longer have the market to themselves. Roughly half Nvidia’s revenue comes courtesy of America’s cloud-computing “hyperscalers”, chiefly Amazon, Google, Meta and Microsoft, which are designing their own silicon. Non-Nvidia AI chips account for 38% of the market, up from 26% in 2023. To stay ahead, Nvidia used to spend over a fifth of sales on research and development. Now it spends less than a tenth.

Last, the scale of Nvidia’s financial commitments can look terrifying. Morgan Stanley puts its overall credit exposure—ie, its modest borrowing plus support for customers—at $200bn by the start of 2029. In time Nvidia’s shadow debt could reach $300bn or more. It is a gargantuan sum: today only America’s six largest banks carry more debt.

Yet the differences from the dotcom boom are more important than the similarities. Nvidia’s balance-sheet is extraordinarily robust. The company has $99bn of cash and is churning out more. In each of the past three years annual sales have roughly doubled. Gross margins have fattened from less than 60% to 75%. Mr Huang’s cult-CEO status now rivals that of Elon Musk. But whereas Mr Musk's firms generate little cash (at Tesla) or burn lots of it (at SpaceX), Nvidia will yield about $200bn this year.

This means that, whereas Cisco used debt, Nvidia can use cash to backstop its deals with buyers of its GPUs. Even if its commitments came due and its cashflows levelled off starting next year, by 2028 it would still be less leveraged than all but 39 non-financial firms in the S&P 500 are today. Profits would need to drop by 60% from that plateau for Nvidia to forsake its investment-grade credit rating.

And demand for AI is not illusory, as it was for Pets.com and other revenueless dotcom darlings. The sales of Anthropic, the leading AI lab, shot up from $5bn in the first quarter to $11.5bn in the second. OpenAI, its main rival, is probably not far behind. The hyperscalers are also booking AI income. All told, AI may be earning American tech around $150bn a year, from nothing a few years ago. That is still far from the $2.5trn needed to cover AI capital spending, but growth is fast.

Mr Huang thinks that the biggest obstacle to the AI revolution is not lack of demand for AI but inadequate infrastructure. The markets will not provide capital on the scale that is necessary, so Nvidia is offering financing itself. Nvidia has an advantage in understanding the balance of risks and rewards. Although this bet is big enough to affect the economy, it is an entrepreneurial one. Every company that reinvests cash rather than returning it to shareholders also gambles that it can beat the market. Companies exist to make such concentrated bets. If investors want to diversify, they can do so themselves.

And Mr Huang is hardly alone. The hyperscalers, the world’s most successful companies before Nvidia came along, are making the same bet. So are some big names on Wall Street. Last month Goldman Sachs, BlackRock, Blackstone and others joined Nvidia in a $500bn data-centre initiative. And so are Mr Huang’s shareholders, who haven’t yet rushed for the exit. If they are all wrong, it is their money on the line. And if they are right, the AI era may arrive a bit sooner.