Huge AI Bonuses in South Korea Spark Fight Over Sharing Tech Wealth

A second-tier port city in South Korea is emerging as the epicenter of a growing global fight over how to split the spoils of the artificial intelligence boom.

Home to a mega-factory run by Samsung Electronics Co., Pyeongtaek nearly shut down this month as employees threatened a strike that risked disrupting global supply chains. The chip giant ended up offering some full-time employees in its memory division bonuses as high as $400,000 — roughly four times Samsung’s average annual salary — to defuse the crisis.

Yet that solution created a new set of problems. Under the tiered payout system, only employees in Samsung’s semiconductors division — about 60% of its Korean workforce — will get the largest payments. Workers who aren’t involved in chip production are set to get only $4,000 apiece, a mere 1% of the biggest winners in the labor fight.

The outcome enraged a sizable chunk of Samsung employees, prompting thousands to quit its biggest union during the wage negotiations. One employee, who requested anonymity to speak openly, said the gaping compensation divide had killed motivation for workers who don’t make memory chips. (…)

How can governments [and companies] equitably spread the benefits of a technology that threatens to render large swathes of the workforce redundant? (…)

Samsung is forecast to generate about 330 trillion won ($221 billion) in operating profit this year, with SK Hynix not far behind. Together, the two firms could contribute more than 100 trillion won in corporate taxes annually. That could exceed the amount the government expected from all companies as recently as April.

The question of how to handle a potential newfound pile of cash and spread the benefits is top of mind for South Korean officials. Kim, the labor minister, has suggested a possible social solidarity wage and vowed in a Facebook post that the government “must not simply stand by and watch the gap between workers continue to widen.” (…)

South Korea is also reportedly looking to allocate some excess tax revenue generated by the chip industry’s super-boom to a “Korean-style sovereign wealth fund.” The government said in a statement that it believed “society would benefit from discussing openly and constructively” how to deal with the AI windfall, and that no decisions had been made regarding the use of future excess tax revenues.

In the Middle East, countries such as Saudi Arabia and the United Arab Emirates have used sovereign wealth funds to channel oil income into global investments and domestic diversification, though these have had limited success in spreading gains broadly across society. Other approaches have focused more on direct redistribution. Alaska’s Permanent Fund pays annual dividends to residents from oil revenues — a rare example of a universal payout that has helped reduce poverty, though the payments are relatively modest. (…)

As workers in South Korea become increasingly vocal about the windfall from the AI supercycle, Barclays economist Bumki Son said it could evolve into a broader social issue with more people feeling “relatively deprived.” (…)

Bonuses from Samsung and SK Hynix — which already distributes 10% of operating profit in cash bonuses — are increasingly becoming reference points for wage negotiations across corporate Korea, and beyond. Asia’s other major chip giant Taiwan Semiconductor Manufacturing Co. reassured workers in the wake of Samsung’s decision that a 30% bump in profit-sharing payouts was coming their way, after a chorus of online complaints. (…)

“A rare historical possibility now lies before Korea,” Kim wrote on Facebook. “The possibility of becoming not only a country that supplies AI infrastructure, but the first country to return the excess profits of the AI era to human life.” (…)

For some, the gap between the bonuses of chip workers and electronics staff hurts all the more because for decades Samsung’s smartphone and home appliances divisions sustained South Korea’s biggest private corporation. Woo Ha-kyung, acting head of the company’s second largest union, said the decision to “gloss over issues” with a mere $4,000 payment to staff not involved in chipmaking undermined their contributions. (…)

Separately, the company has unveiled a $3.33 billion fund to be deployed over the next five years for supplier support, AI talent development and community assistance. (…)

As the Samsung standoff gripped South Korea, Labor Minister Kim Young-hoon emerged as a central figure. After his last-minute mediations helped stave off a strike, the former union leader held a press conference telling reporters he understood it wasn’t just Samsung staff who felt left behind, but also the general public. (…)

Turning point?

Samsung and SK Hynix are private corporations benefitting from the AI boom. They are not the only ones. But the AI riches are not spread equally, far from it.

It’s not only revenues, profit margins are exploding, also unequally:

Even the dash line will eventually stop rising.

Productivity growth has accelerated post-pandemic but labor is not sharing proportionately. This next chart needs explaining.

The nonfarm business price deflator is an implicit price index that measures price changes for the value-added output produced by the U.S. nonfarm business sector, whereas the CPI measures price changes in a fixed basket of goods and services purchased by consumers. They differ in what they cover (business output vs consumer purchases), how they treat imports/exports, and in the index formula (chain-weighted deflator vs fixed basket).

The NFB deflator tracks prices received for domestically produced output. Weights correspond to shares in sectoral value‑added, so capital‑intensive or high‑margin industries carry more weight in the index.

CPI can rise faster than production‑side deflators over long periods because it includes imports and is heavily weighted toward categories like services and housing that may have faster price growth. Researchers like Ed Yardeni  emphasize that the “productivity–pay gap” is much smaller when you use the nonfarm business deflator rather than CPI, arguing that it’s closer to the true split of productivity gains between wages and profits inside the business sector.

The CPI‑deflated compensation line answers a different question: how fast has workers’ pay grown relative to the consumer cost of living, including imports and heavily weighted items like shelter and medical care.

Debatable between economists but consumers will always argue that they have not shared productivity gains since the 1970s. The 1973–74 and 1979 oil shocks drove up consumer energy prices and general CPI inflation much faster than wages, compressing CPI‑deflated pay. Also, housing, healthcare, education, and some services became a larger share of consumer spending and saw faster price growth than many goods, particularly after China joined the WTO and began exporting lower priced goods.

BLS work and later research show that, when you deflate compensation with a production‑side index (like the GDP or nonfarm business deflator), the productivity–pay gap is modest through the 1980s and 1990s and then widens noticeably from the early 2000s. The green NFB‑deflated line tracks productivity fairly well until roughly 2000–02, then lags.

Multiple studies (and simple decompositions of labor share) find that the nonfarm business labor share trends down and mark‑ups/profit shares trend up starting in the late 1990s/early 2000s. Output per hour is rising faster than compensation per hour valued at output prices, boosting profit margins.

The post‑1995 IT‑driven productivity acceleration initially translated fairly well into pay, but over time more of the gains were captured by owner’s income and high‑return intangible capital rather than broad‑based compensation. Offshoring, global supply chains, and increased competition for routine labor further weakened the bargaining position of many workers, so compensation growth lagged the productivity of the firms they worked for.

A larger share of total compensation goes to high‑income workers (equity‑based pay, executive comp, finance/tech) after 2000, while median or typical workers see slower gains.

So the early‑2000s break reflects a distributional shift inside the business sector: an increasing share of value added started going to capital and high‑end labor, not to broad‑based compensation, which is why the green line peels away from productivity.

This was particularly noticeable in services where productivity really accelerated since 2000.

The more recent jump in margins is impressive…

…but not so in mid and small cap companies:

Beware of extrapolations, particularly given workers’ growing reaction to the AI boom, coming right after COVID.

This chart plots real income measures (blue and black) with corporate profits after taxes all indexed to early 2020:

Keeping the trend in profits in mind, this next chart shows how rising inflation has eaten away most of the nominal increase in disposable income since the pandemic.

The AI boom with its exploding corporate profits and stock market gains is compounding the pain from the open inflation wound (pandemic and war “created”).

Profit margins cannot keep rising like they have lately. (Thanks to Ed Yardeni for many charts above).

These two Axios charts sum it up:

 

Data: U.S. Bureau of Economic Analysis; Chart: Matt Phillips/Axios

BTW:

Trump Clears Way for Corporate Tax Dodge Hidden in the Fine Print

A year ago, the Trump administration withdrew from a global effort to curb offshore tax-dodging by multinational companies. That decision has been a huge gift to corporate America, enabling companies to avoid at least $40 billion in income taxes since the beginning of 2025.

A New York Times review of securities filings from nearly 500 companies showed that they avoided taxes by attributing hundreds of billions of dollars in earnings to low- or no-tax foreign locales like Cyprus, Bermuda, Switzerland and the Cayman Islands. Often, corporations funneled the profits through subsidiaries in places where they had no employees, offices or customers.

Tax havens became more appealing after President Trump signed an order on his first day back in office withdrawing the United States from a 13-year international effort to end such schemes. The effort led dozens of countries to impose a minimum corporate tax and rules for pursuing companies using tax havens. After House Republicans passed legislation last year targeting some of those countries with a new tax, international officials agreed to exempt U.S. companies from much of the crackdown.

American Express avoided paying $423 million in taxes last year using the island of Jersey. PayPal trimmed its taxes by nearly half during 2025 thanks to its units in Singapore. Stanley Black & Decker cut its bill by $27 million — nearly one-third — using the island of Cyprus.

A favorite destination was the tiny Mediterranean island of Malta, where Abbott Laboratories, the pharmaceutical giant, has claimed all its global profits were earned by a subsidiary with no employees. Malta helped the company cut its tax bill by $336 million last year, the filings show.

Companies making similar moves spanned nearly every sector of the economy: Walmart and Uber; Mastercard and Pepsi; Crocs and Merck; Honeywell and Cigna. To put the $40 billion in taxes they avoided in perspective, it would be enough to triple the annual budget of the Federal Aviation Administration or U.S. Customs and Border Protection.

On the face of it, the offshore tax strategies don’t necessarily violate any laws. But the Internal Revenue Service says some of the companies have gone too far, and tax advisers say the Trump administration’s actions will make it easier to pursue even more aggressive dodges.

“Accommodating the U.S.’s refusal to participate in the global reforms opens up the door to abuse,” said Philip Marcovici, a former chair of the European tax practice at the law firm Baker McKenzie.

The Times’s analysis relied on a new disclosure required by federal accounting rules. For the first time, in annual 10-K reports filed with the Securities and Exchange Commission, public companies are required to include footnotes reporting the precise amount of tax avoided through each foreign jurisdiction.

Some companies using tax havens to avoid U.S. income tax rely on federal funding for their profits. (…)

The Trump administration’s agreement with the Organization for Economic Cooperation and Development this year frees U.S. companies to park profits in favorable locations — often in conflict with I.R.S. enforcement efforts. (…)

Time is not on Trump’s side:

Trump Says Deal Will ‘Work Out Well’ Even as US, Iran Clash

Trump, in a Truth Social post late on Sunday US time, said constant speculation over whether he’ll agree to a deal — which will likely see the two sides extend their ceasefire by around two months, with Iran reopening the strait and the US lifting a blockade of Iranian ports — weren’t helping.

“It is MUCH tougher for me to properly do my job and negotiate, when political hacks keep negatively ‘chirping,’” Trump said, “that I should move faster, or move slower, or go to war, or not go to war, or whatever. Just sit back and relax, it will all work out well in the end.” (…)

Iran’s semi-official Tasnim news agency, which has close ties to the IRGC, said Sunday that both sides continued proposing amendments to the draft deal, though stressed there’s no guarantee they’ll reach an agreement. (…)

On Saturday, Iranian state TV reported the existence of a new draft agreement, which it said gives the Islamic Republic “exclusive authority to determine the nature of transiting vessels” in Hormuz, a negotiating point the US is unlikely to accept.

The draft also said the US has committed to giving Iran access to $12 billion in frozen funds within 60 days, to be sent directly to Iranian banks without restrictions, according to Iranian TV. It said the document was “unofficial” and not “finalized.”

Meanwhile, Windward last week updated us on the Strait:

Day 89 is the analytical milestone, not just a date. Three months into Operation Epic Fury, the Hormuz operating environment looks structurally different from how it looked on February 28, and the difference is not closure relaxing, but closure restructuring into permission-based control, eastward export bypass, and embedded strait-body IRGC presence.

The April 7 ceasefire and the May 23 announcement that an agreement has been “largely negotiated” sit on top of an operational picture that has hardened. Iranian crude exports are at a fraction of pre-conflict volumes, but Iran has built, and is continuing to build, the channels to keep volumes moving without Hormuz transit at all. The Kuh Mubarak SPM, the Bandar-e Jask sequencing yard, the Chabahar holding area, and the Goreh-Jask pipeline together form an export apparatus that did not exist operationally three months ago.

If the announced diplomatic framework materializes and the strait reopens, this architecture remains. If the framework does not materialize, the architecture absorbs an increasing share of Iranian crude flow.

The indicators to watch in the coming cycle are whether the Kharg loading cadence sustains after the May 25 Suezmax restart, whether the Bandar-e Jask sequencing yard accumulates additional VLCC-class tonnage, whether IRGC small-craft posture remains in open strait-body configuration or returns to coastal patrol, and whether the dark ship-to-ship trade off Fujairah and Muscat continues to expand its operator pool.

The campaign has not de-escalated. It has restructured. The next phase will determine whether the restructuring becomes permanent.

Could Trump’s Iran ‘excursion’ be a bigger global turning point than Vietnam?

This is by Patrick Wintour, The Guardian’s Diplomatic Editor

(…) Assured by limitless military superiority and filled with such noble intent, US presidents have repeatedly been lured into launching wars only to find themselves confounded, ensnared and then broken by their inability to overpower an inferior opponent they wholly misjudged.

(…) Trump’s “little excursion to Iran”, judging by the drafts of the potential peace agreements that are circulating, is being universally perceived as a defeat. Almost regardless of the outcome – most likely a return to the old status quo – the war looks ill-conceived, a monument to confused objectives, bad planning and misplaced assumptions. (…)

By comparison with the Vietnam odyssey, Iran feels more like a day trip.

But in terms of consequence, it is still possible that the “excursion” will prove to be the bigger geopolitical turning point for the unrivalled superpower, the moment when the US will have to concede it mishandled a war not just because it had no convincing battle plan, but also no grand strategy to match how the contemporary world works. In an interconnected world, Trump believes progress is achieved through conflict, not cooperation. (…)

Clearly the domestic US consequences of Iran will never match Vietnam. True, the war was unpopular from the start, but society has not been torn apart by it. Only 13 body bags, each a personal tragedy, have been sent home. At the most, inflation caused by the energy shock will ensure an already unpopular president is punished in the midterms, something he professes not to concern him.

But it is arguable that the international consequences of the Iran war could yet prove more long lasting. (…) Trump’s war of choice looks to be a signal of defeat that will have an effect in several fields.

It marks the collapse of Israel’s 20-year Iran strategy to produce regime change and will accelerate the already rapid decline in the influence of this Israeli government in Washington. Danny Citrinowicz, the former head of an Iran branch of Israeli military intelligence, describes the war as an operational success but a strategic fiasco for Israel.

The war is also prompting Gulf monarchies to profoundly reappraise their geopolitical relationships, including the question of whether the existence of US bases brings the security required for their economies to diversify. (…) claims by Trump that countries such as Saudi Arabia or Qatar would now normalise relations with Israel, or join the Abraham accords, sound absurd – in the words of the former US ambassador to Israel Dan Shapiro: as “delusional as a moon made of green cheese”.

The Gulf states would prefer an imperfect peace because they see no other way out, Barbara Leaf, a former US undersecretary for the Middle East, told a seminar last week. (…)

The fallout is likely to hit Europe hard. As a squeeze on living standards seeps through the global economic system over the next year, centrist incumbents in France, Germany and the UK may face an electoral beating that tears at the architecture of the EU. The task of the incumbents will be made harder if Trump acts on his threat to withdraw US troops from Nato states in retribution for their “cowardly” refusal to come to his aid.

For the US foreign policy establishment, exemplified by the Council on Foreign Relations, the missteps in Iran are the final confirmation that Trump’s highly personalised, instinctive system of predatory diplomacy creates only more disorder.

Last week, the CFR launched a fundamental review of US strategy post-Trump. Its convener, Rebecca Lissner, has already warned the war “has delivered a potentially fatal blow to a US-led international order that was already on life support”. Allies are hedging, middle powers are forming their own coalitions, and regions once firmly in Washington’s orbit are shifting toward new power centres, she said. The former state department official Mira Rapp-Hooper was more brutal at Chatham House, describing it as superpower suicide.

In the short term, two questions from the Iran war have been thrust upon the Democrats, and in effect have already been answered. Has the US interest been furthered by being so close to Israel and its leadership? Would the US not be more powerful if it returned to alliances built on values, and the law, as well as self-interest?

For Iran, weakened, impoverished and yet emboldened, the path is unclear. Tehran may yet have to make concessions in the talks on its nuclear programme, including many it was on the verge of offering in Geneva in February. Iran’s internal politics is unpredictable, but this is a more military government, and at the same time the hardest hardliners in parliament have been marginalised.

Ali Vaez from the International Crisis Group says the war has given Iran three presents: ideological revitalisation, the discrediting inside Iran of foreign military intervention, and the repair of its deterrence strategy. The US deployed its ultimate deterrent on Iran – war – and it did not work. In the strait of Hormuz, Iran has realised how geography and globalisation have given it an immeasurable asset, one that it will take years of new pipeline construction to devalue.

Not surprisingly, so universally damning are the global verdicts on Trump’s war that he agonises and balks at signing a document that will in essence get him back to where he started, at a cost of $50bn. (…)

Trump’s repeated threats to blow countries up have an eerie resemblance to Richard Nixon’s delirium, as described by the former White House chief of staff HR Haldeman in his memoirs.

Haldeman recalled Nixon explaining he “could force the North Vietnamese into legitimate peace negotiations. The threat was the key, and Nixon coined a phrase for his theory … He said: ‘I call it the Madman Theory, Bob. I want the North Vietnamese to believe I’ve reached the point where I might do anything to stop the war. We’ll just slip the word to them that, ‘for God’s sake, you know Nixon is obsessed about communism. We can’t restrain him when he’s angry – and he has his hand on the nuclear button’ – and Ho Chi Minh himself will be in Paris in two days begging for peace.” (…)

Trump’s fallback message that Iran must never have a nuclear weapon had several drawbacks. Iran had agreed to this in the deal signed in 2015 from which Trump withdrew during his first term. Moreover, Trump said he had completely and totally obliterated Iran’s ability to make such weapons in the attacks mounted in the brief war of June 2025.

A succession of experts, including the former EU negotiator for the 2015 deal, Federica Mogherini, tore into Trump’s claim that Iran was close to possessing a bomb. “There was no evidence that Tehran posed an imminent nuclear threat or that diplomacy had been ineffective”.

As a result, she said, the war was illegal and reckless from the very first day. She said: “Analysts predicted that going to war with Iran would empower the country’s most conservative hardliners, spread conflict across the region, and drive global energy prices to punishing levels”. The analysts were largely right.

Increasingly exasperated White House briefers turned to the role Benjamin Netanyahu had played in persuading Trump to attack Iran. In a recent 60 Minutes interview, the Israeli prime minister insisted it was misleading to say he had forced Trump into war. Both he and Trump jointly weighed the risks, but he admitted “the problem of the Hormuz strait became understood as the war went on”.

This was an astonishing admission. Fatih Birol, the chief executive of the International Energy Agency, recently disclosed that in job interviews at the IEA, after asking candidates why they are applying for a job at the IEA, the second is: “What would you do if the strait of Hormuz was closed?” It was a commonplace doomsday scenario, yet the US had to improvise a response.

Equally few in the Pentagon foresaw the extent to which Iran would resort to “triangular coercion” – the attack on oil and gas facilities of the Gulf states, as well as exposed US bases.

International relations literature claims this is a relatively unstudied phenomenon whereby “a coercer who lacks direct leverage over a resilient target coerces a third party who does possess leverage over the target, and to whom the target is vulnerable, and manipulates it into a clash of interests with the target”.

In short, the war might not influence the US itself, but it might get to those that could. It was the alliance of Saudi Arabia, Turkey, Qatar, Egypt and Pakistan that last weekend foreclosed Trump’s return to conflict. They can now hold the reins in the Middle East, and it will be the relationship they can forge with Iran, independent of the US, that matters.

Canada Dips Into a Surprise Recession

Canada edged into a technical recession as weak business and government spending drove a slight contraction in the first quarter, pointing to persistent slack in the economy amid the US trade war.

Real gross domestic product fell by 0.1% on an annualized basis during the first three months of the year. That follows a 1% contraction in the fourth quarter, a downward revision from a previously reported 0.6% decrease.

The weaker-than-expected GDP data coincides with a weaker job market as well, painting a sobering picture of the Canadian economy as US President Donald Trump’s tariffs squeeze Canadian businesses. Business capital investment in the first quarter posted a fifth consecutive decline, shrinking 3% on an annualized basis, driven by lower spending on engineering structures.

CONSUMER WATCH

Americans are spending faster than they earn it

The personal saving rate fell to 2.6% in April, down from 3.2% in March and 4.3% in January — a sharp slide that brings it to its lowest level since mid-2022.

• Consumer spending rose 0.5%, even as disposable personal income fell 0.1%, the Commerce Department said Thursday morning.
• That gap between how fast consumer incomes are rising and how quickly they are spending is driving the drawdown in the saving rate.
• Gasoline and energy goods were the single-largest driver of spending increases in April, one sign of how the war’s energy impact is registering in household budgets.

The Personal Consumption Expenditures Price Index, the Federal Reserve’s preferred inflation gauge, rose 0.4% in April, cooling from 0.7% in March at the height of the energy shock.

• There is still little evidence of the shock spilling over into non-energy-related categories. Core PCE, which excludes food and energy costs, gained 0.2% — cooling slightly from March.
• Still, compared with the prior year, core PCE ticked up to 3.3%, its highest level since 2023. As Fed governor Lisa Cook put it in a speech Wednesday: “Inflation is clearly moving in the wrong direction.”

Before the pandemic, Americans were saving at roughly double today’s rate, though that cushion has been eroded by two consecutive inflation shocks in the span of four years.

Real per capita disposable income — the money consumers can spend after accounting for taxes and inflation — declined 1.4% in April from a year ago. It also dropped 0.4% in March.

Also:

Quarterly results from Dollar Tree, Best Buy and Kohl’s suggest that consumers are still willing to spend on bargains and small indulgences — yet are becoming increasingly selective as higher gas prices and other everyday costs put pressure on household budgets, Axios’ Kelly Tyko writes.

• Their stocks surged today: Dollar Tree (up 17.9%), Best Buy (up 15.8%) and Kohl’s (up 20.5%).

After the market close, Costco reported results that supported the trend, showing 9.8% comparable sales growth.

Goldman Sachs:

Real GDP growth was revised down by 0.4pp to +1.6% (quarter-over-quarter annualized) in Q1.

Consumer spending growth was revised down 0.2pp to +1.4%, largely reflecting downward revisions to healthcare spending from the incorporation of Quarterly Services Survey (QSS) data.

Business fixed investment growth was revised down slightly (-0.4pp to +10.1%), while housing investment growth was revised up (+1.7pp to -6.3%).

Domestic final sales growth was revised down 0.1pp to +2.7%. The contribution of inventory accumulation to GDP growth was revised down 0.3pp to +0.1pp.

Real gross domestic income, an alternative measure of economic activity based on different source data than GDP, grew +0.9% (quarter-over-quarter annualized) in Q1.

Personal income was flat in April, below expectations and partially reflecting a 0.2pp drag from a decline in payments from the Farmer Bridge Assistance program, which had boosted personal income in March. Employee compensation rose 0.2%.

Year-over-year real disposable personal income growth now stands at -1.1%, or -0.4% after excluding the volatile farm income component and the negative base effect from a large increase in Social Security payments last year.

Personal spending rose 0.5% in April, in line with consensus expectations. Real personal spending rose 0.1%, reflecting a 0.2% increase in real services spending but a 0.1% decline in real goods spending.

The core PCE price index increased 0.24% in April, slightly below expectations. Monthly core PCE inflation was revised up by 0.01pp to 0.30% in March and by 0.03pp to 0.40% in February.

Year-over-year core PCE inflation ticked up to 3.29% in April. Headline PCE increased 0.40% in April, and the year-over-year rate increased to 3.77%.

U.S. consumers are seeing prices climb, and not just for fuel

Consumers in the United States are feeling the crush of rising prices. But while the war with Iran gets much of the attention, when volatile products such as food and energy are stripped out, core goods prices are climbing at the fastest pace since Ronald Reagan was on his way out of the White House.

The personal consumption expenditures price index, the measure of goods and services inflation most closely watched by the U.S. Federal Reserve, jumped 3.8 per cent in April from the year before. That was the fastest rate in three years, and it was heavily driven by soaring gasoline prices.

But below the surface, prices for core goods have risen spectacularly.

Two things appear to be at play: tariffs and the artificial intelligence boom. (…)

In a note published by the U.S. Federal Reserve last month, researchers analyzed the impact of tariffs on core goods categories and determined import duties explained much of the increase in prices up to February, 2026. That pressure is set to ease, at least for now, after the U.S.’s top court overturned Mr. Trump’s emergency tariffs, though other tariffs remain in place.

Yet tariffs aren’t entirely to blame, either. Another report published last month by researchers at the Federal Reserve Bank of Minneapolis found prices for some goods are surging more than can be explained by tariffs alone and they pointed to AI-induced demand for electronics equipment, in particular.

Indeed, in Wednesday’s PCE report, the tech category of goods jumped 10 per cent annually. (…)

Here’s Ed Yardeni’s chart on goodsflation. Ed writes: “In his final press conference, Fed Chair Jerome Powell stated he expected the inflationary impact of tariffs to fade within two quarters. That’s not happening yet.”

PCED services is up 3.5% YoY, stubbornly high. The wage-sensitive super-core measure of PCED services (ex energy and housing) also rose 3.5%.

The latest Redbook Retail Sales Index confirms that consumer spending remained strong through May. It rose 8.9% y/y during the week ending May 22, well above the 2025 full-year average of 5.8% (chart).

Chevron CEO warns oil prices to jump over summer as supplies dwindle

(…) “The buffers and the shock absorbers are being steadily drawn down, and the ability for the market to absorb this imbalance is drastically diminished today versus where we started,” he said at a conference organised by the investment bank Bernstein on Thursday.

“Over the next few weeks, we’re likely to see those pressures flow through more directly to physical prices and there’s more upwards pressure that I would expect as we get into June and certainly into July.”

Wirth’s comments follow a 10 per cent fall in oil prices over the past week amid optimism that the US and Iran can agree a deal to end the three-month-long conflict that has closed the Strait of Hormuz (…).

They highlight growing concern among economists that the war’s impact on energy prices will continue to be felt for many months after any deal is agreed to end it. The conflict has removed 12mn-13mn barrels of oil a day from global markets.

The comments by Wirth echo a growing chorus of warnings from other oil executives, including the head of the United Arab Emirates state oil group Adnoc, who cautioned last week that full oil flows through the Strait of Hormuz were unlikely to return before next year even if the conflict is resolved.

“It will take at least four months to get back to 80 per cent of pre-conflict flows, and full flows will not return before the first or even second quarter of 2027,” Adnoc chief executive Sultan al-Jaber said during an Atlantic Council event on May 21. (…)

Wirth said the energy crisis would force governments to focus more on “an insurance policy” by building up oil reserves to insulate them from shocks such as the pandemic and wars in Iran and between Russia and Ukraine. (…)

“If this goes on for long, it tips us into an economic slowdown or a recession, you might have an offset on the demand side, which you can’t rule out.”

• Exxon Mobil warns global oil stocks may hit critical lows within weeks

Exxon Mobil has warned that global crude oil inventories are set to fall to their lowest level on record in the next few weeks, a situation that would force oil prices sharply higher and put pressure on energy markets.

“We are approaching oil inventory levels that have never been this low before,” Neil Chapman, senior vice-president of Exxon, said at a conference hosted by Bernstein in New York.

Chapman warned with concern: “I mean very, very low levels, extremely low. You can debate whether levels that critical will be reached in 2 or 3 weeks, but whenever we get there, you will see oil prices shoot up immediately.”

The executive added that Dated Brent prices in the physical market would jump to US$150 to US$160 a barrel when oil stockpiles fall to record lows in the coming weeks.

He said that once prices rise to a certain level, demand destruction would help pull the market mechanism back into balance. (…)

Americans Are Falling Behind on Their $1.25 Trillion Credit-Card Bill Soaring interest rates and stubborn inflation have led to highest delinquencies since the financial crisis; ‘a pattern of survival debt’

(…) In the first quarter of this year, the percentage of credit-card balances that were at least 90 days delinquent rose to 13.12%, according to data released in May by the Federal Reserve Bank of New York. That’s the highest level in 15 years, and the most since the period following the 2008 financial crisis. (…)

 

America’s total credit-card balance stood at $1.25 trillion in the first quarter, according to the New York Fed, up from $1.18 trillion [+5.9%] in that quarter last year. That’s the highest first-quarter balance since the New York Fed began recording the measurement in 1999. (…)

Americans straining to pay off their debt have flooded credit-counseling agencies. The National Foundation for Credit Counseling, a network of nonprofits that helps people reduce credit-card debt, said it had 24% more clients in January than a year earlier. Its average number of monthly clients was 60% higher this year than in 2018. (…)

Americans carry, on average, about $6,500 to $6,700 in credit-card debt, according to credit-reporting agencies. The percentage of cardholders with balances of more than $10,000 has risen in communities across different income levels since 2018, analysis from the Urban Institute data shows. Last year, 17% of cardholders in low-income communities held balances greater than that amount, 20% in medium-income communities and 25% in higher-income communities. (…)

Middle-class households in particular are struggling to pay down balances as more families “shift to a pattern of survival debt,” said Bruce McClary, spokesman for the foundation. (…)

Bloomberg:

As the graphic below shows, the fortunes of real estate and stocks do a shockingly good job of explaining why some consumers can throw caution to the wind and continue to spend even as savings dip perilously close to zero. Ballooning 401k accounts give people a false sense of security that their retirement financing is ahead of schedule and that they can spend most of their income and even run up debt. Market crashes, in those instances, exacerbate economic pain.

In the extreme cases, the results were recessionary. Wealth is easy-come, easy-go, and an economy that’s sustained by home equity or the stock market is inherently flimsy.

Corporate America Is Starting to Ration AI as Cost Skyrockets Executives are scrambling to track returns on AI investments as the bill for massive computing needs comes due

(…) Some enterprises have hit their annual budget in just three months or reported seeing their AI spending bills double or triple. (…)

But a number of investors and tech executives cautioned against betting on a pullback, noting that sales and usage by corporate AI customers have climbed far faster than forecasts. (…)

Just a few months ago, the prevailing sentiment around AI use at many big companies was the more, the better. All-you-can-eat subscriptions amounted to a subsidy by the model-makers, which often lost money on the intensive activity of power users. Exhorted to embrace the wave of change, employees at some companies engaged in tokenmaxxing, or using as much computing as possible in order to be seen as AI-forward—a practice that continued even as the model companies shifted to usage-based pricing. (…)

Higher costs may eventually steer users toward cheaper models that cost a fraction of the price, but many companies remain wary of such AI systems because several of the cheapest options were developed in China, according to executives. Anthropic, OpenAI, Google and others also offer cheaper versions of their flagship models, and Factory and others have developed systems to help companies triage queries and steer some tasks to cheaper options.

Token use continues to grow immensely. Google said at a recent event that it now processes over 3.2 quadrillion tokens a month, seven times as much as a year ago. The company and others are seeking to reduce the cost of AI use in a variety of ways, including increasing computing efficiency.

That shift to usage-based pricing has forced enterprise customers to reckon with their consumption. An Uber executive said by March, the company had blown through its annual budget for agentic, or autonomous, AI use. Microsoft limited access to an Anthropic program for some employees who can use an internal coding assistant instead. Salesforce introduced a system for tracking how token use ultimately contributes to positive business outcomes. (…)

An Anthropic spokeswoman said the company’s models help customers achieve greater productivity, such as completing complex tasks in less than two weeks that would have taken more than seven months in the past.

“As with any new technology and way of working, teams are still discovering where the biggest gains are and how best to measure them,” she said. “We’re working with customers to give them the tools to make sure the return is something they can see, not just feel.”

Software engineers and startup executives warn that even though it is possible to complete tasks far more quickly, spending on debugging, reviewing and rewriting AI-generated code remains high, indicating that the models still need to be improved.

For companies using advanced AI coding tools, only 18% of spending on tokens is translating into shipped coding products that reach real users, according to EntelligenceAI, a startup that aggregated data on more than 2,000 companies using advanced AI tools for coding.

• Amazon scraps AI leaderboard to stop workers chasing usage scores Senior executive Dave Treadwell tells staff ‘don’t use AI just for the sake of using AI’ as costs rise

Amazon has shut down an internal leaderboard that tracked employees’ use of AI tools after workers tried to boost their scores with unnecessary activity that increased the company’s computing costs.

Employees at the $2.9tn group were told this week its “Kirorank” service — which scored users of Amazon’s Kiro developer platform based on their AI activity — had been taken offline, according to two people familiar with the matter. (…)

Meta employees have similarly sought to boost their position on internal tables by driving up token consumption. (…)

AI labs such as Anthropic have recently shifted to a consumption-based pricing model away from monthly flat fees, in a move that significantly increased the costs of some customers. Amazon uses Anthropic’s AI models extensively.

Amazon had started to use a metric called “normalised deployments”, evidence of engineers regularly using AI to create useful code, to measure the success of its AI tools and adoption of the technology rather than outright token consumption, the people added.

Treadwell told staff that he did not want workers to focus on token use and instead instructed them to focus on building better products.

Compute usage and cost will likely follow patterns similar to the internet in the 1995-2005 period. When internet use became widespread in the 1990s, there was both an immediate problem with individual overuse and a massive consumer cost explosion known as “bill shock”. Because early internet infrastructure was priced by the minute or hour rather than as a flat rate, staying online too long resulted in astronomical, unexpected bills for ordinary households.

The technological novelty combined with the sudden ability to talk to strangers globally triggered immediate societal patterns of overuse.

It was common for families to receive unexpected monthly phone and internet bills totaling hundreds or even thousands of dollars because children or teenagers left the dial-up connection running to browse early websites or use chat rooms.

The cost explosion ended abruptly when AOL shifted to a flat-rate pricing model of $19.95 per month for unlimited access. While this saved consumers from ruinous bills, it created a new infrastructure crisis.

Demand for faster internet services soared as the capabilities and offerings of online services expanded. With the advent of online shopping, streaming media, and an array of cloud-based services, dial-up connections quickly became obsolete. Users required more bandwidth to support these activities—broadband was the solution to this ever-growing demand.

The growth of broadband and e-commerce became deeply intertwined. Each fueled the other’s expansion; e-commerce companies needed customers to have fast and stable internet connections to access their services, while the expansion of broadband opened new markets and opportunities for these companies to reach wider audiences.

• DeepSeek Makes 75% Price Cut on V4 Pro Permanent

From Business Analytics Newsletter:

For the past 18 months, enterprise AI budgets have been locked in a predictable pattern: pay frontier prices, accept frontier performance, negotiate volume discounts. DeepSeek just changed the floor permanently.

The original 75% promotional discount on V4 Pro was set to expire May 31, 2026. Instead, DeepSeek announced the rates are now the standing price, not a promotion. The reason: architectural efficiency, not market pressure.

Long-context inference is expensive. As enterprise AI workloads grow RAG pipelines with large retrieval prefixes, code review agents scanning full repositories, legal document analysis token costs compound fast. A team running 500M input tokens/month at standard frontier rates could pay over $1,000/month in input costs alone, before output.

V4 Pro was engineered from the ground up to cut long-context inference cost, using a Mixture-of-Experts design that activates only 49 billion of its 1.6 trillion parameters per forward pass. Combined with a 1M-token context window and aggressive caching (cache-hit input now at $0.003625/M), the architecture makes the price cut structurally sustainable.

In plain English:

• Baseline: GPT-5.5 estimated at ~$30/M output tokens; V4 Pro original price at $3.48/M output

• After Optimization: V4 Pro permanent pricing at $0.87/M output tokens (75% reduction from original)

• Business Impact: At 1B output tokens/month, switching from GPT-5.5 to V4 Pro saves approximately $2.4M/year; enterprise deployments with cache-heavy workloads can realize even greater savings via the $0.003625/M cache-hit rate

At 34x cheaper than GPT-5.5’s estimated output pricing, this permanently shifts the cost calculus for enterprise AI workloads teams running high-volume RAG pipelines, code review agents, or long-context inference can now achieve seven-figure annual savings compared to closed-source alternatives, while self-hosting the open weights for full data sovereignty.

If your team is still defaulting to GPT-5.5 or Claude Opus 4.7 for cost-sensitive batch workloads, benchmark DeepSeek V4 Pro this week the 80.6% SWE-bench score means coding and reasoning quality is now within striking distance of frontier models at a fraction of the cost.

Anthropic Rockets to $965 Billion Valuation, Topping OpenAI in AI Showdown

The company has emerged as the front-runner in the AI race and is on track next month to hit $50 billion in “annualized revenue”—a metric startups use that employs short-term sales to forecast a yearly figure. That figure grew 80-fold in the first quarter. (…)

The $965 billion valuation more than doubles Anthropic’s previous value. The company’s valuation growth is the fastest in venture-capital history, according to PitchBook Data. Anthropic reached its latest valuation roughly 3 years and 2 months after launching its first product, per PitchBook. (…)

The company’s revenue is set to more than double to $10.9 billion in the second quarter, an explosive rate of growth that will help it turn an operating profit for the first time, The Wall Street Journal has reported. (…)

Instead of chasing chatbot use, Anthropic focused on business customers, especially coding automation, viewing semiautonomous software-writing as a potential takeoff point for more advanced AI capabilities. (…)

OpenAI, which closed a $122 billion funding round earlier this year, has released a competing tool called Codex, and the two companies are locked in a ferocious competition for business users. (…)

The Information says that OpenAI generated $5.7 billion in first-quarter revenue, boosted by demand for Codex.

The FT adds:

Anthropic initially targeted a $30bn raise from financial institutions. The company exceeded that total in part thanks to the participation of infrastructure partners. This added to $15bn in previously committed funding from Big Tech “hyperscalers”, including $5bn from Amazon, to fill out the $65bn raising. (…)

The round comes as private investment giants Apollo and Blackstone were putting together a roughly $36bn debt deal to purchase custom chips designed by Google and Broadcom, which Anthropic intends to lease.

The two firms are sounding out rival investment groups as they prepare one of the largest private credit deals of all time. The loan will be split into multiple tranches, with Broadcom agreeing to step in to make payments to senior lenders if Anthropic misses a payment.

Broadcom’s support means the senior portion of the loan is expected to offer a spread of 1.5-1.75 percentage points above a benchmark compared to a yield of about 8-8.5 per cent on the portion that relies on Anthropic’s ability to meet the chip payments.

Anthropic recently struck a multibillion-dollar deal with Elon Musk’s SpaceX to use one of its data centres, as well as long-term agreements with Google, Broadcom and Amazon potentially totalling hundreds of billions of dollars.

Consumer debt, Investor debt:

Investors are using a record amount of borrowed money to bet on stocks.

Through the end of April, net margin debt hit more than 1.25% of U.S. market cap, near the highest level in records stretching back to 1997. (Axios)

Data: Goldman Sachs Investment Research; Finra; Note: Net margin debt is margin extended to customer securities accounts, after taking account of cash and unused remaining margin credit balances; Chart: Matt Phillips/Axios

Not a timing tool, just a potential fire accelerator.

Speaking of buying on margin: An increasing share of regular folks in South Korea are going all in on the country’s stock market — borrowing money so as not to miss out on the AI-fueled stock boom there.

South Korea’s KOSPI index, which crossed 8,000 for the first time on Tuesday, is up 207% from a year ago. A few big, newly minted trillionaire chipmakers are behind the surge, including SK Hynix and Samsung.

Data: Financial Modeling Prep; Chart: Emily Peck/Axios

Margin loans are at a record high, according to reporting earlier this month in the Korea Times. Middle-aged and older Koreans are increasingly getting in on this — borrowing money so as not to miss out, “mirroring” younger generations but “often with larger sums at stake.”

Some of these folks are risking retirement money with leveraged bets — Korea already has a high poverty rate (40%) for older adults.

• If the AI trade falters, the country’s entire economy is on the line, writes Ed Yardeni at Yardeni Research.
• “Asia’s fourth‑largest economy increasingly resembles a giant leveraged bet on AI.” (Axios)

FYI:

Trump appointees push $250 banknote with his portrait

The director of the Bureau of Engraving and Printing who resisted the effort was reassigned last month. “The buck stopped here,” she wrote in her goodbye email.