Note: I am currently travelling. Hence the more limited postings.
Anthropic Is on Track to Turn a Profit Much Faster Than OpenAI Financial documents from both companies show the different approaches they are taking to the AI boom
Anthropic, which has a growing number of business users because of the capabilities of its Claude chatbot in coding and other arenas, expects to break even for the first time in 2028, the documents show.
By contrast, OpenAI forecasts its operating losses that year to swell to about $74 billion—or roughly three-fourths of revenue—thanks to ballooning spending on computing costs. The ChatGPT-maker also expects to burn through roughly 14 times as much cash as Anthropic before turning a profit in 2030.
The financial road maps, both shared with investors this summer, suggest that the world’s two most-valuable AI startups are taking vastly different approaches to growing their businesses. OpenAI expects thinner margins than Anthropic from its sales for the next five years. Yet it is investing far more in the chips and data centers needed to build its AI technology, and doling out more stock-based compensation to attract top researchers.
The aggressive plan reflects chief executive Sam Altman’s dream of turning OpenAI into a multitrillion-dollar tech giant, his desire to set the pace of the AI boom and his seemingly unending tolerance for risk. The strategy requires near-constant fundraising to keep the startup alive and could backfire if markets cool on the technology or its near-term profitability.
Investors have punished tech companies in recent weeks over concerns about AI spending and whether there will be enough revenue to pay for the extensive build-out in AI infrastructure.
The financial figures for OpenAI came before the startup signed a string of new computing deals with cloud and chip giants—meaning that it is likely set to spend even more in the coming years. Altman said on X that the deals put OpenAI on the hook for up to $1.4 trillion in commitments over the next eight years, leading industry skeptics and some investors to grow doubtful about the startup’s ability to pay for them.
The documents suggest that Anthropic is taking a more cautious approach, with costs growing at a pace more in line with revenue. The company is focused on increasing sales among corporate customers—which account for about 80% of revenue—and is avoiding OpenAI’s costly forays into image and video generation, which require much more computing power. Anthropic’s AI models have also taken off among coders.
Anthropic was founded four years ago by Dario Amodei, a former Google researcher who left OpenAI after a feud with Altman. The startup was caught flat-footed after the viral release of ChatGPT, which gave OpenAI a giant user base and a lengthy head start on sales. Anthropic focused instead on selling its Claude chatbot to businesses, and recently grew its valuation to $183 billion. OpenAI is valued at $500 billion.
Almost every major Silicon Valley investor has a stake in one of the two companies, hoping to cash in on what could be the largest initial public offerings in tech history. The world’s three biggest cloud companies also have tied their growth to the startups. Microsoft is OpenAI’s largest cloud provider, while Amazon and Google are for Anthropic. (…)
The Journal is reporting Anthropic’s base case projections, not its more optimistic forecasts.
The documents show that OpenAI expects to burn $9 billion after generating $13 billion in sales this year, while Anthropic expects to burn almost $3 billion on $4.2 billion in sales—roughly 70% of revenue for both.
Anthropic then becomes a much more efficient business. In 2026, it forecasts dropping its cash burn to roughly one-third of revenue, compared with 57% for OpenAI. Anthropic’s burn rate falls further to 9% in 2027, while it stays the same for OpenAI.
OpenAI’s large upfront investment, particularly for new chips and data centers, could pay off handsomely if demand for its products continues to surge. The company recently launched a new video app called Sora and a web browser named Atlas. It is working on a new consumer hardware device, e-commerce and advertising features for ChatGPT, and humanoid robots.
“Demand for AI exceeds available compute supply today,” an OpenAI spokesman said. “Every dollar we invest in AI infrastructure goes to serving the hundreds of millions of consumers, businesses and developers who rely on ChatGPT to get more done.”
The company is spending almost $100 billion on backup data-center capacity to cover unforeseen demand from future products and research, the documents show. It is setting aside much more computing capacity for new AI research than Anthropic.
“We believe the risk to OpenAI of not having enough computing power is more significant and more likely than the risk of having too much,” Altman recently posted on X.
Last week, OpenAI chief financial officer Sarah Friar said the company had healthy margins and could break even if it wanted to. She highlighted the fast growth of OpenAI’s enterprise business and said the startup was still experimenting with new business models.
Another DeepSeek moment!
This is from The Information yesterday:
Chinese companies are launching open-source AI models built to power coding assistants as cheaper alternatives to those available from OpenAI, Anthropic and Google.
Late last month, Shanghai-based AI startup MiniMax launched M2, a new open-source large language model developed primarily for coding and AI agents. In a post on X, MiniMax noted that the price of M2 is just 8% of that of Anthropic’s Claude Sonnet model. (M2 costs $1.20 per a million output tokens, compared with $15 for Claude Sonnet, according to both companies.)
M2 has already gotten popular. Startups that develop coding agents, such as Cline, Kilo Code and Roo Code, have made M2 available to their users. M2 is currently the fourth most popular model based on token usage on OpenRouter, which offers hundreds of models to developers. M2 trails only xAI, Anthropic and Google on the site.
The low cost isn’t M2’s only selling point. It ranks No. 4 for web development, according to LMArena, which compares AI models on user votes. And it ranks No. 5 in terms of intelligence on another AI model leaderboard, compiled by benchmarking company Artificial Analysis.
MiniMax’s M2 is part of a wave of new Chinese open-source models that are trying to appeal to developers in the U.S. and elsewhere who are looking for decent but more affordable AI models for coding. DeepSeek built a reputation as a provider of strong coding models. More recently other Chinese startups, such as MiniMax, Moonshot AI and Zhipu, have been releasing new AI models that show strong coding capabilities.
They tend to be far cheaper than U.S. models, raising the potential for a price war that could hurt American AI model makers such as Anthropic that owe a significant portion of their revenue growth to powering AI agents.
For the Chinese startups, it is critical they find customers outside China who pay to access the AI models through application programming interfaces, no matter how low the prices are. That’s because it is difficult for AI companies in China to generate revenue from domestic customers where price competition is fierce and business customers are reluctant to pay for subscriptions.
Chinese open-source AI models could face scrutiny from U.S. lawmakers worried about the risk they pose to national security. And many customers are willing to continue spending extra money on more capable American models they consider reliable. But as the overall AI coding market grows rapidly, the Chinese companies are betting that there will be sufficient demand for cheaper and good enough options .
Late last month, Vercel, a U.S. cloud startup that companies use to host and develop websites, said it has partnered with Beijing-based Zhipu to let Vercel users access the startup’s latest model at the lowest price.
Founded by AI researchers at the prestigious Tsinghua University, Zhipu in September released GLM-4.6, a new model built primarily for coding that demonstrated strong performances on various leaderboards
In a post on X, Vercel CEO Guillermo Rauch called GLM-4.6 “an astonishingly good model.”
Beijing-based Moonshot in September released an updated version of Kimi K2 with further enhanced coding capabilities.
Venture capitalist Chamath Palihapitiya last month revealed staff at his VC firm had migrated a large amount of their workload to Moonshot AI’s open-source model, Kimi K2.
“It was really way more performant and, frankly, just a ton cheaper than OpenAI and Anthropic,” Palihapitiya said on the podcast about Kimi K2.
92% cheaper!
What does that do to the biz models showed in the WSJ article?
Investor angst over Big Tech’s AI spending spills into bond market Debt issued by groups building data centres has been hit in recent weeks
A basket of bonds issued by so-called hyperscalers — companies that are building vast data centres, including Alphabet, Meta, Microsoft and Oracle — has sustained a hit in recent weeks.
The spread, or premium in yield investors demand to buy the debt over Treasuries, has climbed to 0.78 percentage points, the highest level since Donald Trump sent markets reeling in April with his tariff plans, and up from 0.5 points in September, according to Bank of America data.
The widening spread highlights how investors are increasingly concerned with the way tech groups are turning to debt markets to finance their investments in AI infrastructure. “The important thing the market woke up to in the past two weeks is that it’s the public markets that are going to need to finance this AI boom,” said Brij Khurana, a fixed income portfolio manager at Wellington Management.
JPMorgan on Monday said building AI infrastructure will cost more than $5tn and “will likely require participation from every public capital market as well as private credit, alternative capital providers and even government involvement”.
The mammoth scale of investment in AI infrastructure has raised concerns about overcapacity, long-term profitability and energy demands. Google, Amazon, Microsoft and Meta will spend more than $400bn on data centres in 2026, on top of more than $350bn this year.
Tech giants are issuing debt at a quick rate to fund their AI expansion efforts despite having large cash hoards, something some investors worry could signal a shift to higher levels of leverage. “The hyperscalers collectively hold [about] $350bn in liquid cash and investments and are expected to generate [roughly] $725bn of operating cash flow in 2026,” JPMorgan said.
“Even so, substantial new debt supply is coming to the credit markets from these high-quality issuers.” In recent weeks, Meta, Alphabet and Oracle have hit markets with blockbuster debt packages, some with maturities as long as 40 years. (…)
Via Reuters:
Bloomberg:
Meanwhile (via The Transcript):
- “And all of a sudden, the past 90 days, we’ve seen definitely an increase in activity and spend. As you’ve seen from the most recent hyperscaler earnings, the forecast for next year has been up, and we’ve seen that. We also are starting to work closer with some of the neo-Cloud and enterprise vendors, and we’re seeing them being more aggressive in wanting to increase their AI CapEx. One of the hyperscalers gave us quite a big increase forecast for ‘26, ‘27 and even into 2028. And we’re seeing our adoption as a companion chip for some of these AI in both the data center and the physical AI accelerate.” Lattice Semiconductor (LSCC) CEO Fouad Tamer
- “Maybe a way to think about it is, back in January of this year, OpenAI with Oracle and SoftBank announced Stargate, which was a $500 billion project to build out data centers over the next number of years. When we go back to where we are now 11 months later, I would say the demand picture for compute is greater than it was at that time. So this is a bit of why you’re seeing all kinds of different accelerated announcements around spend, et cetera, et cetera.” Arm (ARM) CEO Rene Haas
- “People always ask, ‘Is there too many data centers?’ It’s not [a bubble]— because it’s hard to do this. Everyone says they’re going to build something, but it’s not easy to build. You need to find the connectivity, you need to find the power, you need to build the buildings, you have to get the chips, and you need to have $50 billion per AI factory. This can’t be done by everybody… You’re not building enough data centers — first point. Second, the data centers: the shell, the chips inside, they go away in five years. You amortize those in five years. So that will go away, and the chips are going to change — but you’ve amortized them in five years. So we’re not building enough. As much as we’re building today, we’re not even building half of what we can. And you can’t build enough. I actually have the opposite view.” Brookfield (BN) CEO Bruce Flatt
- “Our NVIDIA Blackwell Ultra with GB300 product line now have more than $13 billion in back orders, including the largest deal in our 32-year history, reflecting the tremendous growth potential in hyperscale and enterprise deployments. The B300 platforms are also gaining strong traction following the success of our B200 products as we continue to serve as the leading supplier.” Super Micro Computer (SMCI) CEO Charles Liang
- “What I can share with you at this point is that we are seeing HBM [High Bandwidth Memory] demand grow at a faster pace than supply, and that we have been expanding HBM3E mass production and sales to all of our customers. As a result, in the third quarter, our HBM bit shipments increased by mid-80% Q-on-Q. And excluding some small tail-end sales of legacy HBMs, our overall sales mix is now fully transitioned to HBM3E.” Samsung Electronics (SSNLF)
The Political Left Is Dialing Up Scrutiny of Data Centers Bernie Sanders and others demand White House address high power prices
Sanders, Sen. Richard Blumenthal (D., Conn.) and others on Monday called on the administration to share how it plans to mitigate price impacts and more from data centers needed to train and run AI applications.
In a letter to the White House and Commerce Secretary Howard Lutnick, the group took aim at Meta, OpenAI, Alphabet, Oracle and other firms behind a data-center build-out stretching from the Washington suburbs to rural Oregon. President Trump’s push to fast-track these projects is forcing Americans into “bidding wars with trillion-dollar companies to keep the lights on at home,” they wrote.
The missive is the latest sign of pushback against the AI boom buoying the U.S. economy and markets while straining power producers’ ability to keep up.
Local activists in many areas have begun organizing against data centers that receive sizable tax breaks and require huge amounts of power. At the same time, concerns around increasing electric bills helped fuel Democratic victories last week in New Jersey’s gubernatorial election and down-ballot races in Virginia and Georgia.
“The point is not to stop the data centers but make sure the costs are borne by the gigantic companies that create them, and that electricity prices are checked or even reduced,” Blumenthal said in an interview. “What we need is some federal safeguards and oversight.”
Sens. Chris Van Hollen (D., Md.) and Ed Markey (D., Mass.) cosigned the letter, as did Ron Wyden (D., Ore.), whose state is one of the largest data-center markets in the country.
Trump campaigned last year on cutting energy costs by 50%. While gasoline prices have ticked lower, the Labor Department estimated Americans’ electric bills in September jumped 5.1% from a year earlier, helping to boost inflation above the Federal Reserve’s target.
Price hikes in notoriously opaque power markets can stem from transmission upgrades, volatile commodity costs and expensive fireproofing measures, factors that vary by region. Many analysts warn data centers are adding pressure in areas where development outpaces utilities’ ability to ramp up generation.
The Trump administration has pledged to streamline federal permitting for data centers and accelerate companies’ ability to hook up to electric grids. Officials have focused on powering the build-out through natural gas and coal, slashing subsidies for renewable energy.
Many developers, meanwhile, have increasingly pushed to build out their own power sources to avoid supply-chain snarls and sometimes-onerous regulatory challenges that come with plugging into traditional grids.
Blumenthal said he fears power regulators and local officials “may succumb to the immense dollars that these companies can bring to bear.”
CONSUMER WATCH
Bank of America internal data confirm the inflation squeeze suffered by most Americans:
After-tax wage growth for middle- and lower-income households has not kept pace with sticky 3% inflation, increasing 2% and 1% YoY, respectively, in October.
BofA calculates that “nearly 24% of households so far in 2025 would be classified as living “paycheck to paycheck” based on our central indicator.”
BTW:
Subprime means “bad credit,” not low income. “Bad credit” is a result of being late in paying obligations, or not paying them at all. It does not mean “low income,” though the crisis media constantly equate subprime with low-income.
Low-income people have trouble getting loans, and if they can get loans, balances are relatively small. But the young high-income high-debt dentist into it over his head is a classic example of high-income subprime (he’ll get it cleaned up, but until then, if he gets a new loan, it’s a subprime loan).
The subprime 60-day-plus delinquency rate rose to 6.50% in September, the highest rate for any September, up from 6.12% a year ago, according to Fitch. The seasonal peak was in January at 6.56%, the still reigning all-time high (gold in the chart). (Wolf Richter)
Yet (via The Transcript):
- “The consumer is hanging in there… So, it looks pretty solid right now. The employment numbers that we see through our customer data appear okay… So it’s still a pretty good picture. We were just looking at the sort of deep dive on the October numbers… Spending is robust. And it’s — whilst a little bit stronger at the upper end, it’s still actually hanging in there among lower-end customers. So, that’s been, frankly, given all of the turbulence perhaps a little bit surprising, even with the government shutdown, what we’ve seen is customers that have been impacted by that or drawing down some savings in other places in order to allow themselves to continue to spend to some degree.” The PNC Financial Services Group (PNC) Head of Retail Banking Alexander Overstrom
- “Last thing I’d say is we look very closely at the health of the consumer which, as we’ve described, at least in our book right now, we see a really resilient consumer who continues to spend and grow their spend.” — U.S. Bancorp (USB) Vice Chair, Mark Runke
- “So as we look into those customer bases today, we see the U.S. economy continue to grow, largely driven by the resilient consumer. Consumers remain active. I just got the October data and the money movement to spending by those consumers was up 6% versus last October. Employment remains steady. We can see that in the paychecks coming into our consumer accounts. And we’re getting to see the moderation in the growth of that in our small business customer” Bank of America (BAC) CEO Brian Moyniha
- “I would say, as we’ve commented throughout the year, consumers remain generally healthy and resilient, and we do not yet see any material deterioration in their position… So absent a bigger shock in the labor market, which, of course, is the key to consumer health in our businesses, as of right now, we don’t yet see any material deterioration. In fact, when we are looking at spend trends after a relatively softer, although still pretty solid second quarter, we’ve seen strengthening both in confidence, and that reflected in a little bit of strengthening in spend into the third quarter, and that does continue. We’re still early into the fourth quarter, but we continue to see those trends in the fourth quarter as well. And when you unpack that, it’s really also in discretionary categories, so whether it’s retail or travel and dining.” JPMorgan Chase (JPM) CCB CFO Bori Co
- “In the past few months, almost all range model in production were substantially sold out… Indeed, the order book extends well into 2027.” Ferrari (RACE) CEO Benedetto Vigna
Combining Challenger layoff announcements, WARN notices, initial claims, and earnings call mentions, Goldman Sachs’ layoff tracker
has increased in October and is now higher than before the pandemic. Our quantile regressions based on the level of our layoff tracker, the level of our job growth tracker net of the breakeven pace of job growth, and changes in our slack tracker indicate that the risk of labor market deterioration has increased recently, with the probability of a 0.5pp or higher increase in the unemployment rate over the next six months at 20-25% (vs. 10% six months ago).
Increasing layoffs, rising unemployment, rising fear of unemployment, sticky inflation and slowing wages are all placing an increasing burden on the wealthier stock owners to keep the economy humming.
But so far, so good?
- “Really, it’s just come from the data we see in our own client base in terms of hiring levels. I should add to that in the context, we’re also seeing very low levels of layoffs in the base. So it’s a very static situation… right now, I don’t think it’s a big surprise that hiring is tight. And as such, we’ve just narrowed our expectation within the range that we previously guided towards the low end.” Automatic Data Processing (ADP) CFO Peter Hadley
- “We’re not seeing anything relative to anything of size or meaning with respect to layoffs. The layoffs that have been announced are small relative to those companies and small in general. So there haven’t been any widespread.” Progyny (PGNY) CEO Peter Anevski
For how long?
-
“I think if you think about the low income consumer and you think about the pressures that they face, I mean, right now, you’re seeing across the country, rents are at pretty high levels. You’re seeing food prices, whether it’s in restaurants or grocery, you’re seeing food prices are high, you’re seeing child care is high. There’s just a lot of things that when you think about nondiscretionary spend, there’s some significant inflation there that the low-income consumer is having to absorb. And I think that’s affecting their outlook and their sentiment and their spending behavior, not just in QSR, but across a number of other product categories as well.” McDonald’s (MCD) CEO Christopher Kempczinski
-
“It’s a K-shaped economy without question. You’re seeing significant stability and growth in the high end and some moderate signs of stress in the low end.” Citizens Financial Group (CFG) President Brendan Coughlin
-
“We do see more pressure on the lower-income consumer. We continue to see that in the third quarter, and we expect that to continue into the fourth.” The Wendy’s (WEN) CFO Ken Cook
-
“As the third quarter progressed, we experienced a moderation in trends reflecting broader macroeconomic pressures. Entering the fourth quarter, we’re seeing further moderation as we continue to lap stronger same restaurant sales from the prior year.” CAVA Group (CAVA) CFO Tricia Tolivar
-
Consumers are now expressing worries about potential negative consequences for the economy. This month’s decline in sentiment was widespread throughout the population, seen across age, income, and political affiliation. One key exception: consumers with the largest tercile of stock holdings posted a notable 11% increase in sentiment, supported by continued strength in stock markets. (University of Michigan Survey)
Why Donald Trump is in trouble
(…) The Economist’s data team has delved deep into the results to see whether they are a sign of bigger trouble for Donald Trump and the Republicans.
The most closely watched contests were the governors’ races in Virginia and New Jersey, where centrist Democrats who campaigned on affordability won by bigger margins than expected. Some of that can be explained by turnout. In exit polls from the Virginia race, voters were asked whom they had supported in the 2024 presidential election. Of those who had voted, a larger proportion said Kamala Harris than her actual statewide vote share—suggesting that more of Mr Trump’s supporters decided to stay at home. Exit polls from New Jersey tell a similar story. Yet turnout alone cannot explain the nine-point swing in Virginia and eight-points in New Jersey. Instead, our analysis suggests that Democratic candidates persuaded Mr Trump’s voters to switch sides.
Local election results show where the biggest swings occurred. Passaic and Hudson counties in New Jersey, which last year turned against the Democrats by 19 and 18 points respectively, recorded the biggest swings in the state towards Mikie Sherrill, the new Democratic governor-elect. Both counties have large Hispanic populations, a group that Mr Trump wooed successfully in 2024. Across the state, counties that had moved away from Democrats appeared to snap back.
Other minority communities also tilted back towards the Democrats. In one ward in South Paterson, which has the second-largest Arab population in the country (and home to a neighbourhood known as Little Palestine), voters shifted 56 points to the left. Ms Sherrill won more votes there than Ms Harris, even as turnout fell by 37% from last November. In Edison, where nearly half of the population is Asian, the shift was 23 points to the left.
Our analysis of Virginia points to the same pattern. In precincts with large Asian or Hispanic populations, Ms Spanberger outperformed Ms Harris by wide margins. National polling data confirm that non-white voters are souring on Mr Trump faster than white ones are.
Younger voters, too, appear to be drifting back. In precincts where few residents are over 60, Ms Spanberger’s advantage over her Republican rival was almost 16 points larger than Ms Harris’s (in older areas it was four points). Young voters are less partisan and cite the cost of living as their main worry, an issue that helped Mr Trump last year but now hinders him.
Two other states offered further evidence of the changing tide. Pennsylvanians were voting for judges in their highest courts; Georgians for members of the state’s public-utilities regulator. Turnout was predictably lower than in the presidential election.
But both delivered landslides for Democrats, sharply improving on their 2024 margins. In Pennsylvania the party fended off a multimillion-dollar campaign to tilt the state’s courts rightward. In Georgia Democrats captured two seats on the utilities board, their first non-federal statewide wins in nearly 20 years amid growing anger over soaring energy bills.
Democrats also defeated Republicans in a string of local races and ballot initiatives, including a redistricting effort in California. In deep-blue New York City, Zohran Mamdani, a Democratic Socialist, easily won the mayoral race (although he faced uninspiring opponents, and lost in parts of the city that look most like the rest of America).
Attention now turns to next year’s midterm elections, which decide control of Congress. Much could change before then, but our current polling with YouGov shows that Democrats lead the generic congressional ballot by three points. Such a margin could yield a narrow majority in the House of Representatives. But if the recovery among minorities and younger voters endures, some Republican gerrymanders may prove less secure than intended. In Texas, for example, Republican mapmakers counted on Mr Trump’s strong margins among Hispanic voters in 2024—support that now looks shakier.
The Senate is tougher. To capture a majority Democrats would need wins in states such as Alaska, Ohio and Texas—all still long shots. Success has often bred complacency in the party. But if they learn the right lessons from November 4th, they will be better prepared for the battles to come.
Voters are 10 percentage points less likely to approve than disapprove of Trump’s job performance (44% to 54%), setting a record low since he took office in January. Similarly, his net favorability rating is 12 points underwater, the lowest we’ve measured in more than a year since before the July 2024 attempt on his life.
Perceptions of Trump’s handling of health care, the economy, Medicare and Social Security, the national debt and energy set new record lows over the weekend.
The popularity of the GOP’s brand on Capitol Hill has soured over the past month, with our latest survey showing the net favorability rating for Republicans in Congress 12 points in the red — their worst showing in 2025 — compared to a 7-point deficit at the start of the shutdown. Democrats, on the other hand, have largely made up for a popularity hit they sustained after the shutdown’s first week.
Trump’s reaction: stimulus checks.
Trump just announced the “tariff dividend,” a payment of AT LEAST $2,000 per American. We expect 85%+ of US adults to receive this, resulting in $400+ BILLION handed out. All as US debt nears $40 trillion. What’s next? Let us explain.
@KobeissiLetter
Trump Defends Foreign Students as ‘Good’ for US Universities
President Donald Trump defended allowing foreign students to study in the US as a “good” practice and pushed back on a call to reduce their numbers, saying it would be financially destructive to the nation’s higher education system.
“You don’t want to cut half of the people, half of the students from all over the world that are coming into our country — destroy our entire university and college system — I don’t want to do that,” Trump said in an interview with Laura Ingraham that aired Monday on Fox News.
“I actually think it’s good to have outside countries. Look, I want to be able to get along with the world,” he added.
Trump was repeatedly pushed by the Fox News host on why he wouldn’t reduce the number of students from foreign countries — in particular China — who are studying in the US, with Ingraham casting it as a policy the president’s supporters would back and a change that would make it easier for Americans to enroll in schools.
Trump argued that reducing the numbers of foreign students would cause financial harm to the university system and lead some schools, including historically Black institutions, to go “out of business.” (…)
“We take in trillions of dollars from students. You know, the students pay more than double when they come in from most foreign countries. I want to see our school system thrive,” he added. “It’s not that I want them, but I view it as a business.”
Trump’s comments offer a contrast with policies his administration has carried out that have targeted international students. The administration has revoked thousands of visas, arrested students who were involved in pro-Palestinian activities and imposed heightened application requirements.
The administration has also targeted top-tier institutions over their handling of international student applicants and compliance with visa regulations. Harvard University challenged the administration’s attempt to block the school from admitting foreign students. A judge blocked the administration from enforcing the ban, but the US is appealing that decision.
Trump earlier this year also announced that he would begin revoking the visas of Chinese students, a move intended to obtain leverage in trade talks with Beijing, but later retreated as part of a broader truce on tariffs and export controls between the two countries.
Officials have also moved to reduce the number of foreign students as part of a proposed compact with colleges that would give them preferential access to federal funds. The effort, dubbed the “Compact for Academic Excellence in Higher Education,” would include a cap on international students that limits undergraduate numbers to no more than 15% of students on foreign visas, with no more than 5% of students coming from any one country.
Earlier this year, Secretary of State Marco Rubio ordered US embassies worldwide to stop scheduling interviews for student visas, with the administration later resuming them but with more strict vetting of applicants’ social-media profiles.
China Hatches Plan to Keep U.S. Military From Getting Its Rare-Earth Magnets Beijing considers ‘validated end-user’ system to fast-track certain export licenses
China plans to ease the flow of rare earths and other restricted materials to the U.S. by designing a system that will exclude companies with ties to the U.S. military while fast-tracking export approvals for other firms, according to people familiar with the plan.
The “validated end-user” system, or VEU, would enable Chinese leader Xi Jinping to follow through on a pledge to President Trump to facilitate the export of such materials while ensuring that they don’t end up with U.S. military suppliers, a core concern for China, according to the people familiar with the plan.
If strictly implemented, the system could make importing certain Chinese materials more difficult for automotive and aerospace companies that have both civilian and defense clients. Beijing’s plan could still change and its licensing system wouldn’t be certain until it is implemented, the people said. (…)
The VEU mechanism that Beijing is considering is modeled on U.S. laws and procedures, as is much of Beijing’s export-control architecture.
Under the American version of the VEU system, which has been active since 2007, certain Chinese companies are cleared to buy sensitive goods under a general authorization—essentially a simplified export-approval mechanism—instead of needing individual licenses for each purchase.
This makes it easier to import controlled goods such as chemicals or chip-making equipment, but requires companies to put up with U.S. government inspections of their facilities, among other steps, to verify compliance with the program. (…)
While Trump has said that the issuance of general licenses by Beijing will mark the de facto end of export curbs on critical materials, Beijing appears to be keeping some of the controls in place, even as it tries to ease exports for proven civilian uses.
Beijing hasn’t made clear which companies will be eligible for general licenses, or exactly what benefits such licenses would provide.
It isn’t clear how long VEU protections would last. Under the U.S. system, Chinese companies that are cleared to receive purchases under the VEU sometimes have this clearance taken away, which has caused consternation in Beijing. (…)
Hmmm…somebody will get angry…