Chinese models have gone from 1.2% of token traffic (a common way to calculate AI usage) in 2024 to more than half of the total by the summer of 2026.

 But the biggest threat to that next wave is not technology. It’s money. Between 2023 and 2026, venture funding into U.S. AI companies topped $380 billion; China’s start-ups received barely a tenth of that figure, according to Boston Consulting Group.

Funding the future

In the past, Chinese entrepreneurs looked to state guidance funds and venture capital backing, but policy-driven funds are known to prioritize later-stage startups while early-stage venture capital is only just recovering from a three-year fundraising drought.

There are three reasons why broader funding channels should be a priority for entrepreneurs in the AI economy.

First, inflation is taking hold inside the AI economy. CXMT, for example, has been raising memory prices for months and held firm even when Huawei, one of its largest customers, demanded relief.  

The war for AI talent is just as fierce: postings for AI-related roles surged roughly twelvefold year-on-year in early 2026. Algorithm engineers specializing in large language models command some of the highest pay packages of any technical role in China.

Founders must also outcompete deep-pocketed former employers and U.S. rivals. More than half of studies presented at the world’s top AI conference had lead authors based in China. China’s AI talent is known to be in demand worldwide.

Second, external funding is scarce. Venture investment in China totaled just $20 billion in the first quarter of 2026, against $267 billion in the U.S. China has raised impressive sums this year—assets under management for newly registered VC funds hit 154 billion yuan ($22.8 billion) in the first five monthes, already exceeding last year’s total—but that is still far below what US venture capital regularly deploys.

Meanwhile, China’s state banks, although directed to prioritize technology lending, are absorbing rising non-performing loans elsewhere on their books, which could lead to weaker overall credit supply.

Third, profitability will take a while to achieve. Chinese enterprise software firms primarily sell into the domestic market, which limits their revenue base. U.S. rivals have a head start: a global customer base, stronger brand recognition, and R&D budgets deep enough to fund everything from enterprise-grade cybersecurity to polished customer experience design.

To be sure, the next generation of AI ventures may not need vast amounts of capital to build applications on top of existing models or fill the gaps in the tech value chain.  Enterprise customers can also provide essential development funding.

Entrepreneurs must nevertheless cope without the kind of financial support that is available for the AI sector in the U.S. Closing the performance gap increasingly depends on bulking up in-house computing capacity—yet China’s AI infrastructure spending remains a fraction of what is being spent in the U.S.

Private market alternatives

This is where access to Hong Kong’s finance industry will play a growing role. Its capital markets remain one of the few channels still capable of moving global capital toward Chinese enterprise at scale. More than 430 applicants are in the IPO pipeline in the second half of 2026.

The IPO pipeline, however, tells its own story. Many Chinese technology startups are choosing to list earlier than the previous generation of companies did because they lack an alternative.

Compare that with the U.S., where the likes of OpenAI and Anthropic have grown to an enormous scale by raising private capital. OpenAI closed a round of more than $100 billion earlier this year, while Anthropic raised $65 billion in May. China’s leading model developers, Zhipu AI and MiniMax, beat OpenAI and Anthropic to the public markets—but their Hong Kong IPOs in January raised just $558 million and $620 million respectively, despite heavy over-subscription.

Private credit is another route. Asia-Pacific private credit assets are projected to grow from $59 billion in 2024 to $92 billion by 2027, with China accounting for a fifth of the region’s activity. However, these loan providers tend to prioritize bigger or established companies.

It is already clear that the AI sector, unlike the software sector, will not be dominated by U.S. firms alone. China’s startups have learned the lessons of the previous decade, commanding global respect and winning a growing customer base. The country’s open-weight strategy has given its leading AI companies a cost advantage – even Silicon Valley leaders acknowledge it.

For China’s latest generation of AI entrepreneurs, venture capital and bank loans—the two main sources of start-up funding – may not be enough in the coming years. Keeping pace in the AI race will require using every asset and every channel available. It may mean a public listing earlier than they would have preferred, exploring the growing private credit ecosystem, revenue sharing with customers or leveraging the equity they hold as collateral.

The entrepreneurs involved in the next wave of China’s AI development will need to be creative with funding to maintain their competitive edge at a time when overseas rivals are spending 10 times as much.

The opinions expressed in Fortune.com commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of Fortune.

This story was originally featured on Fortune.com

Well, nearly 60 years after the publication of E.L. Konigsburg’s book about two kids running toward the city and away from their suburb-living parents, there are increasing signs of the complete opposite occurring: Parents are taking their kids out of the New York City school system, and out of New York City entirely, and heading for the burbs, thanks to the increased cost of raising a child here. Families with young children are leaving New York City at more than twice the rate of households without them, and the public school system is shrinking as a result.

A decade ago, the city had over 1 million public school students, but this year it has 885,000, City Comptroller Mark Levine said in a social media post on X—and he expects another 150,000 fewer over the next decade. The drop has left 219 schools with fewer than 200 students each, he said, some too small to field a sports team. “This is not sustainable,” he wrote in the post, and the city will have to consolidate schools with sharply declining enrollment, Gothamist reported.

Levine said in a statement to Fortune that cost is part of the story. “Raising a family in New York City is increasingly difficult as the cost of living rises, and declining student enrollment is one troubling sign of that challenge,” he said. “Affordability is a major factor, but not the only one, and we need to better understand every driver of this trend to respond with viable, data-driven solutions.”

“Families with young kids are twice as likely to move out of the city than the rest of the population,” Emily Eisner, executive director of the Fiscal Policy Institute, told Fortune. “That’s because it’s really expensive to raise kids in the city these days.” Housing adds to the pressure as children get older. “There’s basically no options for families to find an affordable housing situation,” she said.

The cost of raising a kid

An FPI report from June 2024 found that New York City households with a child under 6 were 112% more likely to leave than those without. They made up 30% of out-migrants but 14% of the population. Among New Yorkers who left the state after the pandemic, 36% said they wanted better or more affordable housing, up from 16% before it. The report calls the outflow a policy failure and says it raises long-run fiscal and economic concerns.

It’s not just the Big Apple: The cost of raising a child has climbed nationwide. LendingTree puts the 18-year total at more than $300,000, with childcare the largest expense for families with children under 5. Care for an infant and a 4-year-old averages $28,190 a year nationwide, which would take a household income of $402,708 to stay within the federal 7% affordability benchmark.

For comparison, the average two-child household earns $145,656. Parents spend about 20% of their income on childcare, according to Care.com, and one in five spend more than $30,000 a year. In a BMO survey, 82% of parents said the cost of raising children has “gotten out of control.”

Housing leaves less room still. A buyer needs at least $211,970 in household income to afford a median-price home in New York City, nearly three times the national figure, according to StreetEasy, though that is down $10,625 from a year earlier. Meanwhile, only 15% of city households earn $200,000 or more.

Declining birth rates

Also compounding the strain is the declining birth rate, as U.S. fertility fell to a record low in 2024, to below 1.6 children per woman. Births in 2025 ran about 24,000 below the 2024 total. In New York, the city had about 110,000 births in 2019 and fewer than 100,000 a year since 2020.

A new FPI report shows the people leaving New York are not the wealthiest, Fortune reported last month. In typical years, FPI found, the top 1% of earners leave at one-quarter the rate of everyone else. But of middle-income families, Eisner said, “I don’t think they migrate because of taxes. They migrate because of the high cost of living, which really is about childcare costs and housing costs.”

Cost is not the only explanation. A city Education Department survey found that 40% of families who left the school system moved out of the city, citing a better environment to raise kids, concerns about schools, and more housing space.

Eisner said pushing for universal free childcare and a larger housing supply—referencing policies such as Zohran Mamdani’s pledge during his mayoral campaign to make childcare free for children from 6 weeks to 5 years old—will ease that decline in children in the city: “Those will play a significant role in bringing the cost of living down.”

After all, the kids need a reason to map out their hideout spots in the Met.

This story was originally featured on Fortune.com

The company unveiled a variety of updates to Windows on Wednesday, including incorporating AI agents and other AI tools directly into the PC software, as Microsoft seeks to ensure that its products remain at the center of a rapidly changing tech industry.

Speaking at San Francisco’s Dogpatch Studios, chief executive Satya Nadella said Windows is becoming a “hybrid intelligence” system, meaning there will be capabilities for cloud-based AI, as well as AI that runs locally through products like the coding tool GitHub.

“Just having the [AI] model doesn’t do much for anything,” Nadella said, explaining that Microsoft is bringing together local AI, AI-powered software, and devices built natively for AI, such as the new Surface Laptop Ultra, which the company unveiled at the event. “Some of the things you’ve been using all click together.”

Microsoft was one of the first Big Tech companies to recognize the importance of generative AI, investing $13 billion in OpenAI and partnering with the company years ago. But the alliance between the two companies weakened, and Microsoft is now trying to find its lane in a crowded AI landscape as competitors like Google, OpenAI, and Anthropic take the lead developing the most advanced large language models, and Meta carves out a space with its Muse personal AI assistant.

A big part of Microsoft’s plans involves integrating the reimagined Copilot AI app seamlessly into Windows. This means having Copilot run tasks without having to access the cloud when those tasks don’t need heavy compute, according to Jacob Andreou, the head of Copilot. A user, for example, will be able to have Copilot autonomously pull all of the files they need to file taxes from various parts of their laptop and do so without accessing the cloud, making it cheaper and faster.

“Copilot now truly understands your PC,” he said. “It’s Copilot built natively for Windows.”

While Microsoft operates one of the industry’s largest cloud computing services, the company is betting that having a strong presence throughout the “tech stack” could give it an edge as AI transforms the way businesses operate. Local AI has become a popular way for coders and other users to experience the technology because it can be cheaper and safer than running AI on the cloud.

Microsoft is also tying its updates to its new Surface Laptop Ultra, a high-end device built for developers and creators. A user on the Ultra, the company showed, can play video games while editing a video and running AI workloads. Nadella was joined onstage by Nvidia chief executive Jensen Huang. The new Surface Laptop Ultra uses Nvidia’s RTX Spark “superchip,” which Huang called “a beast” of a product.

The Ultra will cost $2,599, while the Surface RTX Spark Dev Box, which works alongside the Ultra, will cost $5,999. The laptop will be available beginning Oct. 16, and the Dev Box in November.

Nadella has recently refocused Microsoft on the basics after initially also pursuing general consumer AI with Copilot. Microsoft aims to own the enterprise side of AI workflows and cloud computing. Wednesday’s event was an extension of presentations it made at its June Build conference. During the conference, it introduced an AI wearable badge-like device and a cube-like device with a voice-activated screen, though it hasn’t yet provided an update.

Microsoft is reorienting its Copilot AI brand and its leading software for the AI era. It is also positioning its devices as the vehicle best suited to optimize customers’ workflows, without bogging them down with numerous Copilots. Microsoft received many initial user complaints about Copilot’s nonstop presence within the Windows OS.

The company’s redesigned Copilot app will now be a central interface through which customers can chat with AI bots, tap AI tools to help accomplish tasks, write code, and access a new Autopilot feature that brings agentic capabilities. Some of the features Microsoft has shown off, however, are not yet available.

An Autopilot feature on the app is designed to let users name their agent, assign it a role and a goal, and then set it loose on tasks such as project updates, tracking work decisions and deadlines, reminding colleagues when commitments are due, gathering prep materials, and reaching out to partners.

Microsoft is betting that this new “super app” will boost Copilot by connecting AI directly to its already lucrative business tools. As OpenAI and Anthropic push into enterprise offerings, Microsoft is racing to protect its turf and keep pace with customer expectations.

In July, the company said its Microsoft 365 Copilot had surpassed 30 million paid seats, with net seats added doubling quarter over quarter. Its shares have rebounded over the past six months on the strength of its cloud computing business, fueled by AI demand and Copilot growth. Its next quarterly earnings report is expected later this month.

This story was originally featured on Fortune.com

The volume of results stunned many mathematicians, while the way OpenAI has gone about tackling the problems and publishing the solutions divided the field. Some said they were enthusiastic about the results, seeing huge new areas for mathematicians to explore. Others said the approach OpenAI and other AI companies have taken to solving mathematical problems constitutes an assault on mathematics as a human academic discipline.

OpenAI said it achieved the results using an unreleased internal AI model. It said that on average the model took about three hours of computing time to arrive at each solution.

The massive cache of new solutions includes full or partial results for many of the problems mathematicians have considered the most important to the field. The results come weeks after OpenAI said it had used an unreleased internal model to solve the Navier-Stokes equations, one of the seven Millennium Prize problems for which the Clay Mathematics Institute offers a $1 million award. In the most recent batch of results, OpenAI said it had made progress on three other Millennium Prize problems but had not fully solved them.

AI companies have been targeting mathematical problems as a way of showcasing the capabilities of their models. AI researchers have also said that training their AI models on difficult math problems may help them learn many skills that generalize to other domains in the real world. For instance, it may help teach the models logical reasoning skills as well as how to be persistent in the face of difficult problems. It may also teach the models to do well in domains such as physics or economics that involve a lot of mathematics—although so far, it is unclear exactly how a model’s mathematical capabilities may generalize to domains, such as law or business strategy, which involve logical reasoning, but do not have objectively verifiable correct solutions.

Meanwhile, some of the traits learned in tackling very difficult mathematical problems—such as persistence—may increase safety risks. In recent “rogue AI” incidents, AI agents went to extreme lengths to achieve results in an evaluation, including taking unauthorized and illegal actions. Faced with a seemingly impossible challenge, a human might simply give up rather than resort to these kinds of unauthorized steps.

Dan Litt, a professor of mathematics at the University of Toronto, told Fortune he was excited about OpenAI’s results. “My view is that this is great for mathematics,” he said, adding that there were several solutions OpenAI published that impacted problems he was interested in and that he was eager to understand the solutions OpenAI’s model found: “I think that it’s great to have new solutions to questions that I and others are interested in.”

Litt cautioned, however, that he is worried about the effect the solutions may have on the field of mathematics, especially if a perception that AI has “solved math” leads funding organizations to withdraw support for mathematical research or discourages promising young mathematicians from entering the profession: “It’s important that society reaffirms support for human mathematical expertise if we want to get anything out of the progress on these problems that AI has made.”

Showing the work

When OpenAI published its Navier-Stokes solution, two mathematicians, who had also been working on a solution to the problem using AI tools, including OpenAI’s, accused the company of either intentionally or inadvertently feeding their work in progress to its AI model, helping point it in the direction of the solution. OpenAI denied this was the case, saying it did not feed its model the two mathematicians’ work and that the model could not have picked up any clues about their research from its training data because the cutoff for that data preceded the date on which the two mathematicians had begun using OpenAI’s Codex AI product to work on Navier-Stokes.

In response to the latest results, Tristan Buckmaster at New York University, one of the mathematicians involved in the earlier controversy, told the New York Times that it remained unclear whether mathematicians using OpenAI’s models had inadvertently helped point the company’s internal AI system toward the solutions it found. “There’s likely to be a bunch of results where they take someone’s work and then take it to completion,” he told the Times. Given the number of results being released simultaneously, he said, “I don’t think they’ve done their sort of due diligence at all” to ensure the AI model had not plagiarized anyone’s work.

Last month, following criticism from mathematicians in the wake of its Navier-Stokes solution, OpenAI said it was forming an independent advisory group on mathematics and artificial intelligence hosted at the Institute for Advanced Study in Princeton, N.J. 

Late last month, the group released a set of recommendations for the publication of AI-generated mathematical proofs. The recommendations included that AI-generated proofs should be published following the conventions of a traditional mathematical research paper, so that human mathematicians could more easily scrutinize and learn from the results. It also recommended that for each solution, an AI company should make public the name of the model used, the prompts used, the model’s “chain of thought” (or an output of its reasoning steps), the time it took the model to arrive at the solution, and an approximation of how much that computing time cost. It said that the company should also disclose how it decided to have the AI try to solve that particular problem and, if many results were published at once, that the company should publish a report detailing why those problems were targeted and how many other problems of comparable difficulty the model tried and failed to solve.

OpenAI published the latest mathematical solutions to GitHub, the code repository site. It followed some, but not all, of the steps the advisory group had recommended. The group published a statement on Tuesday saying: “We reaffirm our published recommendations on responsible release.” It said its discussions with OpenAI had been “constructive” but that “ultimately it is up to the mathematical community to assess the extent to which our recommendations were followed successfully, and whether there are others we should suggest.”

The company released a blog post on Tuesday in which it said it had “drawn on” the advisory group’s advice about how to publish the solutions. “For future releases, we are committed to further improving the quality of the papers via the citations, mathematical exposition, and presentation of the results for better understanding,” OpenAI said. It said it was sharing formalizations of the proofs for many of the problems—these are versions of the proof that can be verified by specialized computer software—and would share more of these as it obtained them. It also said that for 10 problems it was publishing summaries of its model’s reasoning, estimates of the compute spent, and statistics about the number of attempted problems.

Litt, who was not a member of the advisory group, told Fortune he approved of most aspects of how OpenAI published the solutions. Having them on GitHub made them easily accessible for other mathematicians to study, he said, and he credited the company for not making too much of any particular advance in a blog post or marketing material intended for a nontechnical audience. He also said he thought OpenAI lacked the capability to publish all the results in research papers that would meet rigorous academic standards, both because the AI models don’t write mathematical exposition well enough and struggle to cite prior mathematical work, and because OpenAI doesn’t employ enough mathematicians with expertise in enough areas to understand all the proofs the AI models can generate.

While some mathematicians have complained that AI-generated proofs, such as OpenAI’s Navier-Stokes solutions, are difficult to follow, making it hard for mathematicians to build on the results, Litt said he thought such concerns were “overstated.” He said mathematical writing was often difficult to follow anyway. “I think to extract understanding from [the OpenAI results] there will be a huge amount of human labor involved, but it’s not so different from the labor that mathematicians have been doing forever,” he said.

OpenAI said it wanted its solutions “to push the frontier of human knowledge and enable further progress in mathematics.” It said it would be funding a series of workshops, conferences, and programs focused on helping mathematicians understand the results its AI system had generated.

The end of ‘Math 1.0’

The independent math advisory group said in its statement on OpenAI’s release that “the future of mathematical research cannot consist only of understanding results produced by AI labs. Mathematicians must be able to formulate their own questions, develop their own approaches, and explore directions that have not been selected as examples of an AI system’s capabilities. Equitable access to powerful research tools and adequate computational resources are essential to that freedom.”

Terence Tao, a UCLA mathematics professor considered one of the world’s greatest living mathematicians, has been increasingly critical of the way AI companies have gone after mathematical problems, arguing that it is the process of arriving at solutions—not so much the solutions themselves—that advances mathematical understanding, and that by solving so many interesting problems so quickly, AI companies are discouraging students from becoming mathematicians, robbing the field of its future.

In a social media post on Mastodon Tuesday, Tao reiterated these criticisms. “Problems are being solved autonomously by AI prompters who have no interest in the broader field itself once their initial target is ‘solved,’ and do not understand the AI output well enough to answer questions on the result, give talks, or otherwise interact with the rest of the field,” he wrote. “Many fewer seminars, workshops, collaborations, or other activities are being generated from these results compared to traditional breakthroughs; few people are joining the community around the field as a consequence; and promising open directions are now being withheld from the public in fear that this will cause their own research to be ‘scooped.’”

Tao said that OpenAI’s mass publication of math solutions marked the end of “Math 1.0,” in which finding solutions to unsolved conjectures and problems, even if those solutions could not easily be understood at first, served as the field’s engine. He said there would now need to be a “Math 2.0” era that “will need to decenter the role of raw problem-solving and value mathematical progress more holistically—for instance by elevating the role of exposition, but also that of community building and opening up new directions of study.”

Litt said he agreed with Tao that the field must change. He said OpenAI’s publication of such a massive set of solutions would help get the entire field “on the same page and understanding that we need to be a little bit radical about rethinking” things such as what kinds of contributions it rewards and how it trains PhD students. And while Tao has generally sounded wistful about this transition, Litt said he was “optimistic” about it.

“One of my collaborators told me, ‘I feel like I’ve been crawling my entire life, and now I can fly,’” Litt said of the advent of AI as a tool for solving mathematical problems. “It’s incredible what we can do now.” He said he thought AI would enable human mathematicians to engage in much more “open-ended exploration” than was possible before: “We should expect mathematicians to be way more productive in the future.”

This story was originally featured on Fortune.com

“They’re a national emergency that threatens our security and our very way of life,” he said at the time. “It’s a very great threat to our country.”

But 17 months later, the administration’s import taxes haven’t closed that deficit at all. In fact, the trade deficit is now the largest it’s been since Trump implemented tariffs, and economists are pouring cold water on the very idea that a widening trade gap is even a threat.

The trade deficit grew by 13.7% to $105.6 billion from July to August, according to the Bureau of Economic Analysis, the highest it’s been since March 2025’s $140 billion. While imports increased by 4.3% to $420.8 billion, exports increased, but at a slower rate, growing 2.2% to $205.7 billion.

Experts suggest the growing deficit is largely a result of the AI boom, which has created a mass demand in the U.S. for overseas hardware, spiking imports. AI demand single-handedly added $200 billion to the U.S. trade deficit in April, according to the Federal Reserve Bank of Minneapolis.

“The U.S. economy does a bunch of stuff that other countries can’t do; one of them is … this AI boom,” Tarek Hassan, a professor of economics at Boston University, told Fortune. “And because of the AI boom, foreigners still want to invest in the United States, which is a good thing.” 

Tariffs have been the cornerstone of Trump’s second administration, and the president has justified the levies with promises of more U.S. manufacturing jobs, as well as a cash windfall to be redistributed to the American people—neither of which has come to fruition. 

To Trump, the trade deficit was representative of other countries taking advantage of the U.S., or of America’s wealth leaving the country, and protectionist measures were the only way to restore U.S. trade dominance. 

“Capital goods imports remain at a record high, more evidence that the President Trump’s economic policies have led to a manufacturing renaissance.” White House spokesperson Taylor Rogers told Fortune in a statement. “The president’s push to reindustrialize the United States continues to pay off as trillions of dollars in manufacturing investments materialize.”

But economists see things differently. They argue Trump’s whipsaw tariff approach that was meant to slash the deficit was actually, in part, responsible for its widening. They also say they believe Trump’s insistence on the gap being a negative economic indicator is actually ignoring evidence of a resilient U.S. economy.

Why Trump’s tariffs backfired

The ineffectiveness of Trump’s tariffs in closing the trade deficit have less to do with the magnitude of the duties and more to do with their ever-fluctuating nature. For example, tariffs on China spiked at 145% following Liberation Day, but have since fallen to around 30% following the Supreme Court ruling against tariffs imposed under the International Emergency Economic Powers Act (IEEPA) and a series of trade agreements with the country. While companies often look to make supply-chain adjustments in the face of tariffs, that really hasn’t happened because the trade landscape is too uncertain to make long-term adjustments. As a result, American businesses are still heavily reliant on overseas goods.

“There’s so much volatility that, essentially, companies are not going to change their behavior,” Hassan said. “They’re just going to adjust prices and continue on, and that’s why this tariff regime has been somewhat ineffective.”

Donald Boudreaux, a professor of economics at George Mason University, explained that the AI boom is not just fueling demand for overseas intermediate goods like hardware parts. It has also created economic optimism among U.S. companies of future productivity increases, driving importers to continue buying tariffed goods, despite higher costs.

“American businesses must be looking at the U.S. economy and their expectations about the future of the U.S. economy, and they’re optimistic,” Boudreaux told Fortune. “They’re saying, ‘Even though we have to pay higher prices for steel, higher prices for machine parts, higher prices for whatever it is we are importing from abroad, those higher prices are not high enough to dissuade us from providing these inputs that we believe we’ll need to meet future demands for our outputs.’”

Be careful what you wish for

Ultimately, the continued investment from other countries in the U.S. is what makes the trade deficit a good thing in Boudreaux’s eyes. While economists like Peter Navarro believe lowering the deficit would prevent the hollowing out of American manufacturing, Boudreaux argued that by receiving more goods from foreigners, dollars typically come back to the U.S. via global investments in U.S. equity, buoying the American economy.

“President Trump thinks, on his terms and the administration’s terms, this is unambiguously bad news,” Boudreaux said. “But in my view, it’s, I would say, unambiguously good news. It points overwhelmingly to health in the American economy, and not to any problems in the American economy.” 

If Trump really wants to close the trade deficit, Hassan said, he should be careful what he wishes for. As foreign investors continue to pour dollars back into the U.S., including through Treasuries, they are essentially extending cheap credit to Americans. 

But should the trade deficit close and exports outnumber imports, foreign investment would also taper, Hassan suggested. Bond yields would increase, and financing the ballooning national debt would become even harder.

“It’s going to be a crisis where the trade deficit is going to close because foreigners are unwilling to extend further credit to the U.S.,” Hassan said. “Complaining about the trade deficit is like complaining that the bank keeps giving us all this cheap credit.”

This story was originally featured on Fortune.com

Moments later, Trump signed an executive order onstage that temporarily allows red-dyed diesel on public roads and defers the federal excise tax normally owed on highway fuel through Dec. 31. Despite Trump calling the fuel “tax-free,” the tax hasn’t gone away—at least not yet. The order directs the Treasury Department to defer payment without interest or penalties, and to explore ways to eliminate the deferred bill entirely.

The move comes as diesel prices hover near record highs, squeezing truckers and farmers in the middle of harvest season. It also follows the administration’s retreat from a proposed ban on U.S. diesel exports, which Trump said last week, “We were never going to do.”

What is red-dyed diesel?

Alan Krupnick, a senior fellow at Resources for the Future and director of its Industry and Fuels Program, explained the fuel to Fortune: “It’s the same as regular diesel, but dyed red so inspectors can make sure it’s being sold only for off-road uses, [such as] for agricultural vehicles.”

What’s different about red-dyed and regular diesel, then, aside from the color? “It is [sold] tax-free to give agriculture and other heavy, off-road vehicles and other users a break,” Krupnick said.

Gilbert Metcalf, a visiting professor at the MIT Sloan School of Management and a former deputy assistant secretary for environment and energy at the U.S. Treasury Department, described it the same way. “This is the same diesel that you can buy at a gas station; it just has red dye in it. That’s to indicate that it’s not subject to the federal diesel tax.”

How much could drivers actually save?

That federal tax is 24.4 cents per gallon: a 24.3-cent excise tax plus a 0.1-cent “Leaking Underground Storage Tank” fee, according to the U.S. Energy Information Administration. On a 250-gallon fill-up for an 18-wheeler, that works out to about $61. The White House’s claim that truckers will save more than $100 per fill-up assumes states suspend their own diesel taxes as well, which the order encourages but can’t require.

Measured against the pump price, the federal tax is a sliver. Diesel averaged about $6.20 a gallon nationally the week of Oct. 5, per the EIA, down from a peak of roughly $6.53 in late September. Before the U.S. and Israel launched their war against Iran in late February, the national average was about $3.76, according to AAA. A 24.4-cent break amounts to about 4% of today’s price—a modest savings for everyday motorists, and perhaps not much more for heavier users in trucking, farming, and other industries.

Why economists say it won’t move prices

The bigger problem is that a small cut in per-gallon costs does nothing to change the basics of supply and demand. Metcalf sees little potential for the executive order to have any meaningful impact.

“I don’t see this having really any impact on supply … On the supply side, it’s not going to lead to any more diesel coming online, all it does is change some of the diesel that’s out there [to be] usable by anyone,” Metcalf said. “But that doesn’t change the total amount of diesel that’s out there. And if you look at U.S. refineries, they’re operating at near record capacity levels. There’s no ability to process more or create more diesel. That’s the supply side.”

Could farmers end up paying more?

“On the demand side,” Metcalf continued, “if we shift some of that diesel into road use, what does that mean? It means that the diesel is going to be more expensive for farmers and other off-road use, like construction.”

In other words, opening red-dyed diesel to highway drivers may slightly reduce on-road fuel costs while actually raising the effective price paid by the off-road users the fuel was meant to help. Farmers who already burn red-dyed diesel in their tractors and combines don’t gain a new tax break from the order.

Diesel prices have surged primarily because of two overseas conflicts: the war with Iran, which has disrupted tanker traffic through the Strait of Hormuz, and the Russia-Ukraine war, where Ukrainian drone strikes on Russian refineries prompted Moscow to restrict its own diesel exports. With global supply tight and U.S. refiners already running hard, the system has little slack.

Only a return to more stable global production and refining will have a true and lasting impact on domestic diesel prices. A temporary tax deferral is unlikely to deliver much relief, and it may even backfire if it drives up fuel costs for farmers, construction firms, and other off-road users.

This story was originally featured on Fortune.com

Alabama-Quassarte Tribal Town in Oklahoma launched PokkeTrades, while three California tribes launched their own apps: Kletsel Dehe Wintun Nation introduced Sovereign Edge through its economic arm, KEDA; Greenville Rancheria launched Greenville; and Alturas Indian Rancheria launched AIR Markets, according to a press release Kalshi shared with Fortune.

“These markets offer tribes a chance to participate in an emerging industry, especially those whose geography has limited revenue generation,” said Michael Stopp, vice president for economic development of the Alabama-Quassarte Tribal Town, in the release.

The announcements came less than a month after the Tunica-Biloxi Tribe of Louisiana introduced SaltTrade Derivatives, the first prediction market app launched by a Native American tribe.

Each tribe owns its app and controls its brand. It also handles marketing and works directly with customers. Kalshi runs the trading behind the scenes, matching buyers and sellers and processing their trades. It also holds funds and monitors trading activity. All four apps connect to Kalshi’s national exchange, giving customers access to the same pool of buyers and sellers.

The partnerships also mark a sharp departure from what had been a uniformly adversarial relationship between Native American tribes and prediction platforms. For decades, casino gaming has served as the primary economic engine for tribal self-governance, generating over $46 billion in revenue in 2025 alone, according to the National Indian Gaming Commission. Prediction markets are competing for the same customers. Platforms such as Kalshi and Polymarket—both of which have valuations that exceed $20 billion—let users trade contracts tied to sports and other events with a few taps on their phones.

The rise of prediction markets has unleashed a flurry of lawsuits from Native American tribes across the United States. Kalshi faces lawsuits from tribes including Blue Lake Rancheria in California, the Ho-Chunk Nation in Wisconsin, and the Mescalero Apache Tribe in New Mexico, which allege its sports contracts violate tribal gaming rights.

The latest four tribes to partner with Kalshi see these apps as a way to earn money without needing to build a gaming facility. The wealth built from tribal gaming remains heavily concentrated. Just 9% of operations bring in 56% of all industry revenue, while more than half of all venues share less than 5%, per a National Indian Gaming Commission report.

“The digital commodities marketplace gives us an opportunity to build beyond those limits,” said Eric Wright, CEO of KEDA. “Economic self-determination means having the authority to evaluate emerging opportunities and decide for ourselves what serves the long-term interests of our Nation.”

This story was originally featured on Fortune.com

The study, led by researchers at the Veterans Affairs St. Louis Health Care System, found that use of these GLP-1s, as they are classified, was associated with reduced risks of a number of health problems, including substance abuse disorders, suicidal ideation, schizophrenia and other psychotic disorders. Risk reduction was also found for neurocognitive disorders like Alzheimer’s and dementia, along with infections, liver cancer, and even life-threatening clotting disorders, like blood clots in the lungs.

How and why the drugs have such wide-ranging health benefits is subject to further research, and the GLP-1 drugs were found to increase risks of some conditions, including kidney stones and low blood pressure. On balance, though, the study’s results suggest a new frontier of potential benefits—subject, of course, to further investigation and the test of time.

“I think we need more data to help us understand whether these benefits are sustained over long periods of time, and we don’t know whether these medications carry serious long-term risks after five years or 10 years of use,” says Ziyad Al-Aly, who led the study and is director of the Clinical Epidemiology Center at VA St. Louis Health Care System. But given the evidence in total, Al-Aly says, “I’m more encouraged about the future of GLP-1 and its potential promise.”

A lower risk of dozens of diseases

The study included nearly 2 million patients, making it the largest ever conducted on this group of glucagon-like peptide-1 receptor agonists. All the participants, who were followed for a median of about three-and-a-half years, had type 2 diabetes. About 215,000 used GLP-1 drugs, while the rest were taking various other diabetes medications.

Though semaglutides like Ozempic and Wegovy dominate the market in the U.S., Al-Aly and his team included all GLP-1 drugs in the study. They evaluated the drugs’ effect on a comprehensive catalogue of 175 health outcomes and found that compared with usual care, where people continued use of their regular medications without a GLP-1 drug, GLP-1 use was associated with decreased risk of 42 outcomes (24%) and increased risk of 19 (11%).

Some of the results were unsurprising. Beyond glucose control and weight loss, benefits like a reduction in the risk of stroke, heart attacks, major cardiac events and kidney disease were expected, since some GLP-1 medicines have already been shown to be beneficial in reducing the incidence of these disorders in people with type 2 diabetes in clinical trials. But Al-Aly was particularly struck by the drugs’ apparent neurologic effects when compared with the care patients would normally receive—that is, without a GLP-1.

“There’s very clearly some neurotropic aspect to GLP-1, in the sense the drugs actually reduce risk of quite a number of neurologic conditions,” the researcher says. In the category of substance use disorders, for example, risk reductions were seen for alcohol, marijuana, stimulants and opioids.

GLP-1s may suppress the centers in the brain responsible for cravings—the same mechanism that could contribute to weight loss, says Al-Aly. In a similar vein, a study published this summer showed that patients taking semaglutide had a significantly lower risk of seeking medical care or counseling for tobacco use compared to those on other anti-diabetes medications.

Weight loss and reduced inflammation 

Like any study, there are limitations. The data are drawn from the VA database, which skews heavily male, white and older. (The large size of the study may reduce that effect somewhat.) And Al-Aly said this test, for so many health outcomes, might miss some weaker signals that could be detected by more narrowly focused studies of similar sample size.

The study found mild reductions in the risk for dementia (8%) and Alzheimer’s (12%)—not overwhelming, but still noteworthy. “It’s not the panacea,” Al-Aly says, but rather a sort of side benefit of GLP-1 agonists. The researcher likened the drug’s effect on those conditions to “a side gig. Nobody designed GLP-1s to reduce Alzheimer’s disease.”

Preclinical studies have highlighted the potential brain-protective effects of these drugs and the role they may have in treating various neurodegenerative diseases. Use of these in individuals with Alzheimer’s and Parkinson’s disease is already the subject of multiple trials.

Studies have shown that a GLP-1 drug, exenatide, sufficiently crosses the blood-brain barrier in rodent models of Parkinson’s disease, leading to improvements in motor performance, behavior, learning and memory. Human studies have seen more mixed results with respect to the drugs lowering the rate of Parkinson’s, slowing disease progression or improving motor function in affected individuals.

Nigel Greig, who heads drug design and development within the National Institute on Aging at NIH, says that part of the apparent usefulness of a drug like Ozempic may lie in the fact that its intended target, the GLP-1 receptor, has wide distribution across the tissues of the body. That includes the heart, lung, pancreas, digestive tract, arteries and veins, brain, kidney, and more

“Equally important, numerous diseases are likely underpinned by common fundamental mechanisms—like inflammation, insulin resistance and oxidative stress—that GLP-1 based drugs can positively impact,” Greig says. Clearly, those mechanisms can extend well beyond diabetes control or weight loss promotion.

GLP-1s also reduced the risk of respiratory failure, sepsis, chronic obstructive pulmonary disease and pneumonia in Al-Aly’s Nature Medicine study. A previous review of 28 randomized control trials showed that GLP-1s lowered the incidence of respiratory diseases.

“I think the elephant in the room here is that GLP-1s reduce the risk of obesity–and it’s possible that obesity is kind of the mother of all ills,” Al-Aly says. “Obesity impairs metabolic health and leads to chronic low grade inflammation, affecting multiple body systems and making the immune system less efficient.” By assisting with weight loss, Ozempic or GLP-1s can improve that underlying immune response and reduce inflammation, among other things, the researcher says.

The future of GLP-1s

According to research by the Kaiser Family Foundation, about 1 in 8 adults in the U.S. has taken a GLP-1 drug to treat diabetes, promote weight loss or prevent heart attacks or strokes (in those with a history of heart disease). But nearly 60% quit within 12 weeks, before they’ve been on the drug long enough to meaningfully gauge its effectiveness for their condition.

There are many possible explanations for that, including side effects and the drug’s cost. According to a KFF analysis, the list price for GLP-1 drugs can range between $936 and $1,349 before insurance, manufacturer coupons or rebates. As the GLP-1 patients in Al-Aly’s study were all veterans, with full drug coverage, their retention rate was higher—roughly 70% after a year.

How the study’s results will be interpreted or acted upon remains to be seen. But “off-label” use of drugs, in which they’re prescribed to treat conditions for which they weren’t originally designed, has such a long history in the U.S. that it wouldn’t be uncommon for GLP-1s to find multiple uses over time. The Food and Drug Administration has approved semaglutide to reduce the risk of cardiovascular death, heart attack and stroke in adults with heart disease and obesity.

As for side effects, Al-Aly noted that the study found signals for drug-induced pancreatitis, arthritis, as well as interstitial nephritis, a disease characterized by inflammation within the kidneys. Other adverse effects include gastrointestinal issues like heartburn, vomiting or rare cases of stomach paralysis, although those were already well known with some GLP-1s.

The researchers also saw an increased risk of fainting spells and a higher risk of having episodes of low blood pressure. While, they don’t fully understand why, it’s possible, Al-Aly says, that when patients lose weight on the GLP-1s, their blood pressure improves—but their anti-hypertensive medications are never adjusted, so they faint or experience low blood pressure.

One thing researchers cannot yet answer is what would happen to patients who stopped taking a GLP-1. If they experienced enhanced protection against certain conditions while on the drug, would their risks increase again once the regimen ends?

“One thing we know is that for people who discontinue GLP-1s, their weight gain rebounds. We know that very, very clearly,” says Al-Aly. “I think we would need more data to help us understand whether the protective effect of these medicines (against certain conditions) is sustained.”

That, says Al-Aly “is the billion- or even trillion-dollar question.” A new frontier of possible uses for these drugs likely depends in part on the answer.

More on Ozempic and GLP-1s:

• Fake Ozempic and Wegovy are flooding the market. Here are 4 steps to avoid dangerous counterfeit weight loss drugs
• Zepbound vs. Wegovy: New clinical trial says this weight-loss medication sheds more pounds
• They lost weight on Wegovy. Here’s what happened when they stopped taking it
• How drugs like Ozempic and Wegovy work for weight loss: ‘Food tastes good, but I don’t want it’
• Our list of the best online semaglutide programs

This story was originally featured on Fortune.com

The 18-year-old had been under investigation for nearly two years, trading messages with a confidential informant, before agents sold him the rifle and ammunition and arrested him on the spot, they said.

“This was a radicalized young man who needed to be watched,” U.S. Attorney Daniel Rosen said as he praised the FBI investigation. Agents “were able to detect a threat, and disrupt it, before the defendant carried out his intention to kill, in his words, maybe 30 to 60 people.”

Sheikhdoon Abdullahi Mohamud, 18, was arrested Tuesday in a public park in Richfield, several miles northwest of the Mall of America, after he paid $600 to an undercover FBI employee posing as a member of ISIS for an AK-47 and 200 rounds of ammunition, according to the criminal complaint.

Mohamud appeared in a yellow jail uniform for his initial hearing Wednesday in federal court in St. Paul.

In text messages with the informant, Mohamud repeatedly affirmed his intent to carry out an attack to kill non-Muslims at the mall on Oct. 24, the date of a scheduled Hmong arts celebration, the complaint said.

It was not immediately known if the teen had an attorney who might speak on his behalf. The Associated Press could not immediately identify any of his family members to contact for comment.

Rosen said Mohamud is an American-born U.S. citizen of Somali descent. According to the complaint, in his text messages he expressed admiration for recent terror attacks as well as an interest in traveling to Somalia to join ISIS. Ultimately he decided he couldn’t afford to travel there and would attack the mall instead, it said.

The teen first came to the FBI’s attention in December 2024, the same month the source began exchanging messages with him, the complaint said. By May 2025, agents were concerned enough that they went to his home. His mother told them he wasn’t home from school yet and called the boy’s father, who declined to speak with them and said he would give their contact information to his lawyer, the complaint said.

“What I would say to parents all over Minnesota, and all over America: If you get any indication — especially if you hear from federal law enforcement — that your son or daughter is being radicalized, to the point where federal law enforcement wants to come and give you that warning, those are warnings that ought to be heeded,” Rosen said.

Tens of thousands of people of Somali descent live in Minnesota, dating back to when refugees began arriving in the 1990s. Over the years several have been charged with attempting to join ISIS or other terror groups.

This story was originally featured on Fortune.com

While recent studies have revealed that these drugs may help stop opioid addiction, smoking, and even help people live longer, a new study published in the journal Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association now also links semaglutide to a decreased risk of Alzheimer’s disease.

In the study, researchers at Case Western Reserve School of Medicine analyzed three years of electronic records of nearly one million Americans with type 2 diabetes. They compared patients using semaglutide programs—a glucagon-like peptide-1 (GLP-1) receptor agonist—to those prescribed one of seven other anti-diabetic drugs, including metformin, dipeptidyl-peptidase-4 inhibitors (DPP-4i), and other GLP-1RAs (which mimic the effects of GLP-1s). Patients prescribed semaglutide had a 40% to 70% reduced risk of first-time Alzheimer’s diagnosis, compared to other antidiabetic medications.

“The result is what we expected,” Rong Xu, biomedical informatics professor at Case Western Reserve University and lead researcher on the study, tells Fortune. 

Given other studies that show semaglutide can help reduce inflammation in the body, prevent neuron damage, promote weight loss, control diabetes, and prevent cardiovascular disease—in addition to helping curb the urge to smoke and drink—Xu says the drug is already lowering risk factors for Alzheimer’s.

“If we can target these risk factors…semaglutide can be beneficial in preventing or slowing down Alzheimer’s disease symptoms or development,” Xu says.

Xu makes it clear that this study wasn’t meant to examine whether semaglutide lowers the risk for Alzheimer’s in people without diabetes.

“Diabetes itself is a risk factor [for Alzheimer’s],” she says. Because of that, Xu and her team wanted to see if people at heightened risk—those who have diabetes—would have better outcomes with semaglutide versus other diabetes medications.

Researchers can’t say that semaglutide does lower Alzheimer’s risk, only that there is a link. More research is needed.

But Xu is also optimistic for what these study findings might mean for people currently diagnosed with Alzheimer’s. She’d like to see if semaglutide could improve outcomes or have a therapeutic effect on those currently managing Alzheimer’s symptoms.

For more on Alzheimer’s:

• Why Barbara Corcoran of ‘Shark Tank’ ‘should’ve seen a psychologist’ while caring for mom with Alzheimer’s
• These 5 lifestyle changes improved brain function for those with early Alzheimer’s
• Belly fat tied to brain health in middle-aged adults at high risk of Alzheimer’s, study finds—but it’s worse for men

This story was originally featured on Fortune.com