Two Chinese AI firms unveiled last week models that they claim can compete credibly with OpenAI’s and Anthropic’s best systems. It was swift, and the response predictable. Commentators said Silicon Valley was shaking, while policymakers used the usual language about arms races and wake up calls.
The Associated Press claimed that a Chinese AI model “took the US tech industry by shock.” Bloomberg called it a “surprise” breakthrough, which is “roiling” markets and sending global stocks tumbling due to concerns this could force US companies to rethink spending massive amounts on AI infrastructure, data centers and chips. Business Insider asked if the launch was “The Next DeepSeek?” referring to the Chinese AI model which blindsided US AI last year. Xprize’s founder Peter Diamandis called the launch America’s AI Sputnik Moment, referring to a Soviet satellite launched at the height Cold War which encouraged US investment in science and space programs. DeepSeek, too, was widely referred to as America’s AI Sputnik Moment. This comparison felt less gratuitous at the time, as DeepSeek arrived without warning and challenged the assumptions made about frontier AI costs. It also sparked immediate responses across the financial and technology sectors.
It is surprising that these model announcements came as a shock. We have known for years that China is catching up to the US in AI. The world was shocked to see that the time may be here.
US and Chinese firms train nearly all the most popular AI models in the world. Six of the 10 top AI tools that appear on OpenRouter’s leaderboard, which tracks token consumption and benchmarks, are Chinese. Since some time the performance gap is narrowing, and recent models by companies such as Z.ai or DeepSeek are seen to be highly competitive against top-tier US labs Anthropic and OpenAI. Chinese models can be used at a much lower cost, according to reports. US firms are turning more to Chinese providers as domestic ones become increasingly expensive.
Beijing is also keen to encourage and fund homegrown AI initiatives, as well as crackdown on companies that are trying to cut ties with China. Washington’s AI policy has been a mix of heavy-handed interventions that have left its allies doubting America’s credibility and laissez-faire assumptions that the markets would see it right. This is not an easy approach against a competitor who will mobilize all the power of the government behind one technological goal.
Moonshot AI of Beijing, China, is one of the leading AI developers in China. They unveiled a brand new model Friday. Moonshot claims that it has outperformed nearly all US models, and trails only OpenAI GPT-5.6 sol and Anthropic Claude Fable 5 by a small margin. Moonshot also charges a high price for Kimi K3, charging $15,000 per million tokens. This is compared to $30 for the GPT-5.6 sol and $50 for Fable 5 Moonshot claimed that the demand was so high that the company temporarily stopped new subscriptions due to the overwhelming response. Most of the responses are focused on this particular release.
Alibaba, the Chinese tech giant, followed a few days later with an exclusive preview of Qwen3.8. The company described it as one of the “most powerful models available today,” and second only to Fable 5. This added to the controversy that Kimi K3 caused.
Both companies intend to release their flagship models in public. Moonshot and Alibaba both plan to make their new flagship models publicly available. Developers will be able to use and modify core values that were created by the AI during its training to shape responses. This is in sharp contrast with the proprietary, closed approach taken by many leading US AI laboratories, such as OpenAI, Anthropic and Google, to their frontier models.
The economics are worth a close look. The debate is still unresolved over the question of whether and how much Chinese firms are using US-based models for their training, as American companies claim. This could lead to better performance at lower costs. The tokens used to train AI systems aren’t directly comparable, nor is the price of each model. For example, a more expensive model could generate better answers with less tokens. In order to attract customers, companies also subsidise inference costs. In other words, cheaper does not necessarily mean that it is better or less expensive.
Even so, it is possible that Chinese laboratories will eventually develop models which aren’t just cheap alternatives, but could actually match or surpass their US competitors. Companies that are a little behind the curve could have a huge impact, if they produce models which are better, cheaper, and easier to use, or come with more favorable terms. It could directly affect US businesses, the economy and national security.
Anthropic, and OpenAI both are preparing for what may be trillion-dollar IPOs. These valuations depend in part on their expectation to dominate the AI market globally. Chinese models that are capable of competing with the US could undermine this assumption and potentially drive away customers. They may also squeeze margins and change growth assumptions. Some US startups have already reported that they are turning to Chinese AI models because of the high cost. A wider market threat exists, affecting more than just a few AI companies. The US tech market is dominated by stocks, with a large share of the US stock markets attributed to recent expectations of continued growth in AI demand. Companies have invested hundreds of billions in data centers, chips and energy infrastructure, based on the belief that American companies will dominate. Investors would question the costs if Chinese laboratories could capture a portion of this demand or produce and operate models for less. Any reassessment would be a ripple effect on all these industries and the pensions or savings of millions.
Security is also a concern. Even if they are behind the US, highly capable Chinese open models could provide advanced AI systems to a wider range of users. This is especially true if US companies limit access or enforce stronger security measures. Cybersecurity leaders were concerned that limiting access to Anthropic’s latest models would hinder the ability of defenders and security experts to identify vulnerabilities. These restrictions become harder to defend if similar models are readily available in other countries. If organizations are denied US-made models, they may be forced to turn to Chinese options to protect their networks. Or else risk being more vulnerable to attacks by attackers who can use similar tools. Kimi K3 has already been reported to have identified and corrected cyber vulnerabilities which OpenAI Codex or Anthropic Fable wouldn’t touch because of safety safeguards. Some models that are less capable can be a danger, and they already seem to do so. Z.ai, a Chinese company, claimed in June that its GLM-5.2-model could compete with Anthropic Mythos for cybersecurity, even though the model lacked on more general tasks.
It is difficult to assess the capabilities of either model as neither has been released fully. Therefore, benchmarking claims made by companies should be treated cautiously. Despite this, it is unlikely that companies have misrepresented their performance.
The exact ranking of the models is not important. The fact that China’s AI leaders are producing systems which could rival the best US laboratories is the main conclusion. They are doing this with such regularity, that it is no longer necessary to treat each release as something that would be as shocking as the next “DeepSeek”, or even a “Sputnik Moment”.

Then he continued, “I wouldn’t be a purchaser, because of interest rates. Even if the inflation rate was only 2%, a 10-year bond would be worth 4.5 to 4% and 3.25%-3.5% for unsecured loans. They are almost at that level today.”
Dimon speaks about two factors that are likely to influence the yields of government bonds in the future: inflation expectations and borrowing by the government.
The long-term Treasury bills — 10-, 20- or 30-year notes — act as a thermometer for economic forecasts. In addition to incorporating inflation expectations into the yields of longer-term Treasuries, they also provide benchmark rates that lenders can use for their loans. The asset with low risk, government borrowings, is compared against interest payments consumers pay.
These yields reflect the interest rates charged to other borrowers in the market, such as those who borrow for cars, houses and credit cards.
Fear of crisis
The bond market’s foundation is the belief in the government’s capacity to repay its debts. This is a solid assumption, given the strength of the U.S. economic system and the Central Bank of America’s power to affect the price of debt by changing its money supply.
The Treasury is accumulating debt rapidly. It’s currently at over $39 trillion, and the interest payment rate for the Treasury is $24 billion per week. Economists, as well as private sector experts, are concerned that lenders may start to demand higher rates at some point in order to reflect the increased risk of funding.
The U.S. operates at around a 120% debt to GDP ratio, Europe is at about 90% and the UK is a bit over 95%.
Dimon stated that “these are high numbers for debt and deficit, but we’re doing well.” To have such high numbers, you usually need a major recession, depression or war.
Dimon, who has repeatedly suggested this issue would come to an end and once again urged policymakers to “deal with maturely” and “sit down,” has made the same recommendation.
The banker said, “That’s the better option.” The other option is to just wait until it becomes a big problem. My guess is this is what will happen. And that will exhibit itself with higher interest rates, the market getting rattled a little bit, people talking about it constantly–remember the bond vigilantes–hopefully not worse than that, but it could be worse than that.”

Financial math that would have you fleeing from high-tax states to a Florida oasis is hit by a roadblock.
Miami has the highest cost of living in the United States, thanks to a combination rising property taxes and insurance rates, as well as everyday inflation.
Bloomberg found in a recent analysis that U.S. Bureau of Economic Analysis data indicates that the cost of living for the Miami, Fort Lauderdale, and Palm Beach regions — dubbed “the Gold Coast” by some — are now approximately 5% more expensive than the New York metro area and surrounding suburbs.
CEO: MIAMI’S LUXURY BOOTH FUELS MECCA’ FOR THE WEALTHY, AS OTHER BUYERS FEEL PRICED OUT
According to the U.S. Bureau of Labor Statistics, South Florida’s consumer prices have risen by 36% in 2019. This is the second highest inflation rate amongst major American markets behind only Tampa.
S&P CoreLogic CaseShiller data shows that South Florida home values have increased 79% since pandemic. Meanwhile, the average homeowners’ insurance premium is $8,292, the most expensive in the nation and four times as much as the cost to insure a house in New York.
U.S. Census Bureau figures also show that Miami’s typical household income is about $1,000 less than the median national figure. The cost of dining out, for example, has risen 4% in one year to $94 per head per bill. This compares to the average restaurant check of New York City, which comes to around $79 each.
GET FOX BUSINESS ON THE GO BY CLICKING HERE
Florida could see a price reduction after the Governor. Ron DeSantis, the Florida Legislature and other state officials approved the proposed amendment to the Florida Constitution for November 2026’s general election ballot. The proposal would increase the homestead exemption. If voters approve, the proposal would allow eligible homeowners to receive a maximum exemption of $250,000 from property taxes other than school-related.
The constitutional amendment, if approved by 60% of Florida voters at the minimum, could result in lower property taxes and savings of millions for Florida homeowners.

Climate change and record temperatures are driving demand for air conditioners, while the Trump administration has been dismantling federal programs that promote energy efficiency.
Christine Egan is the executive director of Collaborative Labeling and Appliance Standards Program. This nonprofit organization works to improve efficiency policies around the world. Energy prices and demand are increasing, so you would think that we’d want to make our products as energy-efficient as possible.
When electricity prices are increasing faster than the inflation rate and gas is about $4 per gallon, consumers may end up burning more gasoline and using more energy. Even though more energy-efficient cars and appliances are more expensive initially, they can save consumers money in the long run because they consume less electricity.
Both parties, since the 1970s have issued rules to help reduce energy consumption in cars, buildings, and appliances. These efficiency standards could also reduce the amount of emissions which are heating the earth.

HONG KONG – The Chinese Kimi K3 artificial intelligence system, which caused quite a stir among the U.S. technology industry, suspended subscriptions following a surge in demand that exceeded the capacity of the machine within days after its launch.
This development highlights the challenges Chinese AI models face in serving a rapidly growing number of users at home and overseas at a moment when U.S.-China technology is escalating.
Chinese advanced AI models that are open source and lower cost, such as DeepSeek V4, have gained global attention. DeepSeek’s V4 model, which was released in early 2025, helped China become a major competitor with the U.S.
Moonshot AI (based in Beijing) wrote late Sunday that “Kimi K3 received much more love than expected.” The demand for our products has increased over the last 48 hours.
Moonshot stated that existing customers are given priority and new subscribers will be put on hold temporarily. Moonshot said that it was adding capacity at a rapid pace and would reopen subscriptions in small batches. A similar message was also shared by the AI startup on Chinese social networks.
K3’s late-last week public launch, which aimed to challenge more advanced AI models from Anthropic and OpenAI, among others, has rattled U.S. competitors. It’s the largest AI model in terms of parameters. AI models are measured by a number called parameters.
Lian Jye su, chief analyst of the Omdia technology advisory and research group, said that new model releases can cause a surge in interest which strains existing computing infrastructure. This does indicate that Moonshot AI doesn’t have enough compute chips to meet the current surge of demand.
Su stated that Moonshot did not anticipate the popularity of K3 to the extent it has. He said that K3’s compute needs are “very high”, making it difficult and costly to allocate compute resources.
Kimi K3 was ranked first in the rankings of major AI models after their public release, using a measure called front-end programming capability.
K3 has put pressure on the stocks of American tech titans, as they fear that cheaper Chinese AI models could undercut pricing and demand for U.S. companies. This is despite the fact that U.S. restrictions had already prevented China from accessing the latest technologies in the world, such as the cutting edge chips.
K3 AI is the latest Chinese AI model that has attracted global attention.