Kevin O’Leary announced plans to reduce the scale of his AI data center venture in Utah following political pushback. The prominent entrepreneur and "Shark Tank" star revealed in a letter to the president of the Utah Senate that he intends to scale back the ambitious project, originally spanning 40,000 acres, by nearly 50%.

Kevin O’Leary announces plans to downsize his Utah AI data center project following political backlash.

A moody dusk scene in the Utah desert: Kevin O’Leary in a dark suit stands in the foreground, backlit by the last natural light as he surveys a half-built AI data center with idle cranes and partly dismantled scaffolding. Scattered protest signs reading “Stop Expansion” lie abandoned in the dusty dirt, and a single “AI Data Center” sign hangs askew on the building’s shrinking shell.

Kevin O’Leary says he will shrink his Utah AI data center project after political backlash - NBC News

- Kevin O'Leary plans to reduce his Utah AI data center project by half.  
- The decision follows political backlash regarding the project's size and impact.  
- O'Leary communicated the changes in a letter to Utah's Senate president.  

Kevin O’Leary says he will shrink his Utah AI data center project after political backlash

The “Shark Tank” mogul said he would cut the 40,000-acre project by roughly half in a letter sent Thursday to Utah’s Senate president.

Kevin O’Leary says he will shrink his Utah AI data center project after political backlash

Your choices regarding privacy are essential to us. By selecting 'Accept all', you agree to the use of technical identifiers along with your browsing and search data for analytics purposes, personalized advertising, content measurement, and audience research, as well as for service development. If you choose to proceed, we, along with our partners—which include over 250 affiliates under the IAB Transparency & Consent Framework—will store and access data on your device via cookies. This may also involve utilizing precise geolocation and additional personal data to enhance the analytics processes and tailor advertising and content to your preferences.

Should you prefer that we and our partners refrain from using cookies and your personal data for these extra aims, simply click 'Reject all'.

For those interested in personalizing their privacy choices, selecting 'Manage privacy settings' will assist you in making adjustments tailored to your needs.

Please be aware that you have the ability to revoke your consent or modify your selections at any moment by engaging with the 'Privacy and Cookie settings' or 'Privacy dashboard' links available on our platforms and applications. To understand our practices regarding your personal data further, please consult our Privacy Policy and Cookie Policy.

Your options for privacy

A lone adult standing before a wall of semi-transparent glass panels, each etched with a privacy icon—padlock, fingerprint, shield, on/off slider—bathed in warm natural light as they reach out to tap one option. Modern minimalist interior, soft-focus background.

AI is quietly becoming an unofficial and potentially unwanted 'third' in relationships - Yahoo

- Users can accept or reject cookies and personal data usage for various purposes.  
- Clicking 'Manage privacy settings' allows customization of privacy choices and data usage.  
- Consent can be withdrawn anytime through the 'Privacy dashboard' links on the site.  

Your privacy choices

Accept all', we and technical identifiers and browsing and search data, for analytics, personalised advertising and content, advertising and content measurement, and audience research and services development. If you click '', we and our partners , including 250 who are part of the IAB Transparency & Consent Framework, will also store and / or access information on a device (in other words, use cookies) and use precise geolocation data and other personal data such asand browsing and search data, for analytics, personalised advertising and content, advertising and content measurement, and audience research and services development.

If you do not want us and our partners to use cookies and personal data for these additional purposes, click 'Reject all'.

If you would like to customise your choices, click 'Manage privacy settings'.

You can withdraw your consent or change your choices at any time by clicking on the 'Privacy and Cookie settings' or 'Privacy dashboard' links on our sites and apps. Find out more about how we use your personal data in our Privacy Policy and Cookie Policy.

Your privacy choices

Three years back, Microsoft CEO Satya Nadella propelled the company into the forefront of the AI landscape, earning accolades and comparisons to a "superhero" from a former executive. In February 2023, after early investments in OpenAI paid off, Nadella introduced an AI-enhanced Bing search engine at a Seattle event, signaling a challenge to Google's dominance. "A race starts today," he proclaimed, receiving waves of praise. Later that year, amidst a crisis within OpenAI's board, Bill Gurley highlighted Nadella's role in transforming the company’s reputation, and CNN Business awarded him "CEO of the Year."

However, as the tech industry shifts, Nadella's legacy is now on the line. Microsoft's stock has declined over 24% in the past year, far worse than other major tech players. There’s increasing investor skepticism over whether the company’s massive investment in AI will yield returns. For instance, Copilot, Microsoft’s leading AI offering, is falling behind rivals like ChatGPT and Claude. On another front, LinkedIn faces backlash due to an influx of AI-generated content, and Xbox is reportedly struggling, leading to layoffs and a reorganization following its record $69 billion acquisition of Activision Blizzard.

Internally, employees are questioning the rationale behind a planned $190 billion investment this year in AI infrastructure, particularly as generative AI transforms workplaces, software development, and information consumption. The effectiveness of three of Microsoft's key businesses—Microsoft 365, GitHub, and Azure—hangs in the balance. Investors are anxiously awaiting Wednesday's fourth-quarter earnings report to gauge the situation.

As generative AI becomes ubiquitous in corporate America, software companies are scrambling to avoid what many are calling a SaaSpocalypse. This predicament is particularly daunting for Microsoft, which boldly staked its future on AI. What was once a guiding star could now pose a risk.

For decades, Microsoft’s productivity tools, including Word, Excel, and PowerPoint, have been staples for knowledge workers. Now, many are beginning to utilize AI tools for these tasks. Gartner analysts warned early this year that AI might disrupt traditional productivity suites like Microsoft 365 and Google Workspace, impacting a $58 billion market.

Microsoft leaders cite ongoing adoption of Microsoft 365 and Copilot as evidence of continued customer interest. "The M365 business is seeing tons of new adoption and M365 Copilot usage," one executive noted, emphasizing the push for capacity to accommodate demand.

GitHub, the software development platform acquired in 2018, is experiencing similar challenges. Although it has maintained its stronghold among developers and capitalized on early AI coding advantages, competition is fierce. Emerging tools like Cursor—recently targeted for acquisition by SpaceX for $60 billion—and Anthropic’s Claude Code have gained traction, leading Microsoft to contemplate substantial overhauls to GitHub to remain competitive. Increased AI demand has also led to numerous outages this year.

In the cloud computing segment, Microsoft is grappling to meet skyrocketing demand for Azure. Despite this year’s heavy investments aimed at expanding data center capacity for AI workloads, executives indicate that building new capacity remains a challenge. As a result, the company has raised sales quotas for its Azure sales team by as much as 30%.

While Azure continues to be the fastest-growing segment within Microsoft, executives reveal that balancing demand with production capabilities has become increasingly difficult. Chief Financial Officer Amy Hood acknowledged that the company is prioritizing its own AI products over cloud services for external customers. In doing so, Azure growth could have surpassed 40%, instead of a reported 39%, had resources been allocated differently, Hood suggested.

Consequently, Microsoft has been sourcing additional cloud capacity from competitors. Following outages at GitHub, Amazon provided temporary assistance. Microsoft also explored leasing Oracle's cloud infrastructure but ultimately withdrew due to compliance and security concerns. Now, the company is in talks with providers like Amazon and Google to alleviate constraints, indicating, "We are shopping for capacity everywhere," according to insiders.

This strategy of prioritizing internal services has drawn mixed responses from Wall Street but remains a clear focus within Microsoft. Internally, conversations have shifted toward the implications of these choices, especially decisions like prioritizing Adobe's growth over that of M365, leading to confusion about how to communicate these strategies externally.

As Nadella faces rising pressure on core business areas, he has been reshaping the company’s leadership structure. He promoted Judson Althoff to oversee Microsoft's commercial sector to allow for more direct leadership in AI initiatives. Althoff's previous role in sales now gives him a prominent position. This reorganization is described by Nadella as part of a major transition into an AI-centric approach.

Significant changes are unfolding throughout Microsoft, including adjustments to its performance review system to better align with current goals. These revisions, perceived as a return to the competitive practices of the Steve Ballmer era, aim to boost accountability but have raised concerns about a renewed culture of fear.

For years, Microsoft has thrived on controlling the workspace environment for both users and developers. However, the rise of AI threatens to disrupt those foundations. Now, Nadella's legacy may hinge not on the quest to develop superior AI, but on his ability to safeguard the essential businesses that have historically contributed to Microsoft's standing as one of the world’s most valuable enterprises.

Exploring Nadella's Most Challenging Year

A shadowy office at night: Satya Nadella seen from behind, standing before a bank of flickering screens showing plunging stock charts, urgent cloud-outage alerts, and AI ethics headlines. The Microsoft logo looms large, softly backlit on the wall. Deep shadows, a single overhead warm spotlight, tense and cinematic.

Inside Nadella's hardest year: Is Microsoft's North Star also a noose? - Business Insider

- Satya Nadella's leadership is under scrutiny as Microsoft faces AI growth challenges.  
- Microsoft's stock has dropped 24%, raising investor concerns about its AI investments.  
- Internal restructuring and pressure reflect the urgency to maintain Microsoft's competitive edge in AI.  

Three years ago, Satya Nadella catapulted Microsoft to the front of the AI race and became "like a superhero," one recent former Microsoft executive said.

In February 2023, after betting early on OpenAI, the CEO unveiled Microsoft's AI-powered Bing search engine before a packed audience outside Seattle, and declared a war on Google's search dominance. "A race starts today," he said. Waves of adulation followed. When Nadella helped navigate OpenAI's board crisis later that year, Bill Gurley called it an "amazing shift in corporate reputation." CNN Business chose Nadella as the CEO of the Year.

Inside Microsoft and across the tech industry, Nadella was hailed for seizing the future. Now, his legacy is at stake.

Microsoft's stock is down more than 24% from 12 months ago, significantly worse than the rest of the Magnificent 7. Investors have grown increasingly skeptical that the company's multibillion-dollar AI bet will deliver. Copilot, Microsoft's flagship AI product, lags behind other AI tools like ChatGPT and Claude. LinkedIn has drawn criticism for becoming flooded with AI-generated hustleporn. Xbox's business is "not healthy," its CEO recently said, and undergoing layoffs and restructuring as it tries to justify the company's record-breaking $69 billion Activision Blizzard acquisition.

And inside the company, employees are questioning Microsoft's plans to spend a record $190 billion this year to build AI infrastructure. As generative AI changes how people work, write software, and consume information, three of the company's core businesses hang in the balance: Microsoft 365, GitHub, and Azure. Investors will get a report card on these challenges on Wednesday, when the company releases its fourth-quarter earnings results.

As AI adoption spreads through corporate America, every software company is fighting to fend off the so-called SaaSpocalypse. But the battle is particularly fraught for Nadella's Microsoft, which made an early and loud bet on AI to propel the company's future. Now the company's north star has also become a potential noose.

For decades, Microsoft's productivity software has been the default homeroom where knowledge workers start their day. They opened Word to write, Excel to analyze data, and PowerPoint to build presentations. Now, millions of those workers are beginning to do all these things directly inside AI tools. Gartner analysts earlier this year predicted AI would threaten to dethrone traditional productivity suites like Microsoft 365 and Google Workspace in a $58 billion market shakeup.

Microsoft executives point to continued growth in Microsoft 365 and increasing Copilot adoption as evidence customers still want Microsoft's products at the center of their workdays. "The M365 business is seeing tons of new adoption and M365 Copilot usage," one executive said, who said the company is specifically chasing computing capacity to meet the demand.

GitHub faces a similar challenge. Since acquiring the software development platform in 2018, the company has held a dominant position with developers and had an early advantage in AI coding through GitHub Copilot. And it continues to grow: The platform recently had its "best month ever," an executive told employees in internal meeting comments viewed by Business Insider, though he didn't say by what measure.

But upstarts have swarmed in, as millions of engineers have adopted Cursor — which SpaceX recently announced plans to acquire for $60 billion — and Anthropic's Claude Code. As Business Insider previously reported, executives have discussed internally the need to overhaul GitHub to better compete with those AI-native coding tools. AI demand has also strained Github. As AI usage surged GitHub has experienced dozens of major outages this year.

The company is also struggling broadly to keep up with the demand for compute capacity. Despite this crunch, Microsoft is raising salespeople quotas for selling its cloud computing platform, Azure, some by 30% this year, according to people familiar with the change.

Azure remains Microsoft's fastest-growing strategic business, but internally executives say it has become a constant balancing act. Demand for computing infrastructure has outpaced the company's ability to build new capacity, forcing Microsoft to make difficult decisions about where its resources go. Even with this year's $190 billion capital expenditures — largely to expand data-center capacity for AI workloads — executives say the company is still constrained.

Earlier this year, Chief Financial Officer Amy Hood suggested Microsoft was prioritizing scarce computing resources for its own AI products before allocating the remaining capacity to Azure customers.

"The first thing we're doing is solving for the increased usage in sales and the accelerating pace of M365 Copilot, as well as GitHub Copilot, our first-party apps," Hood said during Microsoft's January earnings call. "Then we make sure we're investing in the long-term nature of R&D and product innovation... Then what you end up with is the remainder going towards serving the Azure capacity that continues to grow in terms of demand."

Why would Satya prioritize growing Adobe over growing M365? Microsoft executive

If Microsoft had allocated the GPUs that came online during the first half of its fiscal year to Azure instead of its own AI products, Azure growth would have exceeded 40% instead of 39%, Hood said. Microsoft previously reported $75 billion in Azure revenue for its 2025 fiscal year.

That earnings report triggered one of Microsoft's biggest post-earnings stock declined by more than 10% as investors questioned the company's slower Azure outlook despite record AI spending and growing concerns that Microsoft was diverting capacity away from cloud customers.

Executives who spoke to Business Insider say those tradeoffs have intensified.

Microsoft is so desperate for capacity that it's turning to competitors to help relieve some of those constraints. Following a series of GitHub outages, Amazon bailed Microsoft out. The company also explored leasing Oracle cloud infrastructure but Microsoft walked away due to security and compliance concerns.

Microsoft is now seeking additional cloud capacity from other providers, including evaluating Amazon and Google, according to people familiar with the discussions. "We are shopping for capacity everywhere," one of those people said.

While prioritizing internal services has a mixed reception on Wall Street, the strategy is clear within Microsoft.

"All of the supply is gone once you solve for frontier labs and our internal businesses like M365 and Microsoft AI," one executive said.

Those decisions have created difficult conversations internally.

"Why would Satya prioritize growing Adobe over growing M365?" the person said. "I have no idea how we're going to land that message with customers."

As the pressure on Microsoft's core businesses mounts, Nadella has been bearing that pressure down on his workforce, and reshaping the structure of the company and its leaders.

As Business Insider previously reported, Nadella promoted Judson Althoff to CEO of Microsoft's commercial business to free himself and the company's engineering leaders to focus more directly on AI. Althoff was previously Microsoft's longtime sales boss but the role gave him a bigger profile. In an internal memo viewed by Business Insider at the time, Nadella described the moment as "a tectonic AI platform shift."

The mounting pressure on Nadella has trickled down through Microsoft's ranks from the executive suite to the rank-and-file employee.

At the same time, Nadella has remade his inner circle. Business Insider previously reported that Microsoft effectively retired its traditional senior leadership team structure in favor of smaller, flatter leadership groups. AI CEO Mustafa Suleyman has narrowed his focus to Microsoft's superintelligence efforts, top Nadella lieutenant Rajesh Jha retired, longtime product and marketing leader Yusuf Mehdi is preparing to leave the company, and more executive changes are expected.

According to people familiar with the succession planning, Hayete Gallot, who recently returned to Microsoft from Google to lead the company's security business, is viewed internally as the long-term successor to Althoff as sales chief. Gallot previously worked for Althoff and left in what one executive told Business Insider was "not an amicable departure." Nadella recruited Gallot back to replace Charlie Bell, who moved into an individual contributor role focused on engineering quality. Rodrigo Kede Lima, who Microsoft just put in charge of a $2.5 billion AI sales unit, is also a rising star, one of the people said.

The changes extend beyond the executive suite. Business Insider has learned that Microsoft overhauled its performance review system this year, simplifying ratings into five categories while making performance distinctions significantly sharper.

Executives say the new process feels like a return to "stack ranking," the controversial system that evaluated employees relative to one another during the Steve Ballmer era. At the same time, managers have been instructed to reduce the number of employees in higher-level engineering roles as Microsoft continues flattening parts of the organization, emblematic of a broader hardcore work culture that's spread across Big Tech in the last few years.

"It's almost like the old era of Microsoft is back," one former executive said. "The old Windows era where you lead with a lot of fear and a billy club in your hand."

For years, Microsoft's greatest strength was that it owned where people worked and where developers built software. AI is beginning to challenge both assumptions at once. Now Nadella's legacy won't be defined by whether Microsoft can build the best AI, but by whether it can keep AI from eroding the businesses that made it one of the world's most valuable companies.

Ashley Stewart is a chief technology correspondent at Business Insider.

Inside Nadella's Hardest Year

HONG KONG (AP) — Artificial intelligence from China is making waves in the United States marketplace.

Raffi Krikorian, the chief technology officer at Mozilla, based in San Francisco, started using the new Kimi K3 model from Chinese startup Moonshot for his regular work shortly after its release last week. He remarked on its responsiveness, finding it quicker than the more established and pricier Claude Fable chatbot from Anthropic. Before K3, he was utilizing another strong Chinese model, Z.ai’s GLM-5.2, for tasks like managing his calendar and emails.

Krikorian's switch to Chinese AI reflects a broader trend among Americans who are increasingly picking Chinese AI solutions that offer higher efficiency at lower costs. Companies such as Coinbase, a cryptocurrency exchange based in the U.S., are turning to these models as a means to cut expenses.

The rise of these Chinese systems has generated concern among some U.S. tech leaders, yet unless there is a broad ban, their appeal to independent developers both in the U.S. and globally is likely to persist.

U.S. policy has restricted China’s access to some advanced technologies, including state-of-the-art AI chips. Treasury Secretary Scott Bessent has suggested that additional sanctions could be implemented to safeguard American intellectual property.

The administration of former President Donald Trump accused Moonshot of potentially using "covert" methods to develop K3, implying some reliance on Anthropic’s Fable. Meanwhile, some U.S. politicians and AI firms, notably Anthropic, have accused Chinese startups of illegally using techniques to replicate their models—a claim that China refutes as unfounded.

The emergence of powerful Chinese AI models this year indicates a shift in the U.S.-China AI rivalry, following the disruptive entry of DeepSeek’s AI model early last year that demonstrated performance similar to U.S. offerings but at a lower cost.

Recent models from startups such as Z.ai (released in June) and Moonshot (released in July) are becoming nearly as capable as leading models from OpenAI and Anthropic. Additionally, Alibaba showcased its Qwen 3.8 Max AI model in July, while DeepSeek revealed a preview of its V4 model earlier in April.

These advancements are yielding tangible results. Curt Meinhold, a technology executive from Greensboro, North Carolina, noted his growing preference for DeepSeek for its utility in generating sales leads and conducting business.

“Most of us don’t really need cutting-edge models like Anthropic’s Mythos or Fable,” he commented. “We just want something that works well enough.”

A July report from Goldman Sachs characterized the Chinese AI models as being at a pivotal moment for broader adoption, spurred by a rising global demand for "agentic" AI that can efficiently manage complex tasks. 

As AI services often charge based on output token counts, the cost differences with Chinese models are significant, as noted by analysts including Alex Colville of the Australian Strategic Policy Institute. “The Chinese models are quite competitive in quality,” Meinhold pointed out, emphasizing the financial advantages.

Currently, the Kimi K3 model from Moonshot is leading in popularity, with recent tracking data showing that the five most utilized models on OpenRouter are Chinese. According to Sensor Tower, K3 amassed over 930,000 downloads shortly after its July debut, marking a staggering 200% increase from the previous week, with around 86,000 of those downloads occurring in the United States—an impressive jump of 387%.

So much interest in K3 led Moonshot to temporarily halt new subscriptions due to overwhelming demand.

Nevertheless, while Chinese AI systems are advancing rapidly, experts indicate they still fall short when it comes to comprehensive capabilities compared to their American counterparts. Anastasios Angelopoulos, co-founder and CEO of Arena, noted that American firms are also exploring competitive models to drive down prices.

The advantage of open-source development is another aspect where Chinese models are thriving, allowing anyone to modify and improve them, in stark contrast to the closed-source models from U.S. companies like Anthropic and OpenAI. Chinese developers could capitalize on this openness to enhance their global reach and adoption, according to Lian Jye Su from the advisory group Omdia.

As the quality of Chinese open-source AI models increasingly rivals their U.S. counterparts, Krikorian pointed out that the landscape is shifting towards models that are predominantly Chinese-built. While discussions around tightening restrictions on Chinese AI models occur in the U.S., several American tech giants, including Microsoft, Meta, and Nvidia, have endorsed a commitment to open AI initiatives.

Moreover, specific U.S. policies have inadvertently given Chinese firms an advantage. For instance, shortly after the U.S. government imposed export controls on Anthropic’s models, Z.ai launched its GLM-5.2, quickly filling the gap left by the U.S. restrictions.

"Whenever an American model faces restrictions, a Chinese alternative can quickly emerge," Angelopoulos explained.

China is also focused on promoting AI technologies internationally. At a recent tech summit in Shanghai, President Xi Jinping advocated for open-source AI and emphasized the objective of increasing global equity, while articulating China’s aim to enhance AI capabilities, particularly in developing nations.

The competitive landscape is shifting, with Chinese firms eager to extend their reach globally. Leading AI startups in China are in the process of securing more investment to subsidize their expansion efforts, including through public offerings.

Both nations are driven to encourage a wide uptake of their AI technologies while carefully safeguarding developments that could enhance their strategic competitors, explained Chelsey Tam from Morningstar.

"The competition has evolved beyond the U.S. versus China, with Chinese labs increasingly competing against each other as well," Angelopoulos noted.

Nevertheless, the surge in AI investment, while promising, raises sustainability questions for these rapidly growing Chinese companies. For instance, Z.ai reported a remarkable 132% revenue increase to 724 million yuan ($107 million) last year, but their net losses rose to 4.7 billion yuan ($694 million).

Looking ahead, the momentum of Chinese AI models is likely to persist, particularly in the U.S. “I encourage anyone serious about their AI needs to evaluate these options,” Krikorian recommended.

Affordable, accessible, and smart: Chinese AI models are making progress in the US market.

A sunlit modern server lab with floor-to-ceiling windows revealing a U.S. city skyline at dawn. In the foreground, sleek racks of servers bear a single Chinese flag sticker and “Open AI Model” labels. A nearby workstation displays open-source AI code running, warm natural light flooding the room to symbolize the surge of cheaper, intelligent Chinese AI models making inroads in America.

Cheaper, open and intelligent: Chinese AI models gain ground, as they make inroads in the US - AP News

- Chinese AI models are gaining popularity in the U.S. for their affordability and efficiency.  
- The newly launched Kimi K3 exceeded demand, prompting temporary subscription suspensions by Moonshot.  
- U.S. restrictions on advanced technologies may inadvertently benefit Chinese AI model adoption.  

Add AP News as your preferred source to see more of our stories on Google.

Add AP News on Google Add AP News as your preferred source to see more of our stories on Google. Share

HONG KONG (AP) — China has a hot new product in the United States: artificial intelligence.

San Francisco-based Raffi Krikorian, the chief technology officer at Mozilla, which runs the Firefox browser, switched to Chinese AI startup Moonshot’s Kimi K3 for many of his day-to-day activities within days of the new, powerful model’s launch more than a week ago.

“It just seems snappier,” he said of K3, comparing it to the acclaimed and more expensive Claude Fable chatbot from San Francisco company Anthropic. Earlier, he had been using another high-performing Chinese model, Z.ai’s GLM-5.2, for routine tasks such as managing his calendar, documents and email.

Krikorian is one of a growing number of Americans adopting Chinese AI systems, which are gaining ground around the world because they are more affordable and increasingly efficient. U.S. companies like cryptocurrency exchange Coinbase said they’re switching to Chinese AI models to help trim costs.

Their popularity has frustrated some U.S. tech giants, but short of an outright ban, they’re likely to continue to appeal to independent software developers in the U.S. and elsewhere.

American-led restrictions block China from accessing some of the world’s most advanced technologies including cutting-edge AI chips and U.S. Treasury Secretary Scott Bessent has warned more sanctions could be coming to protect American intellectual property.

U.S. President Donald Trump’s administration on Wednesday accused Moonshot of using “covert” but not necessarily illegal methods to build K3 off the back of Anthropic’s Fable. Some U.S. politicians and AI companies including Anthropic have accused Chinese startups of illicit “distillation” of their models to extract their technologies, a claim that Beijing rejects as “groundless.”

China’s AI models are cheaper and ‘good enough’

The rapid rise of cutting-edge Chinese AI models this year shows the China-U.S. race in AI has moved on from early last year, when Chinese startup DeepSeek shook up the U.S. technology industry and put China on the map with an AI model that performed on par with those in the U.S. but much more cheaply.

Read More

The latest AI models launched by startups Z.ai, or Zhipu, in June, and Moonshot in July are nearly as intelligent as the frontier models from OpenAI and Anthropic, experts say. Also in July, China’s Alibaba previewed its Qwen3.8 Max AI model. In April, DeepSeek rolled out previews of its latest V4 model, challenging OpenAI’s GPT and Google’s Gemini.

Those constant advances are bearing fruit. Curt Meinhold, a Greensboro, North Carolina-based technology executive, said he increasingly prefers to use DeepSeek for tasks such as finding business leads or ways to generate sales.

“At the end of the day, most of us, the vast majority of us, 90 plus percent, don’t need (Anthropic’s) Mythos or Fable,” said Meinhold, who founded digital legacy platform LilyList. “Like, we just don’t need it, we need something good enough.”

Sign up for Morning Wire: Our flagship newsletter breaks down the biggest headlines of the day. Email address Sign up By checking this box, you agree to AP's Terms of Use and acknowledge that AP may collect and use your data pursuant to our Privacy Policy

Chinese AI models are reaching a “critical stage” for wide adoption, U.S. investment bank Goldman Sachs wrote in a July research report, especially when a surge of global “agentic” AI usage, which requires the AI models to autonomously conduct multistep, complex tasks, is driving higher demand for cost effective AI models.

AI prices are calculated per million tokens for both input and output, so the use of AI “agents” is dramatically compounding cost differences, according to analysts including Alex Colville of the Australian Strategic Policy Institute.

“The Chinese models, you know, I find them to be pretty close on code and research,” said Meinhold. “If I can pay a handful of cents per million output tokens versus 30 bucks or 40 bucks or 50 bucks, then it’s good enough.”

Moonshot’s Kimi K3 leads, but Chinese AI models have limits

Based on data over the past month, the top five most popular models on OpenRouter, a platform that tracks data across AI models, were Chinese.

Estimates from Sensor Tower, a market intelligence firm, showed Kimi had more than 930,000 downloads during the week after K3’s July release, an increase of 200% from the week before. In the U.S. it was downloaded around 86,000 times, a 387% jump.

K3 proved so popular that Moonshot had to temporarily suspend new subscriptions after overwhelming demand pushed its capacity close to its limits.

Still, while Chinese models are becoming serious competitors to the U.S., they lag American AI leaders across their overall, full-range capabilities, said Anastasios Angelopoulos, co-founder and CEO of Arena, a platform for evaluating AI systems.

And U.S. AI firms are also looking for more options to reduce pricing, such as with cheaper alternatives of competitive models, added Yasir Atalan, with the Center for Strategic and International Studies.

US policy could boost Chinese AI models

Most Chinese AI models are open-source — meaning anyone can examine and build on them — at a time when frontier models from U.S. companies like Anthropic and OpenAI are closed-source.

Chinese model vendors are expected to leverage open-source software to promote their global usage and adoption, said Lian Jye Su, of the technology research and advisory group Omdia.

As Chinese open-source models are increasingly almost as good as closed systems made by leading U.S. companies, said Mozilla’s Krikorian, “the open frontier is becoming increasingly Chinese-built.” While the U.S. weighs restricting Chinese AI models, a group of American tech firms including Microsoft, Meta and Nvidia signed an open letter published Friday backing “open” AI models.

Other U.S. tech policies also at times give China an upper hand, some experts say. China’s Z.ai released its GLM-5.2 model in mid-June, not long after the Trump administration put export controls on Anthropic’s Fable and Mythos models, keeping them offline for more than two weeks.

“Restricting an American model can immediately create an opening for a Chinese competitor,” said Arena’s Angelopoulos.

Chinese AI models look to expand globally

In China, people and businesses have rapidly embraced AI technologies that have prospered with state support. Chinese tech companies like Huawei and Tencent are also embedding AI in devices such as smartphones, AI glasses and humanoid robots.

Now China is increasingly looking to dominate AI abroad, too. At a flagship technology summit in Shanghai, Chinese President Xi Jinping championed open-source AI models and the importance of promoting greater global equity, while pledging Chinese involvement in raising AI capabilities especially in developing nations.

As is true for many other industries in China, intense competition at home is driving companies to expand globally. Leading Chinese AI startups are now raising more funding to support that effort, including through public share offerings.

Both China and the U.S. will want to “encourage widespread adoption of their AI ecosystems, while safeguarding technologies that could materially strengthen strategic rivals,” said Chelsey Tam, with investment research firm Morningstar.

“The competition is no longer simply the United States against China; the Chinese labs are also putting a lot of pressure on one another,” Angelopoulos said.

But as is the case in the U.S., huge investments in AI have also raised concerns over sustainability of these Chinese startups. Z.ai, for example, reported its revenue surged 132% to 724 million yuan ($107 million) last year, but net loss jumped 60% to 4.7 billion yuan ($694 million).

For now, Chinese AI models will likely continue to march ahead including in the U.S. “I would highly recommend anyone doing any serious AI load to at least evaluate it,” said Krikorian.

___

O’Brien reported from Providence, Rhode Island. AP Business Writer Kelvin Chan in Toronto contributed to this report.

Cheaper, open and intelligent: Chinese AI models gain ground, as they make inroads in the US

In light of growing environmental issues and the potential effects of AI data centers on the workforce, Jobs for Justice East Tennessee is taking the initiative to facilitate discussions among residents, as there currently isn't an established platform for public dialogue. Officials are addressing community concerns by responding to inquiries about these developments.

Chuck Corra, co-founder of AppPodlatcha, emphasizes the necessity of keeping the public well-informed about the implications of AI data centers, especially regarding their influence on local employment. "Understanding the scope of these projects and their repercussions on jobs is crucial," Corra remarked.

Event organizers note that while the recent one-year moratorium on AI data centers marks a positive step, its protection only extends to the unincorporated areas of the county. 

Regarding ongoing projects, Nathan Shew from the Tri-State Musicians Union expressed optimism about converting the old Hamilton County Jail into a music studio. However, he cautioned that if the AI component is not carefully managed, it could ultimately be detrimental. "If we end up with a large generative data center simply churning out art, it won’t contribute anything meaningful," Shew explained.

The impact of AI data centers extends to everyone, as highlighted by Sandy McRae, the president of the League of Women Voters. She pointed out that the implications of AI reach all job sectors in the region, noting that it's imperative for voters to stay informed about the decision-making processes behind these projects.

"The representatives are accountable to us; they manage our funds, and we deserve transparency regarding what we're getting in return for our investment," McRae stated.

AI data center workshop seeks to address inquiries from Hamilton county residents

Prompt: A bright community center bathed in warm natural light, adult residents seated around sleek server racks with soft white rack lights, an expert at the front pointing to a large screen titled “Hamilton County AI Data Center Workshop,” attendees with notepads raise their hands, a banner overhead reads “Hamilton County Workshop,” potted plants by the windows, wooden floors conveying a welcoming, informative atmosphere.

AI data center workshop aims to answer Hamilton county residents questions - WTVC

- Jobs for Justice East Tennessee addresses local concerns about AI data centers and workforce impacts.  
- A one-year moratorium on AI centers only protects unincorporated county areas, raising concerns.  
- Community leaders emphasize the need for public awareness and informed decision-making on AI projects.  

From environmental concerns to how AI data centers could impact the local workforce, officials from Jobs for Justice East Tennessee say with no current way for the public to have a formal discussion about the topic, they’re doing it themselves by answering an questions residents may have.

"It's really important that people are fully informed about what they are, the implications, and most importantly, with this discussion today, what the impact on the local jobs are going to be, " Cofounder of AppPodlatcha Chuck Corra says.

According to event officials, the recently passed one year moratorium on AI data centers is a step in the right direction, but it only protects the unincorporated parts of the county.

Photo: WTVC.

As for projects currently underway, Nathan Shew with the Tri-State Musicians Union says the plan to turn the old Hamilton County Jail into a music studio could be beneficial, but if the AI portion of it isn’t checked, it could do more harm than good.

...But if we just have a big old generative data center that's creating that artwork, that doesn't do anything. Shew says.

But how do AI data centers like the purposed one for the old Hamilton County Jail affect you?

AI impacts everyone. I mean, it impacts all jobs in this area. Sandy McRae League of Women Voters President says.

Sandy McRae with the League of women voter of Chattanooga says AI Data center is a complex issue, but its up to the voters be informed on how decisions are being made.

The representatives do work for us, and they are spending our money, and we need to know what we're getting for for the investment. McRae says.

AI data center workshop aims to answer Hamilton county residents questions

Your choices regarding privacy are important to us. By selecting "Accept all," you agree to share technical identifiers along with your browsing and search information. This data will assist us in analytics, providing tailored advertising and content, measuring ad performance, and conducting audience research and development of services. If you opt for '', our partners, including 250 members of the IAB Transparency & Consent Framework, will also have the ability to store or access information on your device, which includes the utilization of cookies, as well as gathering precise geolocation and other personal information for similar purposes.

If you prefer that we and our partners do not use cookies and personal data for these additional functions, please click "Reject all."

Should you wish to tailor your preferences, simply click on "Manage privacy settings."

You retain the right to revoke your consent or modify your preferences at any time by navigating to the "Privacy and Cookie settings" or "Privacy dashboard" links available on our platforms. For additional information on how we handle your personal data, please refer to our Privacy Policy and Cookie Policy.

A sunlit home office with an adult seated at a desk, adjusting oversized mechanical levers labeled “Share Data” and “Keep Private” on a console shaped like a smartphone, digital bits and lock icons fluttering in the air.

Quote of the day by Vladimir Putin on AI: 'Whoever becomes the leader in this sphere will become the ruler... - Yahoo

- Users can choose to accept or reject cookies for personal data usage.  
- Accepting cookies allows for personalized content and targeted advertising analytics.  
- Privacy settings can be managed or modified at any time by users.  

A Seattle resident named Lisa Booth, who has been battling stage 4 metastatic breast cancer for over 11 years, has partnered with fellow cancer survivor Steve Brown to create an innovative AI tool known as CureWise. This platform is specifically designed to support cancer patients in managing their medical records, understanding their diagnoses, exploring treatment options, and potentially discovering clinical trials that could be life-saving.

The inception of CureWise took place in February 2025, and since then, it has proven effective in assisting Booth, Brown, and other users in navigating the complexities of cancer treatment. The tool emerged from their own experiences with cancer, addressing the challenges they faced in tracking treatments and care.

Booth, having lived with cancer for more than a decade, often used spreadsheets and various technological tools to monitor new therapies and clinical studies. This experience sparked her desire for a more streamlined method to process vital information regarding her health journey.

CureWise employs AI to analyze a patient’s diagnosis, medical history, and overall health, aiming to connect them with appropriate treatments and clinical trials more efficiently. It’s important to note that while CureWise is not a cure for cancer, it significantly aids in uncovering treatment possibilities.

As the vice president of operations for CureWise, Booth utilized the tool during its prototype phase last summer, particularly when her cancer returned, providing her with crucial information about available trials.

“I’m not joking; this has been the most enjoyable experience in both my professional and personal life,” Booth shared, reflecting on the product’s development.

Co-founder Steve Brown, who was diagnosed with blood cancer after being displaced by the Palisades fire in January 2025, leveraged the AI tool to educate himself about his condition. “I directed this incredible AI tool at my own medical records, which fundamentally changed everything regarding the uncertainty of my illness,” he remarked, having learned about a rare mutation through the platform.

Regina Brown, another cancer survivor, recounted her experience with CureWise in an interview with KIRO 7. After an incident in Olympia last October led to the discovery of a growth around her thyroid, she faced a rapid cancer diagnosis months later. With her husband battling stage 4 lung cancer, she felt the weight of having to inform their daughters about her diagnosis. 

In stark contrast to her husband's lengthy research efforts over a decade, Brown utilized CureWise to hasten her treatment process, stating, “From my diagnosis on June 1st, I already had surgery by July 13th.” Her husband is also now using the tool to locate targeted therapies in far less time than it took previously.

The developers of CureWise have incorporated strict privacy protocols to safeguard patient information. Booth assured users that all personal data is de-identified, securely stored, and not employed to train large language models. “Users have complete control over their data and can remove it at any time,” she emphasized, underscoring the platform's commitment to user privacy.

Expressing profound appreciation for the tool, Regina Brown described it as transformative for her and her family, feeling it played a crucial role in their journey. “It was an answer, an answer to prayer,” she said, highlighting the peace of mind CureWise provided her in managing her cancer. This newfound empowerment enabled her to speak with confidence about her diagnosis to her daughters, attributing her upcoming post-surgery appointment to the support she received from the AI platform.

Cancer survivors from WA with tech expertise create AI platform to assist fellow cancer patients.

Two adult cancer survivors—one woman, one man—in a bright, sunlit modern office. Warm morning light streams through large windows as they stand side by side, one pointing at a large touchscreen showing an AI-powered cancer support dashboard with a gold awareness ribbon icon and friendly patient charts. Behind them a whiteboard features a hand-drawn outline of Western Australia, hinting at their roots, while potted plants and laptops on the desk underscore a hopeful, collaborative spirit.

Cancer survivors with WA roots and tech skills develop AI platform to help other cancer patients - KIRO 7 News Seattle

- Lisa Booth and Steve Brown developed CureWise, an AI tool for managing cancer treatment.  
- CureWise helps patients navigate medical records and find clinical trials for faster treatment.  
- The tool ensures patient privacy with de-identified data and user control over information.  

A Seattle woman living with stage 4 metastatic breast cancer for 11 years, Lisa Booth, has co-developed an artificial intelligence tool called CureWise.

The tool, created with fellow cancer survivor Steve Brown, who is originally from Wenatchee, aims to help countless cancer patients manage their medical records, cancer diagnoses, treatment and maybe even help find clinical trials that could save lives.

Development of CureWise, according to Brown and Booth, began in February 2025, and it has already demonstrated success in aiding its creators and other patients in navigating their cancer journeys.

CureWise was born from Booth’s personal struggle as well as Brown’s.

For Booth, living with cancer for more than a decade, she had experience tracking new treatments and clinical trials using spreadsheets and other tech tools. She was doing all this while living with cancer and still believed there had to be a simpler way to process all the information.

The AI platform she helped design actually can help learn about a patient’s diagnosis, their overall medical conditions and history, and maybe find specific treatments or clinical trials faster and more effectively that could help them.

While developers emphasize CureWise is absolutely not a cure for cancer, it can help patients find treatment options.

Both Booth and Brown used the tool they built to navigate their own illnesses; they admit the builders became patients also.

Lisa Booth, who serves as vice president of operations for CureWise, says she applied her medical records to the tool during its prototyping stage last summer.

She used it to understand available trials when her breast cancer recurred.

“I’m quite serious, I’ve never had this much fun in my entire professional and personal life,” Booth said about the development process.

Steve Brown, co-founder of CureWise, was diagnosed with blood cancer after being displaced from his Southern California home by the Palisades fire in January 2025.

He used the AI tool to educate himself about his condition.

“I kind of took this amazing tool of AI and pointed it at my own medical record,” Brown said, explaining that gaining visibility into his condition “changes everything” by addressing the “scariest thing, the unknown.”

Through the tool, Brown learned he had a rare mutation.

Another cancer survivor, Regina Brown, also spoke to KIRO 7 about her journey with cancer and the AI platform. She credits CureWise with significantly accelerating her treatment path.

After being assaulted in Olympia last October, a visit to the emergency room revealed a growth in the area around her thyroid. Months later, she found out the growth had doubled, and in June, testing revealed an unexpected cancer diagnosis.

“My husband has stage 4 lung cancer so the fact of sitting our daughters down once again and telling them now their other parent has cancer was very daunting,” Brown said.

She contrasted her rapid treatment with her husband’s decade-long experience of “old-fashioned” research.

With CureWise, Brown said, “I went from diagnosis date of June 1st, I already had surgery by July 13th.” Her husband also began using the tool to find targeted therapy in months, rather than years.

The developers say they have implemented safeguards to protect patient privacy when using CureWise.

Lisa Booth confirmed that patient data is de-identified, stored securely, and not used to train any large language models.

“You have absolute control. You can remove your data at any point in time,” Booth stated, highlighting user autonomy.

Regina Brown expressed deep gratitude for the AI tool, describing it as life-changing for her and her family. She believes CureWise helped her, definitely helped her husband, and absolutely did save her family.

She teared up, saying, “It was an answer, it was an answer to prayer, it was just an answer.”

She admitted that finding peace of mind in her ability to manage her cancer with the tool’s assistance was transformative.

It also gave her confidence to talk to her daughters about her diagnosis, and she felt empowered with all the information she received through using the AI platform.

She says she is scheduled for a post-operation appointment next week, an outcome she attributes to the assistance of the AI tool.

©2026 Cox Media Group

Cancer survivors with WA roots and tech skills develop AI platform to help other cancer patients

Oklahoma has introduced an innovative artificial intelligence platform designed to enhance the efficiency of state agencies and elevate customer service standards.

The Office of Management and Enterprise Services (OMES) has rolled out BEACON, a comprehensive state platform that establishes a consistent and secure approach to utilizing AI. State officials emphasized that this platform comes equipped with robust security measures, governance structures, and data standards aimed at minimizing redundancy among agencies while providing a unified framework for creating AI-driven applications.

“Oklahoma aspires to rank among the top ten states for business, which means leveraging the most advanced tools at our disposal,” stated Governor Kevin Stitt. “We are empowering our agencies to optimize their operations, deliver quicker outcomes, and serve the residents of Oklahoma more efficiently.”

The launch of BEACON follows the recent appointment of Tai Phan as Oklahoma’s inaugural Chief AI and Technology Officer, whose role is to guide the responsible and secure implementation of AI throughout government operations.

The initial offerings developed through BEACON are referred to as “lighthouse applications,” designed to demonstrate how AI can tackle governmental challenges and establish reusable models for future initiatives. Currently, two lighthouse applications are in beta testing: the Administrative Rules Modernization tool and the Procurement Guidance Assistant.

The Administrative Rules Modernization tool is crafted to assist lawmakers and state employees in assessing administrative rules, pinpointing opportunities for regulatory simplification, and alleviating unnecessary burdens on both citizens and businesses.

“We have invested significant hours in reviewing administrative regulations to ensure alignment with the state Legislature's intent and the needs of Oklahomans,” commented Senator Micheal Bergstrom, R-Adair. “This tool will enhance our effectiveness in this crucial work and uphold government accountability.”

On the other hand, the Procurement Guidance Assistant is designed to offer state workers real-time advice on purchasing policies and procedures, aiming to streamline decision-making processes while ensuring consistency and compliance.

“Our agencies provide essential services daily, and procurement is vital to that mission,” explained Amanda Otis, the state purchasing director. “This technology will enable teams to operate more efficiently while adhering to regulations, allowing them to concentrate on serving the people of Oklahoma.”

OMES highlighted that these tools are designed to support employees in their day-to-day tasks, reducing the time spent searching for information and enhancing precision, thereby assisting agencies in better serving the citizens of Oklahoma.

Oklahoma introduces a new AI platform to enhance state agencies.

A sunlit modern government office in the Oklahoma State Capitol, with a single Oklahoma flag billowing outside the window. At a polished wooden desk floats a translucent AI dashboard showing vibrant service icons and data charts—“Licensing,” “Permits,” “Health Services”—as a suited official interacts with it, symbolizing the new AI platform streamlining state agencies.

Oklahoma launches new AI platform to improve state agencies - KOKH

- Oklahoma launched BEACON, a new AI platform for state agencies to enhance efficiency.  
- The platform includes security, governance, and data standards to reduce agency duplication.  
- Two "lighthouse applications" are in beta testing to address government challenges effectively.  

Oklahoma has launched a new enterprise artificial intelligence platform aimed at helping state agencies work more efficiently and improve customer service.

The Office of Management and Enterprise Services announced the rollout of BEACON, a statewide platform designed to provide a shared foundation for using AI “safely, consistently and effectively.” State officials said the platform includes built-in security, governance and data standards intended to reduce duplication across agencies and give them a common framework for developing AI-powered tools.

“Oklahoma is committed to being a top ten state for business, and that means using the best tools available,” Gov. Kevin Stitt said. “We are equipping our agencies to work smarter, deliver results faster and serve Oklahomans more effectively.”

The launch follows the appointment of Oklahoma’s first Chief AI and Technology Officer, Tai Phan, who is leading efforts to implement AI “in a secure, responsible and practical way across government.”

The first tools built through BEACON are described as “lighthouse applications,” projects intended to show how AI can address government challenges while creating models that can be reused for future work. Two lighthouse applications are currently in beta testing: an Administrative Rules Modernization tool and a Procurement Guidance Assistant.

The Administrative Rules Modernization tool is designed to help policymakers and state employees review administrative rules, identify opportunities to simplify regulations and support efforts to reduce unnecessary burdens on citizens and businesses.

“We’ve spent countless hours reviewing administrative rules to ensure they reflect the intent of the state Legislature and serve the people of Oklahoma,” said Sen. Micheal Bergstrom, R-Adair. “This tool will help us do that work more effectively and ensure government stays accountable.”

The Procurement Guidance Assistant is intended to provide state employees with real-time guidance on purchasing policies and procedures, with the goal of helping agencies make decisions more efficiently while improving consistency and compliance.

“Our agencies deliver critical services every day, and procurement plays a key role in that work,” said Amanda Otis, state purchasing director. “This technology will help teams move more quickly while still following the rules, so they can stay focused on serving Oklahomans.”

OMES said the tools are intended to support employees in daily work by reducing time spent searching for information and improving accuracy, with the broader goal of helping agencies better serve Oklahomans.

Oklahoma launches new AI platform to improve state agencies

When Joe Spector's corgi, Eddie, ingested four M&Ms, the co-founder of Hims faced a frantic challenge finding an emergency veterinarian to help. A clinic directed him to a pet poison hotline, while another option was over an hour away with a wait time of four hours. Ultimately, the vet bill totaled a staggering $2,000.

This distressing experience in 2021 led Spector to realize, "it doesn’t have to be this way," as he recounted in an interview with Fortune.

This revelation prompted the creation of Dutch, a veterinary telehealth platform launched by Spector on July 1, 2021, following his success with Hims & Hers, a direct-to-consumer telehealth company recognized for addressing issues like hair loss and sexual health. Both businesses have since expanded significantly to encompass a wider range of health matters.

Dutch recently celebrated its fifth anniversary, providing members with telehealth services that allow video consultations with veterinarians for a subscription fee of $100 annually, which covers up to five pets. This fee is comparable to the price of a single visit to a conventional veterinary practice. With operations now stretching across 35 states—up from just eight at its inception—Dutch is benefiting from a more favorable regulatory environment for veterinary telemedicine. The platform has facilitated millions of consultations and boasts over 100,000 active members, as reported by the company. In early 2022, Dutch secured $20 million in funding to further enhance its platform, with the investment led by Forerunner Ventures.

While the company did not disclose its current revenue or valuation, Spector noted significant revenue growth, with an increase of 5.5 times from 2022 to 2023, followed by a 2.2 times rise from 2023 to 2024, and a further 1.8 times boost anticipated between 2024 and 2025.

In a strategic shift to invigorate the business, Spector made a bold decision last year to terminate the services of an outside marketing agency that had been costing the company around $50,000 monthly for just five creative pieces, which averaged $10,000 each.

"This approach was both costly and restrictive, especially since marketing new products is all about testing what resonates," Spector explained.

He then assembled an in-house marketing team that utilizes AI tools to generate diverse content types, including static ads and fully animated commercials. This shift has allowed Dutch to produce about 50 pieces of content monthly—a tenfold increase. The internal operation now costs approximately half a million dollars less than the previous agency arrangement, facilitating more effective testing to engage pet owners, many of whom may be unaware of telehealth options for veterinary care.

"We need to enhance trust and educate consumers on how this service functions and its benefits," Spector emphasized.

Dutch's advertising strategy has evolved, with its investment on Meta increasing from 5% of ad spending to around 25%. The efficiency of the new team has already yielded a 20% reduction in customer acquisition costs from January 2026 to April and an impressive 20% audience growth per month between February and March, according to company data. AI is also being harnessed for measurement tasks such as A/B testing, brand-lift studies, and refining the search strategy.

Spector is quick to credit his marketing team for these advancements, stating, "It’s not just the software—it's the people behind it that drive success."

The aim for Spector is to elevate Dutch’s market presence. Initially, Dutch capitalized on immediate needs through bottom-funnel Google searches targeting pet owners in crisis. The next step is to enhance brand awareness so pet owners consider Dutch before emergencies arise.

To further this initiative, Dutch enlisted former Olympic figure skater Tara Lipinski as a brand ambassador. Lipinski, who became a customer after losing her home in the Los Angeles fires and needing veterinary care for her dog, connects emotionally with the brand's mission.

"I want our customers to feel a sense of relief,” Spector concluded.

A cofounder of Hims dismissed his entire marketing team for AI, and now his tele-vet is experiencing 20% monthly growth.

A bright home office flooded with warm morning light: in the shadowed corner, two empty desks and chairs pushed back hint at the departed marketing team. Center stage, the Hims logo is emblazoned on a sleek laptop showing a tele-vet video call—an enthusiastic veterinarian examining a happy dog while a bold “+20%” growth arrow climbs on a side monitor.

A Hims cofounder fired his entire marketing team for AI. Now his 24/7 online vet service is growing 20% a month - Fortune

- Joe Spector founded Dutch after struggling to find urgent vet care for his corgi.  
- Dutch offers video vet visits for $100 annually, expanding to 35 states since launch.  
- The company shifted to in-house marketing, increasing content production and reducing costs significantly.  

When Joe Spector’s corgi, Eddie, ate four M&Ms, the Hims cofounder spent hours trying to find an emergency vet who could see him. One office pointed him to a pet poison hotline, and another was an hour away and had a four-hour wait. The bill came out to $2,000.

Spector thought at the time “this does not have to be that way,” he told Fortune.

That stressful scramble in 2021 sparked the idea for Dutch, the veterinary telehealth company Spector launched on July 1 of that year after helping build and take public Hims & Hers, the direct-to-consumer telehealth brand best known for treating hair loss and sexual wellness. Those companies have since grown exponentially to cover an array of health concerns.

Dutch, which just marked its fifth anniversary, offers members video visits with veterinarians for a flat $100 a year covering up to five pets. That amounts to roughly the cost of a single in-person visit at a regular veterinary office. The company now operates in 35 states, up from eight at launch, as more states loosen restrictions on veterinary telemedicine. The company has logged millions of interactions and now counts more than 100,000 active members, according to the company. Dutch raised $20 million in early 2022 to scale the platform, in a round led by Forerunner Ventures.

The company declined to share its current revenue and valuation, but its revenue grew 5.5x between 2022 and 2023, 2.2x between 2023 and 2024, and 1.8x between 2024 and 2025.

Firing his agency to grow the business

Although Dutch is clearly charting its growth path now, it was a bit stagnant for a while. Now Spector is making a bigger bet on how the company markets itself.

Late last year, Spector said, Dutch was paying an outside agency roughly $50,000 a month to produce five pieces of creative material, or about $10,000 apiece.

“It’s both expensive and it’s just limited because so much of marketing, especially with a new product, is just testing to see what’s resonating,” Spector said.

So he fired that agency. That drastic move is somewhat reminiscent of Bolt’s CEO telling Fortune at its Workplace Innovation Summit in May he had ditched his entire human resources team, although the executives had very different motivations for their moves.

In its place is a new in-house marketing team leaning heavily on AI tools to generate content, from static ads to full animated and live-action commercials. Now Dutch produces about 50 pieces of content a month internally, or a 10x jump, Spector said. That’s enough work to cost roughly half a million dollars a month through an agency, he said. Just as important as the savings, he said, is the ability to test what actually lands with pet owners who often don’t know virtual vet care exists.

“We have to get over the trust factor,” he said. “We have to educate consumers on how this even works, or like what’s the use case.”

Dutch’s presence on Meta has grown from about 5% of its ad spend to roughly one-quarter. Since the new team’s first full month, the company has cut its customer acquisition cost by 20%—comparing January 2026, before the team started, with April—and grew its audience 20% month over month from February to March, according to the company. Spector said Dutch is also using AI on the measurement side, for A/B testing, brand-lift studies, and its search strategy.

To be sure, Spector is quick to credit the people, not just the software.

“I shouldn’t take the credit—it’s the marketing folks,” he said. The team, he added, is what makes leaning into the tools possible.

For Spector, the goal is to move Dutch upstream. The company built its early growth on bottom-of-funnel Google search, reaching pet owners in the middle of an emergency, he said. Now he wants to grow brand awareness, so that owners think of Dutch before the crisis hits.

That’s partly why Dutch brought on former Olympic figure skater Tara Lipinski as a brand ambassador. Lipinski, Spector said, came to Dutch as a customer after losing her home in the Los Angeles fires and needing care for her dog.

“One of the feelings I want people to have is this sense of relief,” Spector said.

A Hims cofounder fired his entire marketing team for AI. Now his tele-vet is growing 20% a month

This week, a power line failure occurred just outside Washington, DC, which typically allows the electrical grid to recover within seconds. However, this incident took over 10 minutes to stabilize due to a simultaneous shutdown of more than 3 gigawatts worth of power from data centers.

Data collected by Ting Labs, a startup utilizing an IoT sensor network, highlighted that the voltage across the PJM grid surged from Northern Virginia to Chicago as a result of this event. While the incident did not lead to a blackout, it did cause noticeable flickering of lights throughout the area, illustrating the significant impact large data centers can have on the grid — a trend that experts anticipate will occur more frequently.

Northern Virginia, part of the PJM territory, boasts the highest concentration of data centers globally. Ricardo de Azevedo, CTO of ON.Energy, described the situation as indicative of broader issues, pointing to similar events becoming increasingly common.

This recent occurrence resembled a similar incident from two years prior, also affecting the PJM grid, raising concerns about the potential for more severe disruptions if data centers do not adopt strategies to handle such power outages effectively. The PJM Interconnection is the largest grid operator in the United States, managing the electrical supply for 67 million customers across states from New Jersey to Illinois.

During this week's event, data centers quickly transitioned to backup power systems, leading to a rapid drop in demand of around 3.1 gigawatts within 30 seconds, according to PJM data. Although the grid initially showed signs of recovery, further load reductions caused it to take an additional 11 minutes to stabilize. The data centers that disconnected accounted for roughly 3% of PJM’s total demand at that time, as reported by Reuters.

While 3% may seem minor, the grid must maintain a delicate balance between supply and demand. An imbalance can lead to voltage fluctuations. While minor deviations can be tolerated, substantial changes can activate grid failsafes, resulting in disconnections.

As disturbances from the failed power line reached Northern Virginia data centers, they swiftly shifted to backup power, exacerbating the initial load drop and contributing to voltage instability that made lights flicker. Ali Zain Banatwala, a senior market models specialist at the Independent Electricity System Operator, mentioned that the affected data centers acted almost instantaneously, disconnecting in unison to the voltage dip.

To mitigate similar issues in the future, Banatwala emphasized the need for a more structured protocol that allows neighboring data centers to disconnect or reconnect in a staggered manner. This change could enable grid operators to devise more effective contingency plans.

Alternatively, there’s potential for reengineering data centers to better withstand such disruptions. ON.Energy has been innovating a new uninterruptible power supply designed for entire data center campuses. Their system integrates batteries and advanced power conversion technology, allowing the grid to perceive a stable load without fluctuations from the various components operating within the data center. This setup enables operators to adjust computing tasks efficiently while minimizing the impact on the grid.

Importantly, this system allows data centers to manage power variations: excess energy can be stored in batteries during peak demand, while surplus can be dispatched immediately if grid supply dips, preventing issues like those experienced during the recent outage.

ON.Energy is currently in the process of installing 3 gigawatts of its systems across four different data center locations, as per de Azevedo’s statement.

Grid operators are becoming increasingly aware of the challenges posed by large power loads such as data centers. For instance, ERCOT is now mandating that these facilities be capable of "riding through" power disruptions.

However, urgency remains. The recent mass disconnection event was double the scale of a previous incident in 2024, where 60 data centers together disconnected, pulling 1.5 gigawatts from the grid. At that time, data centers represented about 6% of PJM's total load, a figure projected to rise to 24% by 2040. Without timely solutions, the implications for grid stability could worsen significantly.

A downed power line highlighted an escalating issue with AI data centers. Here’s the solution.

Dark, stormy night outside a glass-walled AI data center. A wooden utility pole has snapped, live wires sparking orange-red on wet pavement. Inside the building, rows of server racks flicker between dim red warning lights and blackout. A lone technician in a high-visibility vest kneels under the beam of a handheld lantern, inspecting an emergency generator. Tense, dramatic shadows and warm spark glows highlight the vulnerability—and hint at the fixes needed.

One fallen power line exposed a growing AI data center problem. Here’s how to fix it. - TechCrunch

- A downed power line caused 3 gigawatts of data centers to disconnect from the grid.  
- The PJM grid experienced voltage spikes, causing flickering lights without leading to a blackout.  
- Experts warn that data centers need improvements to handle power supply disruptions effectively.  

A power line went down outside of Washington, DC, this week. Normally, the grid would only need a few seconds to recover from such an event. But this one took more than 10 minutes because more than 3 gigawatts of data centers stopped drawing power nearly simultaneously.

The event caused voltage across the PJM grid to spike from Northern Virginia to Chicago, according to data collected by Ting Labs, a startup that runs an IoT sensor network out of people’s electrical sockets.

The event didn’t cause a blackout, but it did cause lights across the region to flicker. The incident demonstrated the effect that data centers can have on the grid — an outcome that experts believe will become more frequent.

Northern Virginia, which is in PJM’s territory, is home to the highest concentration of data centers in the world.

“It’s the canary in the coal mine,” Ricardo de Azevedo, CTO at ON.Energy, told TechCrunch. These sorts of events involving large loads like data centers are “happening more and more,” he added.

The event echoes one that happened two years ago, also on PJM’s grid, and it could foreshadow larger events if data centers aren’t built to more elegantly handle disruptions to power supplies. The PJM Interconnection manages grids from New Jersey to Illinois and serves 67 million customers, making it the largest grid operator in the United States.

When the power line went down this week, it triggered data centers to switch to backup power, and about 3.1 gigawatts of load vanished in about 30 seconds, according to PJM data. The grid appeared to recover somewhat, but a short time later additional loads dropped off. At its peak, PJM’s grid had an extra 3.49 gigawatts of electricity on it. It took another 11 minutes before it stabilized. The disconnected data centers represented around 3% of total demand on PJM at the time, according to Reuters.

A few percent may not sound like much, but the electrical grid needs to operate in a state of near-perfect balance, with supply and demand closely matched. If they don’t, voltages can sag or spike. The grid and devices connected to it can tolerate small fluctuations, but if those fluctuations grow too large, they’ll trigger failsafes within the grid or within individual facilities, causing them to disconnect.

When data centers in Northern Virginia sensed the fluctuation caused by the failed power line, they switched to backup power, which removed their load from the grid. As more data centers made the switch, they removed greater amounts of load from the grid. What started as a relatively small drop in supply became an even larger drop in demand, sending supply surging and causing light bulbs to flicker.

Most data centers make decisions in a split second, and those that disconnected this week appear to be no different. When the voltage dip reached them, they all decided to disconnect within a few seconds of each other, Ali Zain Banatwala, senior market models specialist at the Independent Electricity System Operator, told TechCrunch.

“We need to figure a way for these loads that are located next to each other to sequentially either disconnect or reconnect,” he said. A more orderly process would allow grid operators to develop more robust procedures in advance.

Alternatively, data centers could be built to absorb disruptions and not turn their backs to them. One startup, ON.Energy, has been working on a product to help data centers — and the grid — ride through events like the one that occurred this week.

The company has developed an uninterruptible power supply for an entire data center campus, covering not just servers but also chillers and other equipment. The company essentially hides the data center behind a bank of batteries connected to sophisticated power conversion equipment. All the grid “sees” is one consistent, well-behaved load rather than the peaks and valleys from each individual part of the data center. ON.Energy’s system allows data centers to ramp computing workloads up and down, including AI training, without bothering the grid.

Perhaps more important, it also means that data centers can absorb power fluctuations from the grid. Rather than disconnecting from the grid, ON.Energy’s system can use any extra power to charge its batteries, and if the flow dips, the system can dispatch power to servers. Plus, it can follow the grid’s lead within milliseconds, preventing sags or surges like the ones that caused this week’s problem for PJM.

ON.Energy is currently installing a total of 3 gigawatts worth of its systems at four different data center campuses, de Azevedo said.

Grid managers have also woken up to the problem.

ERCOT, for example, is going to require large loads like data centers to “ride through” disruptions, de Azevedo said.

The clock is ticking, though. The mass disconnection this week was twice as large as a similar event in 2024, when 60 data centers simultaneously disconnected, pulling 1.5 gigawatts of load from the grid. Back then, data centers accounted for about 6% of PJM’s load, according to Synapse Energy Economics. By 2040, they are expected to make up 24%. If the problem isn’t addressed soon, things could get a lot worse.

One fallen power line exposed a growing AI data center problem. Here’s how to fix it.

Last March, Anthropic, an innovative AI firm known for its chatbot Claude, released a study examining the implications of artificial intelligence on the labor market. This analysis was intended to address concerns that advanced AI could dramatically alter human life and diminish the need for human workers.

In May of the previous year, Anthropic's co-founder Dario Amodei expressed the belief that AI might eliminate 50% of entry-level positions within five years. By January, he indicated that AI could serve as a “general labor substitute for humans,” and in June, he highlighted the potential dangers of an economy trapped in a cycle of rapid growth and increasing inequality.

Despite these dire predictions, Anthropic's report indicates that the impact of AI has not met these expectations. The findings reveal that since late 2022, there has been no significant rise in unemployment among workers who are most vulnerable to AI automation. The current implementation of AI technology is still only a small fraction of what is theoretically possible. For example, Claude currently handles merely 33% of all tasks in the computer and math domain, even though it could potentially manage nearly 100%.

While there is a surge in investment in data centers, productivity gains have not matched the high hopes set by early tech predictions. In fact, labor productivity growth has been slower in the AI era’s initial three years compared to the tech boom of the 1990s.

OpenAI’s Sam Altman, one of the leading figures in AI, has recently voiced skepticism regarding fears of mass job loss. He remarked in May that the anticipated “jobs apocalypse” may not materialize as some industry advocates claim. MIT economist David Autor echoed this sentiment, noting that many have observed slower changes in the job market than anticipated.

This shift in perspective has sparked discussions around a more nuanced understanding of how automation interacts with human employment. Analysts are now questioning whether AI will truly bring about the profound changes it promised. Since peaking in early June, the tech-heavy Nasdaq index, largely driven by AI stocks, has seen an approximately 8% decline.

Critics of the idea that AI can seamlessly transform industries point to the "O-ring" theory, derived from the Challenger space shuttle disaster caused by a faulty O-ring. The argument suggests that until AI can perform every task flawlessly, the value of remaining tasks will increase. This could either elevate the worth of higher-skilled workers or create new opportunities for lower-skilled labor as AI takes on more complex duties.

A recent study highlighted that while AI substitutes tasks at a significant rate, the overall employment impacts have been modest. Reduced demand in affected roles can be balanced by increased labor demand driven by productivity gains in companies that adopt AI.

However, the landscape could shift, as Jed Kolko emphasizes that research into AI’s effects on the labor market is still in its early stages. Amodei's predictive window remains open for nearly four more years, suggesting that significant changes could take time. The Federal Reserve notes a rapid rise in AI adoption among businesses.

Autor also points out that AI is continuing to improve, with progress showing no signs of stagnation. The potential for a dystopian or utopian AI future remains alive, particularly for those in control of the technology. Daron Acemoglu, a Nobel laureate economist, observed that insiders remain optimistic, with many still anticipating the arrival of artificial general intelligence soon. Elon Musk continues to promote the vision of a world where AI and robots enable universal income, allowing work to become optional.

Historically, economists have noted the discrepancy between technological advancements and productivity statistics. Robert Solow remarked, "You can see the computer age everywhere but in the productivity statistics." However, it eventually took time for these technologies to be integrated in ways that reflected their impact on productivity.

Yet, challenges loom on the horizon for AI. It may not eliminate all human labor, nor live up to its promises of vast economic benefits at an acceptable cost. Public sentiment has turned against the construction of AI data centers, with 70% of Americans opposing them, partially due to concerns over rising energy costs and fears of societal upheaval.

Numerous issues remain unresolved; despite significant progress in AI, questions persist about its ability to meet the diverse needs of a modern economy. Autor highlights that not all problems can be solved computationally, as AI excels at replicating language but struggles to apply it accurately to reality. Its limitations continue to result in significant errors.

Additionally, the economic feasibility of AI remains in question. Even if AI could one day address all societal challenges, the financial implications of widespread AI deployment appear daunting, with discussions around investing 20% to 40% of GDP in AI data centers becoming increasingly common. The International Energy Agency forecasts that the energy demands from data centers will more than double to around 945 terawatt-hours by 2030, surpassing the entire energy consumption of Japan.

The financial outlook for AI is tenuous, as rapid model obsolescence means that investment may not yield returns. Acemoglu warns that companies creating AI systems could face severe financial losses, amounting to hundreds of billions annually.

Altman's newfound caution could be seen as a strategic maneuver, suggesting that equating the AI revolution with mass unemployment may be politically unwise. However, broader skepticism about AI’s capabilities is more than just a marketing angle. The ambitious promises surrounding AI’s transformative powers face significant challenges as the possibility emerges that it may not deliver what society is willing to pay for.

The impending AI job crisis is unlikely to occur in the near future.

A bright, sunlit open-plan office where diverse professionals sit at sleek desks collaborating with a friendly humanoid robot that’s handing a document to a smiling engineer. Potted plants and floor-to-ceiling windows fill the space with warm natural light, conveying harmony between human workers and AI rather than replacement.

The AI jobs apocalypse probably isn’t coming anytime soon - The Guardian

- Anthropic's report indicates no significant unemployment increase from AI exposure since 2022.  
- Productivity gains from AI have been slower than expected, resembling the IT boom's early years.  
- Public opinion on AI has soured, with 70% opposing local AI datacenter construction.  

In March, Anthropic, the cutting-edge artificial intelligence business that gave us the chatbot Claude, published an analysis on the impact of AI on employment, to help us assess the claim that intelligent robots were about to redefine human existence, ending demand for human labor.

Last year in May, Anthropic’s co-founder, Dario Amodei, claimed AI could wipe out half of all entry-level jobs in one to five years. Last January, he told us AI would probably become a “general labor substitute for humans”. In June he said we risk “a world where the economic trade-off dial is stuck on the hypergrowth, hyper-inequality setting”.

And yet, Anthropic’s report suggests that, so far, AI’s impact has fallen short of expectations: “We find no systematic increase in unemployment for highly exposed workers since late 2022,” the report stated. Deployment of the technology “remains a fraction of what’s feasible”. Claude covers just 33% of all tasks in the computer and math category whereas theoretically it could take over nearly 100% of them.

Spending on datacenters is going through the roof, for sure, but productivity has not been experiencing the galloping gains that the technorati’s epochal prognostications lead one to expect. Labor productivity was, in fact, slower in the first three years of our AI era than during the information technology boom that began in the mid 1990s.

Even OpenAI’s Sam Altman, AI’s most public face, says he now doubts its job-killing potential. “I don’t think we’re going to have the kind ​of jobs apocalypse that some of the companies in our space advocate or talk about,” he said in May. As Massachusetts Institute of Technology economist David Autor noted: “A lot of people have noticed that the world is not changing as fast as they predicted.”

This has opened the public conversation to a less cataclysmic narrative of the evolution of the technology. The emerging new story not only puts more emphasis on the complexity of the relationship between automation and human work across history. It is also raising doubts about the very feasibility of the threatened AI transformation of the universe. The tech-heavy Nasdaq index, which had been propelled almost exclusively by the rise of AI-related stocks, has fallen about 8% since its peak in early June.

One strand of critique of the “AI-will-do-everything” story might be called the O-ring argument. It comes from the mid-flight explosion of the space shuttle Challenger 73 seconds into its flight on 28 January 1986. A lengthy investigation concluded that the demise of the multibillion dollar spaceship was caused by a rubber O-ring that didn’t work at low temperatures.

That cheap O-ring proved critical. The analogy suggests that as long as AI cannot perform every task perfectly, it will increase the value of the remaining tasks. Depending on which the AI takes over, it could increase the value of high skill workers who are relieved by AI of the low-end part of their job, or increase the opportunity of lower skilled workers by taking over the more expert tasks.

As one recent study noted: “despite strong substitution at the task level, overall employment effects are modest, as reduced demand in exposed occupations is offset by productivity-driven increases in labor demand at AI-adopting firms.”

Things could change. As Jed Kolko points out, research on the labor market impact of artificial intelligence is still in its infancy. There are almost four years to go in Amodei’s one-to-five year window. And maybe devastation hits in year six. According to the Federal Reserve, adoption of AI is expanding fast across businesses.

Moreover, Autor argues, AI is getting better. And its progress shows no sign that it will soon hit a ceiling. “Skepticism about the stochastic parrot is behind us,” he told me. The dystopian AI future – utopian, if you get to own and run the AI – is still on the cards.

“Insiders are as gung ho as ever,” noted Daron Acemoglu, the Nobel prize-winning economist. “They still believe artificial general intelligence is around the corner.” Indeed, Elon Musk has not budged from the dream that “AI+Robots will be able to do everything, resulting in universal high income. Work will be optional.”

One may recall the quip by Nobel prize-winning economist Robert Solow in the early years of the computer revolution: “You can see the computer age everywhere but in the productivity statistics.” It took another 10 years or so, as businesses reorganized around the new technology. But computers did eventually show up in the stats.

Clouds are nonetheless gathering on the AI horizon. It’s not just that AI may not end all human work. AI may not deliver on its promise of vast economic opportunity at a price that humanity is willing to pay.

The politics have decidedly soured on the project. Seven in 10 Americans oppose building AI datacenters in their area. While this has to do with their insatiable demand for energy, which drives up local electricity costs, AI’s unpopularity is no doubt related to the proposition that it will destroy society as we know it.

There are other bumps in the road. Despite its vast progress, big doubts remain on whether AI can do everything a modern economy needs. “Not everything is a computational problem,” notes Autor. AI is good at replicating language, but it cannot connect language to the reality around it. Despite its progress, it still makes plenty of critical mistakes.

And then there are the impossible economics. Even if AI could eventually solve all our problems, the solution looks expensive. How much of GDP are we willing to invest in AI datacenters, 20%? 30%? 40%? According to some estimates, that is where we are headed. The International Energy Agency estimates that power demand from datacenters will more than double by 2030 to about 945 terawatt-hours, more than the energy consumption of Japan.

The economics look more fragile considering how fast the investment in AI depreciates, as new models overtake those developed just a few months ago. Companies developing AI models “are never going to make money”, Acemoglu said. “They are losing hundreds of billions of dollars every year.”

One may discount Altman’s new modesty as a PR feint. Somebody may have told him that equating the AI revolution with mass joblessness was not smart politics. But misgivings about AI’s vaunted capabilities are more than a marketing twist. The grand, epochal promise may be in trouble. Maybe artificial intelligence can’t deliver at a price society is willing to pay.

Kevin O’Leary announces plans to downsize his Utah AI data center project following political backlash.

More from AI

Your options for privacy

Exploring Nadella's Most Challenging Year

Affordable, accessible, and smart: Chinese AI models are making progress in the US market.

AI data center workshop seeks to address inquiries from Hamilton county residents