Good morning. A quick read of the headlines finds artificial intelligence advancing seemingly unchecked across all fronts, swamping every process, every industry.

But for some within the maelstrom, namely those vendors selling AI's promise to the enterprise, it's a different story, as the WSJ Leadership Institute's Isabelle Bousquette reports.

"Vendors say big companies have become more cautious about what they buy. They're taking longer to evaluate solutions, involving more internal stakeholders from legal and finance teams, and placing more emphasis on the kind of financial returns they might get out of the investment. The breakneck pace of AI innovation—like recent updates around Anthropic's Claude—is also making potential customers wary of sales commitments."

The picture Isabelle paints is of a market tapping the brakes ever so slightly. Maybe that's no bad thing.

"Overall, we've gotten a lot more disciplined in making sure that we understand what the outcome of a specific purchase is, not just following the hype," said Kyle Chu, senior manager of Business Intelligence at phone-accessory maker PopSockets.

AI entered the enterprise on a wave of hype, corporate FOMO and the promise of pilots. Buyers now want the math and assurance that solutions work consistently across the business.

The pace of innovation, though, cuts both ways. Better AI should accelerate buying, but sign a three-year deal today and something dramatically better could ship in six months. Customers aren't wrong to be cautious, and vendors who want to close deals may need to rethink how they structure deals.

Kathy Kay, chief information officer at Principal Financial Group, said she thinks carefully about whether a given vendor will even be around or useful in a few years, "which could make it seem like the sales cycle is longer to a company calling on us."

The easy sales are over. Now comes the hard part: Showing up with results.

AI could be making America's housing shortage worse. Landowners and developers are discovering that selling parcels to data-center builders is far more lucrative than any other use. Local zoning often makes it faster and easier to build server farms than homes, the WSJ's Will Parker reports. Northern Virginia, the world's data-center capital, is also one of its most housing-starved regions. It is not alone.

In 2024, Stream Data Centers purchased and then knocked down an entire 55-home subdivision in Elk Grove Village, Ill., a data-center hub near Chicago, to build three data centers totaling 2.1 million square feet.

Compounding the problem, data-center developers are winning the competition for scarce materials and labor needed to build new housing.

Activist investor pushes iconic toilet-maker to make an AI play. FT reports that Palliser Capital in a note to Japan's Toto suggested they invest more in their advanced ceramics segment, source of a crucial ingredient for NAND memory chips. The segment also generates some 40% of Toto’s operating profit.

Toto is not the only non-tech Japanese company critical to the AI boom. Nittobo, a century-old Japanese textile company, makes the cloth-like material known as T-glass used in advanced chips, the WSJ reported earlier. And food company Ajinomoto makes a specialized film used in the underside layer of a chip alongside T-glass.

Hits keep coming from Anthropic. The AI startup on Tuesday launched Claude Sonnet 4.6, its second major AI model in under two weeks, CNBC reports. Anthropic said 4.6, which includes improvements in coding, design, data processing and computer use, will become the default model for free and Pro users on Claude and the Cowork tool. “The model certainly still lags behind the most skilled humans at using computers,” the company said in a blog post. “But the rate of progress is remarkable nonetheless.”

Meanwhile, the Pentagon is weighing labeling Anthropic a supply-chain risk, a designation typically reserved for foreign adversaries, the Journal reports. The standoff centers on Anthropic's refusal to allow its technology to be used for domestic surveillance and autonomous lethal operations. Currently Claude is the only large-language-model that can be used in classified settings.

$50 billion. The amount Microsoft said it intends to invest by the end of the decade to expand AI in the 'Global South,' Reuters reports. 

75,000. The number of homes by which the region known as “Data Center Alley” falls short, according to the Virginia Association of Realtors.

About 9%. The percentage of Nvidia's revenue that comes from Meta, which this week committed to using “millions” of Nvidia’s AI chips in a chips-and-networking deal, Bloomberg reports.

Mistral AI to make first acquisition. The French AI maker said infrastructure startup Koyeb will help its developers run and scale AI without infrastructure snags so they can focus primarily on writing code, WSJ reports.

ServiceNow CEO takes a stand. Bill McDermott is putting $3 million of his own money into company stock after shares fell nearly 30% this year, swept up in a broader software selloff, MarketWatch reports. Four other executives have also canceled automated stock-selling plans. 

Another quantum computing company goes public. Infleqtion made its debut Tuesday, raising over $550 million through a SPAC merger with Churchill Capital Corp X, Barron's reports. The startup aims to achieve a utility-scale, fault-tolerant quantum computer by 2028, but its ambitions extend beyond computing into atomic clocks and quantum sensing, including a NASA gravity sensor project.

Europe turns up heat on U.S. social media. The Spanish government is requesting that prosecutors investigate X, Meta Platforms and TikTok for potential crimes against minors linked with the dissemination of  AI-generated images, the WSJ reports.

AI advancements have made it easy for users to generate images. Elon Musk’s X has come under tight scrutiny in Europe lately after a public outcry over sexualized deepfake images created by its Grok AI software. Last month, the European Union launched an investigation and on Tuesday, Ireland’s data protection watchdog opened its own investigation.

AI-powered cameras on everything. As Meta’s bet on smartglasses pays off and rumors swirl around OpenAI’s own gizmo, Apple, the OG of smart devices, is focusing on three wearable devices of its own: Smartglasses, a smart pendant and AirPods with more AI, Bloomberg reports. VisionPro, it seems, can wait. All three will be built around Siri, the digital assistant still waiting for its big AI upgrade, and will be equipped with cameras.

Surveillance creep. Apple's smartglasses could draw comparisons to Meta's version, so sleek that users won't notice the tiny camera. That unobtrusiveness is the point, and the problem: a growing number of people aren't comfortable having their actions recorded by a stranger on the street, the New York Times reports. Meta also is looking to add facial recognition to its smartglasses, the NY Times reports.

Vice President JD Vance said Iran had failed to acknowledge core U.S. demands in talks here Tuesday, after which Washington said it had agreed to give Tehran two weeks to close the gaps between the sides. (WSJ)

The U.S. presented new seismic data Tuesday to buttress its recent allegation that China has secretly carried out low-yield nuclear tests, challenging Beijing’s insistence that it has scrupulously observed an international accord banning all nuclear detonations. (WSJ)

Mayor Zohran Mamdani proposed raising New York City property taxes by nearly 10%, having failed so far to persuade New York Gov. Kathy Hochul to increase taxes on corporations and the wealthy. (WSJ)

Nine skiers are missing after they were overwhelmed by an avalanche in Northern California, authorities said. The avalanche in Truckee, Calif., hit Tuesday morning, engulfing a group of backcountry skiers. (WSJ)

The WSJ CIO Journal Team is Steven Rosenbush, Isabelle Bousquette and Belle Lin.

The editor, Tom Loftus, can be reached at thomas.loftus@wsj.com.