The AI party is getting pricier, and investors are lowering expectations

This week's business and economy stories point in the same direction: AI is increasingly part of everyday work, but the bill is becoming clearer to everyone who has to pay it.

Money is harder to come by

The mood has shifted at the big AI companies. OpenAI has closed a first portion of a new round of about $11 billion at a valuation of $1.25 trillion. The round could end up at about $20 billion, against a target of up to $30 billion. That is below what the company originally wanted. According to investors, Anthropic wants to go public before Thanksgiving, but its valuation looks likely to come in lower than the $2 trillion some had hoped for. Tech stocks have fallen, and it is being felt all the way at the top.

The physical side is also getting more expensive to build. In the November 3 election, American voters restricted data center expansion in 38 of 39 decided ballot measures. The count is preliminary, but data centers have become a political issue that matters all the way up to governor races. Data centers are the large buildings full of computers that AI runs on, and without them growth grinds to a halt.

Workplaces put the tools to use

At the same time, AI is becoming a regular part of companies' daily operations. The cement group Cemex has built an HR agent in Microsoft Copilot Studio that about 800 employees in Central Mexico can query in Teams. The case comes from Microsoft itself, and the impact has not been quantified. Airbnb's AI assistant now resolves about 40 percent of support cases without a human employee, and an AI agent for the entire trip is planned for the first half of 2027.

As usage grows, so does the need for oversight. Anthropic has made Smart Reports generally available for Claude Enterprise. The reports sort a team's usage into workstreams and point out problems and waste. It is a sign that companies are starting to ask what they are actually getting for their money.

The raw material is not free

A third angle is about who owns the content AI learns from and answers with. Meta has postponed new commercial terms indefinitely after complaints from publishers, who fear the terms give Meta the right to pull content from their sites. According to unnamed legal experts, the revised wording is still broad. An advocate general at the EU Court of Justice also believes that chatbot providers can, under certain circumstances, be liable if their answers recognizably reproduce the form of a news article. The case concerns Google's Gemini and a Hungarian publisher. It is a non-binding opinion, not a ruling, but it shows the direction.

What you can do

If your workplace uses AI, look at what it is actually being used for before buying more. Vendors' own success stories, like the Cemex example, are good to have, but they are not proof. Also keep an eye on whether Anthropic's IPO plans and OpenAI's round end up lower than hoped. Over time, that could affect prices and terms for the tools you use every day.