What's up: AI meets aggressive accounting in Meta's new data center project; AI investors want more making it and less faking it; Google leapfrogs rivals with Gemini 3.

Good morning. As the third anniversary of OpenAI’s ChatGPT release approaches, it’s remarkable to consider the catalyzing effects. Win or lose, good or bad, for better or worse, that event has reshaped business, markets and the way people live. It has reframed the direction of technology and investment.

ChatGPT woke up rival model developers such as Google, DeepSeek, Meta and xAI, triggering an epic amount of spending on AI infrastructure and the Nvidia systems that power them. Investors are anxious, and given the scale of capital pouring into AI, who can blame them? Nonetheless, demand remains strong, measured by financial and nonfinancial metrics alike. Nvidia results continue to beat investors’ expectations, and the number of tokens processed by Google, a proxy for broad AI usage, is soaring.

Perhaps most telling, corporate leadership, including CIOs, CTOs, CEOs and their boards, continue to invest in AI in an aggressive but mostly disciplined way. We have seen evidence of that in story after story by the WSJ Leadership Institute. Generally speaking, that’s not because they are reckless. The fact is that the corporation, often stereotyped as stodgy and protective of the status quo, has a longer investment horizon than public investors who are focused on the next few quarters. In their approach to risk, their timeline is more closely in sync with the venture capital mindset. Early adopters in the enterprise including Walmart, Intuit and BNY also told the WSJ Leadership Institute in recent weeks that those investments are having a meaningful positive impact on the company.

Looking ahead, we can see the beginning of a new AI-driven model for commerce, in which companies have a presence in ChatGPT itself, a sign that AI has avenues for much broader expansion in the world.

Unquestionably, air is coming out of the AI bubble, and that’s almost certainly a good thing. Investors are right to be skeptical and apply discipline to the market (see more below).

It serves no one’s interest for valuations to run mindlessly ahead of current reality. But the fact that the entire AI ecosystem hasn’t collapsed under that pressure, at least so far, is a sign of resilience.

About last week: The startup model meets AI economics.

Silicon Valley’s traditional culture of “fake it until you make it” has collided with the economics of AI, the WSJ's James Macintosh reports. Many AI companies lose money on every user because providing AI services costs more than customers pay. The hope is that fast growth will eventually lead to profitable scale, but that assumption is coming under pressure.

Market signals turn cautious. Two events last week highlighted the shifting mood. Nvidia and Microsoft announced a major investment in Anthropic, paired with Anthropic’s pledge to buy computing power from Microsoft using Nvidia chips. Such deals have lifted tech stocks in the past, but this time not so much. Nvidia’s strong earnings also failed to sustain an early rally.

Investors shift to near-term value. Broader volatility and fading expectations for interest-rate cuts have added to the skepticism. Investors now want less faking and more making, says the WSJ.  

Tech companies’ massive spending on artificial intelligence has triggered a surge of corporate bond issuance, straining credit markets and unsettling investors.

Since September, Amazon, Alphabet, Meta and Oracle have sold nearly $90 billion in investment-grade debt. Meanwhile, data-center builders like TeraWulf and Cipher Mining have issued over $7 billion in speculative-grade bonds. The combined flood has pushed yields higher and driven bond prices lower, WSJ reports. 

Pressure on tech company bonds applied unevenly. Oracle now pays some of the highest yields among investment-grade tech firms. Smaller players such as CoreWeave face even steeper financing challenges, says the Journal.

But companies like Alphabet, which focus on selling practical tools, have weathered the downturn better.

The release of its Gemini 3 has vaulted the company to the front of the AI race, with the model outperforming ChatGPT and Anthropic on a wide range of benchmark tests, the WSJ's Katherine Blunt reports. Early testers, including Box CEO Aaron Levie, reported unusually large performance jumps in document analysis and problem solving.

Gemini 3’s strong results mark a major turning point. Google has struggled for years to regain momentum after ChatGPT’s debut. The model now powers tools like Nano Banana and has helped drive Gemini’s user base to 650 million monthly users. 

A case study in aggressive AI financing.

Meta is building a $27 billion data center in Louisiana, financed with debt, and neither the data center nor the debt will be on its own balance sheet, WSJ reports.

Meta shifted the nearly complete data center into a joint venture with Blue Owl Capital, which owns 80% and issued $27.3 billion in bonds. Meta will lease the facility starting in 2029 under terms designed to qualify as an “operating lease,” minimizing recognized liabilities.

Although Meta claims it lacks control over the venture, it directs construction, operations and bears risks such as cost overruns. It also guarantees bondholders the full amount if it doesn’t renew its four-year lease.

A November cyberattack on SitusAMC, which is used by lenders for originating and servicing real estate loans, has prompted large U.S. banks such as JPMorgan Chase, Citi, and Morgan Stanley to assess potential exposure of customer data, New York Times reports.

The Cybersecurity and Infrastructure Security Agency, created in 2018 to warn election officials about potential threats from foreign governments, was absent in several elections this month, the Associated Press reports. The lapse has heightened worries among election officials about how involved CISA will be during the 2026 midterms.

Harvard University says an unauthorized party accessed a database containing information on alumni, donors, some students and faculty after a phone phishing attack, Bloomberg reports. This is the second security incident Harvard has investigated this year. Ivy League schools have faced a surge in attacks in recent months

California's DMV has given approval for Waymo to run its robotaxis in all of the Bay Area, Sacramento and in much of Southern California, the Information reports. 

The White House hailed what it called constructive talks with Ukraine in Geneva on Sunday, saying the two sides had modified the Trump administration’s proposed plan for ending the war with Russia. (WSJ)

On the silver screen, about half of all movies that made their debut last year included appearances of cigarettes, cigars and other tobacco products, up 10% from the previous year, according to a new report from public health nonprofit Truth Initiative and research organization NORC at the University of Chicago. (WSJ)