Financial Activity with Artificial Intelligence: What Does it Mean, and What’s Next?

This topic hasn’t come up for my posting for five months, but that hardly means it has been idle, hidden, or unimportant.

First is “Anthropic Tops OpenAI to Become the World’s Most Valuable A.I. Start-Up” (Mike Isaac and Cade Metz, The New York Times, May 28^th).  That firm “raised $65 billion in new fund-raising that put its value at $900 billion, ahead of OpenAI’s last valuation of $730 billion.”  Both are among the most important model producers.

A headline raising immediate questions leads “Investors Pile On as A.I.’s Rough Summer Continues” (Andrew Ross Sorkin et al., The New York Times, June 26^th).  After the question of how can a season be described so early, how “rough” can it be if investors are still there?  The authors said “A.I. giants already faced high interest rates, rising input costs and corporate customers rethinking their A.I. spend.”  Since then, they have had to deal with growing concern about regulation, though little there has yet happened.  It was followed by “A.I. Has Rewarded Investors.  It May Now Post Their Greatest Risk” (Jeff Sommer, The New York Times, July 10^th).  The author’s main point was that there were many companies which have “become a backdoor bet on A.I.,” meaning those not producing the software but getting large sales from its progress.  The example here, along with other technology firms, is Caterpillar, for its “power-generation turbines” at data centers, a key reason why its “stock has soared.”  Sommer advocated investor diversification, including “cash and bonds,” as “the eventual faltering of the A.I. trade may now be the market’s greatest risk.” 

That author was back a week later with “A.I. Is Running on Borrowed Money,” also in the Times.  We have known that companies involved with the technology have been lending back and forth, but here is Oracle “increasingly selling bonds,” along with Microsoft which might not keep its top credit “rating after its splurges on A.I. data centers,” and Meta, Alphabet, and Amazon, also “giant investors” in those facilities.  Although these companies previously funded these ventures, officially investments and not expenses, from their operations, they are now borrowing more, with “the gigantic A.I. infrastructure expenditures… outpacing growth in profits.”  Since “the capital markets” have “begun to charge premiums for what it considers to be heightened risk,” and some bond trading prices are falling, “a great winnowing is coming.”

Quantifying AI value has been difficult, but now there is a measurement.  In “What Are Companies Getting for All That A.I. Spending?” (Lydia DePillis, The New York Times, August 3^rd), the author described “tokens,” “generally understood to represent a few characters of English text,” as “the units of computing power in which A.I. is sold,” “tokenmaxxing” as what happened when “executives encouraged as much A.I. use as possible,” ““tokenminning” after they rapidly burned through millions of dollars in company money,” and “tokenomics” as “the study of how this limited resource is created, traded and converted into things people want.”  Yet, for various reasons, including that “some models… will keep trying to accomplish a goal even when the mission is hopeless” this measurement scheme is only a vague beginning.

Meanwhile, “The US went all in on AI investment.  Is it driving up consumer prices?” (Rachel Barber, USA Today, September 6^th).  The Bureau of Economic Analysis, which issues the Personal Consumption Expenditures Price Index, “the Fed’s preferred measure of inflation,” found that “the shift in AI-driven demand” had been “responsible” for 0.4% of the 3.3% July core-price year-over-year gain.  The federal Labor Department’s “computers, peripherals and smart home assistant devices” category in its Consumer Price Index went up 3.5% “from June to July,” “faster than the rate for all items combined for the entire year ending in July.”  Other areas of possible increase, such as electricity, were uneven and sometimes nonexistent, but, overall, “if one supply chain has a higher demand for a particular input, and that outstrips demand from another source, then prices go up.” 

Starting the news from the past two weeks, “Weak Links Start to Show in A.I. Borrowing” (Maureen Farrell and Rob Copeland, The New York Times, October 17^th).  Questions of funding and repayment “burst into the forefront this week when investors rapidly lost confidence in one debt-laden data center company, CoreWeave,” shown as “prices of its bonds… crashed, and investors placed a more than 60 percent likelihood CoreWeave would fail to meet its obligations.”  Including other companies mentioned, the authors asked “whether companies facing debt trouble like CoreWeave, Oracle and SoftBank are the industry’s weak link or are harbingers of deeper problems in the A.I. build-out.”  Perhaps they were too aggressive at the wrong time, perhaps they skipped financial safety measures, perhaps investors are just too nervous, or perhaps there will be many more to come. 

One company is not slowing down: “Meta Raises Its A.I. Spending, as Cash Pressures Grow” (Eli Tan, The New York Times, October 21^st).  That firm “is feeling the pressure of keeping up those investments,” and “raised the floor of its capital expenditures this year to $135 billion from $130 billion,” mostly going “toward building data centers.”  As a result, “its free cash flow, which is the money it had on hand after paying for expenses, turned negative for the three months that ended Sept. 30, dropping to negative $2.12 billion.  That was down from $784 million at the end of June and $8.55 billion in June 2025.”  It is covering that through issuing a new class of stock, and has also “struck some unconventional arrangements that do not appear on its balance sheet” such as a data center build “financed by private credit firms.” 

Finally, to answer an exceedingly pertinent question, “What is the ‘AI bubble’ and what happens if it pops?” (Greta Reich, USA Today, October 27^th).  Released one day after Anthropic “filed for its initial public offering (IPO),” and thereby provided “a first official look into its business operations and ownership,” it defined a bubble as “overinvestment in a space that won’t see a positive return in investment, eventually creating market instability.”  As “different areas… could see higher returns on investment than others,” an AI bubble would not be “a homogeneous, consistent kind of thing,” and investors disagree on whether a bubble exists now.  One capital firm estimated that “money companies and individual investors” were getting $600 billion less than “they thought (they) would.”  Anthropic is “expected to start trading before Thanksgiving, which will serve as a litmus test for investor demand and set the stage for other private AI companies, including OpenAI,” which had previously expected to go public this year.

Is there a bubble now?  There must be, but we don’t know just where or how big.  There are enough companies that the herd of AI-related ones will be culled.  As above, we may have already seen the beginning of that in the data center area with CoreWeave.  We most likely do not need all the largest American AI-software producers now in action.  Eventually, though, it will take more than that to permanently damage the AI industry; while it could cause a downturn of some size, it and the economy will rebound.  The next six to 12 months will be critical - if AI’s business side makes it through that, we can then focus on artificial intelligence’s other problems.