Just to step back and get a look at the big picture (or the big pictures), when we talk about AI, which these days almost always comes down to large language models and LLM-driven agents, there are at least four huge stories to choose from.
1. The capacities and limitations of the rapidly evolving technology. This is both the most exciting and infuriating part of AI news, with legitimate breakthroughs and sensationalistic crap fighting for readers' attention.
2. The tremendous and often tremendously negative impact that this huge breakthrough in natural language processing has had on society, including AI slop, new and improved pig butchering scams, AI-generated documents clogging up the court system, hallucinations corrupting data and scientific research, AI psychosis, and people increasingly turning to computers for human interaction.
3. The Doomsday cult (going by the somewhat questionable name the rationalists) that has been quietly gestating in Silicon Valley for 15 or 20 years now and which has come to dominate the rank and file and possibly the leadership of the industry.
4. The magnitude and potential impact of the AI bubble. Which we'll get to in just a minute.
There are lots of factions and subfactions battling each other in these debates, but for me, the most useful distinction is between the Believers and the skeptics. While there are lots of distinguishing traits, the most important is probably a question of focus. The former like to talk about what the technology might be able to do someday. The latter insist on talking about what it definitely can do today.
Regular readers can probably guess which camp we are in.
The skeptics still represent the minority view, pushing back against the rationalists, the tech industry (which has literally trillions of dollars on the line), and a press corps that can no more turn down a Next Big Thing story than a junkie can turn down a fix. That said, the skeptics are clearly gaining ground.
You can list Josh Marshall among those moving toward the Cal Newport/Adam Becker/Paul Kedrosky camp. Marshall, like Paul Krugman, has always suffered from a mild version of Gell-Mann amnesia when it comes to outlets like the New York Times. Within his areas of expertise, he knows better than anyone how misleading the Establishment press's coverage can be, but he still tends to trust the standard narrative everywhere else.
As recently as a few days ago, he was favorably citing Jacob Coxon's Wired interview, but as he has dug into the story, he has since moved to the skeptic side, quoting Gary Marcus, critically reading Yudkowsky (someone who does not stand up to close scrutiny), and writing posts that feature the following graphs.
When it comes to the financial establishment, the skeptical view may now actually be mainstream. As the Financial Times (which, to its credit, started asking the tough questions very early in the game) has observed, the Bank for International Settlements now sounds like Ed Zitron without the profanity.
And then there's The Wall Street Journal piece cited by Marshall. If the WSJ doesn't represent the business and finance establishment, I'm not sure what does.
[Emphasis added throughout.]
The AI Build-Out Is Becoming the Biggest Economic Bet in U.S. History
Konrad Putzier, Justin Lahart
Sept. 23, 2026 9:00 pm ET
The AI build-out is on track to become the biggest economic bet in U.S. history, dwarfing the investments made to fund other huge U.S. infrastructure projects such as the railroads, the highway system and the plumbing for the internet.
Total investment in data centers and related artificial-intelligence infrastructure is projected to total $10.3 trillion from 2025 to 2032, according to new estimates by economist Stijn van Nieuwerburgh published by the Brookings Institution. That is a staggering 3.6% of gross domestic product a year, on average. Never before has the U.S. economy been so dependent on the build-out of a single industry.
The investment is transforming every corner of the economy, creating hundreds of thousands of jobs and minting new billionaires.
It is also creating significant risk, as much of it is built on debt. An abrupt slowdown could ignite shock waves throughout the U.S. economy.
Projecting investment is tricky, and total spending might well end up substantially lower. Still, the money poured into data centers this year already represents an investment unprecedented in recent history. AI investment in the U.S. is projected to hit 1.9% of GDP in 2026, according to new estimates from Goldman Sachs. The railroad boom of the late 19th century marked the last time the build-out of one new industry accounted for a larger share of the economy.
Here are some important impacts all that money is having on the economy:
Construction
The flood of money spent on the build-out of data centers has represented a bright spot in an otherwise dark time for the construction industry.
Through July of this year, a seasonally adjusted $37 billion has been spent on private data-center construction—about $9 billion more than in the first seven months of last year, according to the Commerce Department.
Private construction spending on everything else—houses, apartment buildings, shopping centers and so on—was about $46 billion below year-earlier levels in the first seven months of this year.
Hyperscalers are tying up scarce workers and electricity, raising costs for other businesses. The Federal Reserve Bank of Richmond recently reported that data-center construction is straining labor availability in its area.
Last year, Mississippi was in the running to land an aluminum smelter, along with an estimated 1,000 permanent jobs. But a data center was announced near one of the proposed sites, near Vicksburg, tying up electricity needed for the smelter, according to a person familiar with the operator’s decision-making. So, the smelter chose Oklahoma instead.
It isn’t just electricity: In many places, data centers are pushing up land costs. “It’s crowding out manufacturing,” said Didi Caldwell, a site-selection consultant who works with heavy industry.
Financing and risk
Analysts estimate that capital spending at five of the so-called hyperscalers—Alphabet, Amazon.com, Meta Platforms, Microsoft and Oracle—will be $4.2 trillion in the four years ending in 2029, according to FactSet. A growing share of that spending is financed by debt.
That level of spending raises risks for the financial sector if the boom goes bust. Van Nieuwerburgh said often tech companies use off-balance-sheet entities to borrow from banks and private-credit firms, and those deals usually come with little public reporting. The practice makes it hard to figure out how great the financial risks are, he added. If AI doesn’t generate enough revenue to service the debt raised to build data centers, the fallout could ripple through the financial system.
...The wealth effect
The AI-powered rally has led to huge gains in stock-market wealth. As of the second quarter, U.S. stock and mutual fund holdings came to $63 trillion, according to the Federal Reserve—nearly double the amount at the end of 2022. The trend has buoyed consumer spending even as inflation-adjusted wage growth has faltered. The gains have been particularly pronounced for the well-off, who tend to have more of their net worth tied up in stocks than the middle class do.
Nationwide, home sales have been mired in a four-year slump. But in Silicon Valley, AI money is fueling a surge in luxury-home sales. “This is the best it’s been since 2000,” said Ken DeLeon, a real-estate agent. DeLeon recently listed a five-bedroom home for $9.9 million. He got seven offers, and within two weeks it was in contract for more than $13 million. The winning bidder: an AI entrepreneur.
Inflation
Demand for equipment that goes into data centers—notably, memory chips—is leading to shortages and driving up costs for tech products. The prices importers paid for computers, peripherals (such as hard drives) and semiconductors were 20% higher in August than a year earlier. Those high import prices are in turn putting upward pressure on the costs of some consumer goods, such as iPhones and gaming consoles, and contributing to inflation.
Chicago Fed President Austan Goolsbee recently warned that data-center investment is pushing up wages in related fields. And Fed chairman Kevin Warsh named borrowing by hyperscalers as one reason why long-term interest rates are up, which in turn has made homeownership less affordable for millions of Americans. Electricity bills have surged in some areas with lots of data centers.
No comments:
Post a Comment