{"id":82431,"date":"2025-10-16T15:37:10","date_gmt":"2025-10-16T15:37:10","guid":{"rendered":"https:\/\/www.newsbeep.com\/nz\/82431\/"},"modified":"2025-10-16T15:37:10","modified_gmt":"2025-10-16T15:37:10","slug":"the-ai-boom-is-propping-up-the-whole-economy","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/nz\/82431\/","title":{"rendered":"The AI boom is propping up the whole economy"},"content":{"rendered":"

\"\"<\/a>AI holds up the economy \u2014 for now. (Illustration by DALL\u00b7E)<\/p>\n

In case you missed it, I wanted to share our weekly housing news write-up<\/a> from Writing Fellow Halina. If you enjoy this one, we post the latest in housing every Wednesday and share shorter original pieces every Monday and Friday for paid subscribers. If you\u2019re already a paid subscriber and would prefer not to receive our evening pieces (or would prefer to only receive those), here\u2019s how to opt out<\/a>.<\/p>\n

A \u201cdata center\u201d is basically a big building full of computers. Until recently, you probably didn\u2019t hear very much about data centers, because even though computers are an important part of daily life, building large structures full of computer equipment is not a particularly large share of the economy. And yet it\u2019s in the news a lot these days, because this year the change in spending on data center construction is large relative to the total change in economy-wide spending.<\/p>\n

In other words, as Jason Furman recently noted<\/a>, the Bureau of Economic Analysis\u2019s category for investment in information processing equipment and software accounts for over 90 percent of economic growth in the first half of 2025. <\/p>\n

\"\"<\/a><\/p>\n

This doesn\u2019t mean that without the data center boom, we would have experienced zero economic growth. Much of the financial capital and real resources that are currently going into data centers would have been deployed by some other sector absent the boom. But in terms of the actual American economy, the AI boom is the whole story. <\/p>\n

The story of how the stock market learned to love the Trump administration<\/a> is not without nuance, but it is actually overwhelmingly the case that AI-related stocks dominate the market, and the administration has been quite solicitous of AI-specific interests \u2014 including with tariff exemptions<\/a>. <\/p>\n

But not only is there a lot of money being made off the AI boom, there\u2019s a lot of obvious boom-market behavior. AI start-ups are getting funding with minimal scrutiny, and the AI construction boom increasingly features unorthodox financial arrangements. There\u2019s nothing wrong per se with a little innovation in how deals are structured. But it\u2019s also true that, throughout history, we\u2019ve seen investment booms go beyond a sustainable level, fueled by creative financing that obscures weak fundamentals. I\u2019m hearing more and more questions about an \u201cAI bubble,\u201d and I even see people confidently<\/a> proclaiming that one exists<\/a>. <\/p>\n

That\u2019s not a crazy hypothesis, for reasons that we\u2019ll explore below. But I do also want to caution against motivated reasoning. <\/p>\n

A lot of people are both deeply alarmed by Donald Trump and also deeply resistant to Slow Boring-style calls for significant steps to expand the Democratic Party<\/a> to include those with more conservative views on some issues. <\/p>\n

Democrats\u2019 best chance to defeat Trump through pure acts of stand-your-ground resistance is if the economy experiences a massive setback, just as Barack Obama won a landslide victory in 2008 with a platform that was no more moderate than John Kerry\u2019s (though significantly more moderate than Hillary Clinton\u2019s, Joe Biden\u2019s, or Kamala Harris\u2019s) simply by pairing charisma with very bad fundamentals for Republicans. <\/p>\n

So telling yourself that the economy is on the verge of collapse offers a convenient way to avert political tradeoffs. Similarly, the data center boom is raising electricity demand in ways that are problematic for environmental groups\u2019 net zero timelines. But blocking data center construction would seem to have major economic costs that people will probably not want to bear. However, if all the investment is just a crazed speculative bubble that will end in tears, there is no economic cost to blocking the buildout. <\/p>\n

Those are certainly reasons to want to believe there\u2019s an AI bubble. But they\u2019re not good reasons to believe it. <\/p>\n

The poster child for the bubble thesis is actually a bit removed from the issue of data centers. <\/p>\n

Mira Murati was OpenAI\u2019s chief technology officer, but left the company in the wake of the abortive coup against Sam Altman. She then put together a small but impressive team of people from leading AI labs and successfully raised<\/a> $2 billion in venture capital (at a $10 billion valuation) for her new start-up, Thinking Machines, which by most accounts had no demo product or even really a clear pitch. <\/p>\n

It\u2019s common \u2014\u00a0clich\u00e9 even \u2014 in V.C. spaces to say that you\u2019re really backing a team, not a specific idea. <\/p>\n

I\u2019ve lived that. When we were pitching the idea that became Vox, people absolutely did listen to the details of our pitch, and they asked questions about it and had varying degrees of enthusiasm in response to it. But fundamentally, the people who were interested in the pitch were interested in Ezra Klein and Melissa Bell and Matt Yglesias.<\/p>\n

And that was not a crazy calculation. A lot of our initial big ideas frankly didn\u2019t work out, many of them quite quickly and obviously after we started. But we\u2019d hired a fantastic initial team, and when you have smart people with good values and good work ethic, you drop ideas that don\u2019t work and shift to other ideas that are better. Some of our biggest successes ended up coming in places like the video program, where we had virtually no ideas in our pitch, but we hired two absolute geniuses and they hired more amazing people. <\/p>\n

Still, the other thing that really helped us get off the ground is that, at the time, investors were enthusiastic about ad-supported digital media properties. There was a thesis (wrong, as it turned out) that you could make a lot of money in that business. So a ton of digital media start-ups got funded, and many of them failed. But what really made the idea a dud is that even the ones that didn\u2019t fail \u2014\u00a0like Vox \u2014 did not turn into huge successes that made tons of money for the founders and investors. <\/p>\n

That\u2019s where it gets hard to judge these things.<\/p>\n

Investing in start-ups is inherently uncertain, and the business model depends less on having a very high batting average than on the idea that home runs will be extremely lucrative. <\/p>\n

So when a given sector becomes fashionable, the bar for which pitches get funded is lower. We pitched Vox at a time when the bar for digital media pitches was relatively low. If we\u2019d gone out with that five years earlier or five years later, nobody would have put money into it. With AI right now, the bar is currently very low, and it\u2019s obvious that a lot of the projects getting funded aren\u2019t going to work out. But this doesn\u2019t actually speak to the merits of the thesis that this is a promising field. The thesis lives or dies primarily on the question of how big the hits will be. And if I could tell you that, I wouldn\u2019t be writing a subscription newsletter for a living. <\/p>\n

The most eyebrow-raising aspect of the whole thing, though, is the spate of unorthodox financial arrangements that have been announced this fall. <\/p>\n

Consider last week\u2019s deal between OpenAI and AMD<\/a>. <\/p>\n

Cutting-edge AI work has thus far mostly relied on Nvidia high-end chips. AI labs, naturally, don\u2019t want all the financial returns to this sector to accrue to their chip supplier. So, OpenAI decided that they were going to buy a bunch of chips from a competing company, AMD, that\u2019s been a bit left behind in the AI race. <\/p>\n

The announcement of the deal caused AMD\u2019s stock to go up a lot, in part because the deal should make the company a lot of money. But also because most people aren\u2019t technical experts in training AI models or powering AI inference or in GPU design, but are aware that Sam Altman and his team have an incredible amount of technical expertise in this field. If OpenAI says that AMD chips are useful for the AI infrastructure buildout, that\u2019s a huge vote of confidence that conveys broader information about AMD\u2019s ability to make money off of AI by selling to many companies. <\/p>\n

Where it gets really interesting is that OpenAI foresaw that announcing the deal would cause AMD\u2019s stock to soar. So the terms of the deal included warrants for OpenAI to get AMD shares if the price of AMD stock soared. The upshot of this is that, yes, OpenAI is in a sense getting the chips it agreed to purchase for free \u2014 the profits on the equity investment in AMD exceed the cash value of the chips purchase. But AMD shareholders also benefit from giving OpenAI the chips, because despite the dilution of their shares, the price increase still leaves them better off.<\/p>\n

Each step in this chain makes perfect sense but, as Matt Levine says<\/a>, it\u2019s also the kind of thing where OpenAI\u2019s dealmaking is so impressive, they could be tempted to spend all their time on deals rather than on building AI. <\/p>\n

\"\"<\/a><\/p>\n

Or in other words, while I have absolutely no reason to believe that this transaction is anything other than completely above-board, you certainly could take advantage of the basic market dynamics in play here to run a scam. <\/p>\n

We\u2019re also seeing deals that go in the opposite direction. Nvidia\u2019s stock has been flying high on the assumption that the company will sell more and more high-margin chips to AI companies in the future. And AI companies certainly do seem to want to buy a lot of chips. But those chips are expensive, which means the AI companies need to raise money to be able to buy them \u2014 Nvidia\u2019s future depends on people investing in AI companies who buy Nvidia\u2019s products. So in a recent deal, Nvidia cut out the middleman and made a large investment in xAI<\/a>, and xAI is using the money to buy Nvidia chips. <\/p>\n

The circularity here doesn\u2019t necessarily mean that anything is wrong. In some sense, the whole economy is just money moving in circles. That\u2019s why expansionary monetary policy can help revive a depressed economy \u2014 your expenses are someone else\u2019s income, so simply putting more money into circulation can generate real economic activity if there are idle resources or workers to mobilize. But when you get a very small circle rather than a big series of arm\u2019s length transactions, the potential for shenanigans increases. <\/p>\n

In a true bubble \u2014 like baseball cards when I was a kid<\/a> or Beanie Babies shortly afterwards<\/a> \u2014 people are buying assets purely on the basis of the belief that they\u2019ll be able to unload the asset to someone else at a later date. <\/p>\n

Collector\u2019s items like the aforementioned baseball cards are particularly prone to this dynamic since, by definition, there\u2019s no stream of revenue associated with owning a Ken Griffey Jr. record card. Of course, just because something doesn\u2019t generate a revenue stream doesn\u2019t mean its value is fake. If you have a bunch of Picassos lying around your house, those are assets with very real value. But there\u2019s still nothing like dividends or rents associated with the asset the way there are with shares of stock or homes, so there\u2019s a lot of opportunity for pure speculation to run rampant. <\/p>\n

But it seems to me that if you think about the dot-com stock boom or house-flipping in the mid-aughts, it was incredibly common to believe that you could always offload the asset to some other sucker. There\u2019s a reason that bucket shop scams<\/a> proliferated in the 1990s \u2014 very unsophisticated people were buying stocks on the assumption that stocks would always go up, which meant they weren\u2019t scrutinizing specific pitches carefully. <\/p>\n

There are often echoes of that behavior in cryptocurrency markets, and we also saw a lot of it when people bored at home during the pandemic got really into \u201cmeme stocks.\u201d<\/p>\n

And what I will say on behalf of the AI boom is that I\u2019m pretty sure it\u2019s not a bubble in that particular sense. <\/p>\n

The leading AI labs are genuinely seeing rapid revenue growth<\/a>. I don\u2019t think \u201cthe trajectory of model capabilities improvement and associated revenue will continue on its current path indefinitely\u201d is the most sophisticated investment thesis in the world. But it\u2019s also not unmoored from reality. It\u2019s a plausible but possibly wrong guess about the future state of the world \u2014 i.e., a pretty normal investment hypothesis. <\/p>\n

\"\"<\/a><\/p>\n

I also think that some critics are confusing these companies\u2019 thirst for new capital with the \u201cwe lose money on every sale but make it up in volume\u201d nonsense of dot-com-era companies like Kozmo<\/a>. What\u2019s happening here is that the financial return on any given L.L.M. is pretty good, but each company is continually investing large amounts of money in training the next model on the expectation that it, like its predecessor, will be lucrative. <\/p>\n

Does that mean all\u2019s well? Of course not! This could break down in a number of ways: <\/p>\n

The hypothesis of continued upward progress in model capability \u2014 \u201cscaling laws\u201d that dictate that applying more data and more computing power will predictably generate increasingly powerful models \u2014 might be wrong and you could hit a wall. <\/p>\n

The frontier companies might lack a meaningful \u201cmoat\u201d such that competition renders cutting-edge models actually not that lucrative, no matter how capable they are. <\/p>\n

Increasingly capable models might do something harmful that destroys rather than creates economic value. The Entity from the \u201cMission Impossible\u201d movies was surely bad for stock prices, despite its impressive technical abilities. <\/p>\n

If it turns out that some of these closed loop deals between the big companies are scammy, that could create a huge crisis of confidence, even if the underlying technology and finances are fine, which would become a kind of self-fulfilling anti-bubble. <\/p>\n

I would not be shocked if some huge stock-market crash is around the corner, but I\u2019m also not betting that it is. As a prognosticator, I take the bubble possibilities seriously, but I take the problem of motivated reasoning equally seriously. I\u2019m much more knowledgeable about politics than I am about finance, and I can tell you that many of Democrats\u2019 internal conflicts feel easier to resolve if they assume something like \u201cNot only is the Orange Man bad, but he\u2019s also bad at his job and his effort to prop up the economy by coddling the AI sector will end in tears.\u201d <\/p>\n

I think that\u2019s more or less what Bharat Ramamurti is saying here about why I\u2019m wrong<\/a>, that Democrats don\u2019t really need to waste much time thinking about ideology and public opinion \u2014 they really just need to hash out some forward-thinking policy ideas they can implement when power falls into their laps. I don\u2019t think anyone is per se rooting for a recession or an economic collapse, but convincing yourself that one is coming is a soothing alternative to thinking through the question of how to win in 2028 if unemployment stays low and growth remains robust. <\/p>\n

My take on this is that you shouldn\u2019t take financial advice from your favorite political blogger, but also that a political movement shouldn\u2019t take its strategic direction from their hope that the economy is an unsustainable house of cards. <\/p>\n

It might be, in which case people in 2029 will have fun poking fun at all my circa-2025 hand-wringing. But the nice thing about financial markets is that the people investing in them have very strong incentives to at least try to avoid wishful thinking and motivated reasoning and instead focus on cold hard cash. Those people absolutely get things wrong sometimes, but they\u2019re right more often than not. And it\u2019s dangerous to assume they\u2019re wrong at just the moment it would be most convenient for you. <\/p>\n

Share<\/a><\/p>\n","protected":false},"excerpt":{"rendered":"AI holds up the economy \u2014 for now. (Illustration by DALL\u00b7E) In case you missed it, I wanted…\n","protected":false},"author":2,"featured_media":82432,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[11],"tags":[138,219,111,139,69],"class_list":["post-82431","post","type-post","status-publish","format-standard","has-post-thumbnail","category-economy","tag-business","tag-economy","tag-new-zealand","tag-newzealand","tag-nz"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/82431","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=82431"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/82431\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/82432"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=82431"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=82431"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=82431"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}