There are fresh signs that the AI bubble is wobbling. Again.
In recent weeks, a much-vaunted AI-focused hedge fund, Situational Awareness, run by Valley wunderkind Leopold Aschenbrenner lost $35 billion; nearly its entire fund. South Korea saw its largest-ever stock market collapse as investors pulled bets on AI chips and tech stocks. The circularity of the AI economy, with hyperscalers investing in AI labs and AI labs purchasing cloud compute from the hyperscalers and hyperscalers reporting big profits, is reaching new heights of precarity. Due to a fresh series of moratoria and political opposition, the future of the massive data center buildout has never looked more uncertain. The US markets have dipped then boomed, dipped then boomed.
Where does all of this leave us? Just about everybody—Sam Altman, the Financial Times, AI critics, your local barista—seems to agree that we’re still, somehow, improbably, in some kind of a bubble. But how big? How volatile? When will it burst? How? This week, I dive into all these questions with the infamous AI skeptic, analyst, and industry scapegoat Ed Zitron, author of the Where’s Your Ed At newsletter, host of Better Offline, and CEO of EZPR.
Zitron has been one of the loudest, most vitriolic, and most visible critics of the AI boom since the early days of ChatGPT. He also happens to be one of the most incisive. For my money, he’s done more to track the finances of the AI firms, and note the myriad ways they’re not adding up, than most business journalists and tech pundits combined. To this day, he’s raised questions about AI company revenues, circular financing, and cash burn that the industry still doesn’t have good answers for. As a result, he’s got nearly as many dedicated haters as he does diehard fans. Like a podcasting FDR with a special ire for generative AI, he welcomes their hatred.
Now, as fellow traveling skeptics of the broader AI project, Ed and I don’t agree on everything. But you’d be hard-press to find a sharper—or more entertaining—critic of the dubious economics of the AI boom. He’s been on a real tear lately, too, examining the worrying concentration of spending among hyperscalers and top AI labs, the financial vehicles underwriting the data center buildout, and the lack of demand for AI products outside of the chatbots built by the top labs.
Whether or not you appreciate his acerbic style or not (I do), we’d all do well to pay special attention to Zitron’s warnings. If he’s right, we’re on the brink of a rather spectacular collapse.
As always, if you like what we’re up to with the podcast/YouTube show, please subscribe, follow, and review. It’s starting to get picked up by all the various algorithms and all that stuff helps. And if you really like it, please consider becoming a paid subscriber to make the whole BLOOD IN THE MACHINE project possible.
On the Texas data center pause
This is big news in the data center opposition world right now is that Texas’s deeply conservative governor, Greg Abbott, has issued a pause on data center projects until each can be fully audited.
Via a press release from the state:
Audit Must Be Completed Before Any Data Center Project Moves Forward
Governor Greg Abbott today directed the Public Utility Commission of Texas (PUCT) and the Electric Reliability Council of Texas (ERCOT) to conduct a comprehensive verification and audit of all data centers advancing through ERCOT’s interconnection process. The PUCT and ERCOT must complete this audit before any data center project moves forward. Any project that fails to comply with the requirements set forth by the PUCT and ERCOT will be denied connection to the Texas grid.
“Our top priority is to protect Texans’ safety and quality of life,” said Governor Abbott. “Any project that fails to comply with the requirements set forth by the PUCT and ERCOT, and by state law, must be denied connection to the Texas grid. Simply put, Texans must come first.”
ERCOT is currently considering approximately over 474 gigawatts of requests to connect to the Texas grid, more than five times Texas’ record peak electricity demand for ERCOT. Approximately 90 percent of the new power requests are data centers. That unprecedented load growth could endanger the reliability and stability of the Texas electric grid.
Texas joins New York as the second major state to officially block data center development, though its mechanism is much different than New York’s. New York is a flat moratorium for hyperscale projects of a certain size, Texas’s is a pause to assess whether certain criteria are met before they’re given the green light. But the binary is interesting here; New York and Texas are two cultural polar opposites; red/blue, rural/urban, cowboy/yankee etc etc etc—more evidence of the way that resisting big tech’s data centers is the nation’s great unifier.
There is, I would hasten to add, a potential gap between the efficacy of the approaches here: Texas’s audit is constructing in such a way that makes it possible to greenlight the bulk of the projects in relatively short order after Abbott’s having performed the grand act of Having Done Something. Trump has already criticized Abbott’s move, which will put some pressure on hastening the compliance with industry at scale here.
Still. It’s an unambiguous sign of the current power and bipartisan reach of the data center opposition movement, that even the most pro-corporate of Republican governors feel they must respond to its demands, and to the backlash against AI, in visible and tangible ways.
People hate AI so much tech platforms have been forced to label and ban it
Speaking of the backlash to AI, Reece Rogers has a fun story in WIRED detailing all the ways that non-AI tech companies have recently found themselves forced to respond to people’s overwhelming distaste of encountering slop on their platforms.
From the piece, “The AI Slop Backlash Is Actually Having an Impact”:
Many Americans are having similar reactions to the growing prevalence of generative AI in daily life. “The AI revolution has happened, and everybody hates it,” says Meredith Broussard, a data journalism professor at New York University and author of Artificial Unintelligence: How Computers Misunderstand the World. In a recent Gallup poll, Americans’ increased familiarity with generative AI coincides with negative attitudes about the technology. This is especially true for young people between 18 and 29, where almost half see generative AI as doing more harm than good.
While this AI backlash has been brewing for the last few years, more companies appear to be directly addressing these concerns. LinkedIn added a “seems like AI slop” button for reporting AI-generated content on the platform, despite still hosting plenty of generative AI tools. Snapchat announced that fully AI-generated videos would no longer be eligible for its discovery feed. Substack recently added an AI detection tool to police writers.
The backlash’s “why” is simple at its core. Many Americans feel like they never consented to any of this. I had no idea that my online interactions would one day be scraped en masse for AI model training. I didn’t ask for Instagram to enable deepfakes on my account. I didn’t want Google to push AI-generated answers at the top of search results. I woke up one day, and generative AI features were stuffed into every corner of all the software I rely on without my consent.
Whether it’s AI models scraping your data or noisy data centers popping up in your backyard, the AI revolution seems to be an event happening to us without making us feel like active participants.
“Consent is really important,” Broussard tells WIRED. “We can connect that to the way that women, queer people, people of color, and minorities get exploited and abused online. Tech companies have never really been sensitive to issues of consent.”
Worth noting too that the story spent much of the week hovering in the MOST POPULAR column at WIRED, our most prestigious magazine about technology.
I would slot DuckDuckGo’s recent viral stunt into this category, too. It recently unveiled its “Normal F***ing Sunglasses,” which, it touts, were made with “No AI” and “no cameras.”
It’s a delightful little rejoinder to Meta’s AI glasses, which for their propensity for helping men record video of women without their consent have been dubbed “pervert glasses” by seemingly everyone outside the company.
From the ‘AI is ushering in a boom in carbon emissions’ dept
Amazon is investing in a large-scale natural-gas-burning power plant as part of a huge data center in Texas, the company confirmed on Friday, a facility that could become the largest single source of climate pollution in the United States.
Amazon and other tech giants are racing to build enough data centers to keep up with surging demand for artificial intelligence services, and to secure the electricity to power them. A company spokeswoman said in a statement that its planned data center, in Pecos County, would be “powered by new on-site generation that won’t raise electricity costs for Texas families.”
The new gas-burning plant, if built to specifications, would be permitted to release 33 million tons of carbon dioxide a year, regulatory records show, more planet-warming gases than any other power plant in the country.
Is AI solving climate change yet?
Funny I should ask. The Amazon power plant story, which broke last week, underscores the findings of new research conducted by former Microsoft employees. As the Guardian put it, “AI’s potential climate benefits outweighed by role in boosting fossil fuels, study finds.”
From the story:
AI-driven productivity gains enable more planet-heating pollution from fossil fuels than they avoid from renewables, a study has found.
Researchers modelled the technical potential for AI to boost clean power generation along with projections for how it can help produce coal, oil and gas. Across 64 scenarios, they found net yearly carbon pollution rose by 0.47-1.8 gigatonnes, or about 1-5% of the energy sector’s annual emissions.
The study is the first to quantify the climate impact of AI across the full power sector. Previous research has explored its indirect climate benefits – such as reducing renewable downtime and optimising electricity grids – while ignoring the extra pollution from increasing productivity in drilling oil wells and extracting gas.
The notion that AI would soon be a net benefit for the climate has always been dubious, and this study shows exactly why. Those alleged and ultimately marginal benefits will be wiped away by increased demand for fossil fueled power—precisely of the sort that Amazon is trying to build out, at scale, as we speak.
Alright, that’s it for now. Thanks for listening and reading, and thanks again to everyone who chips in to make this work possible. You’re the best. Until next time—hammers up.
