Is Artificial Intelligence Going to Tank the U.S. Economy?
AI Giants are spending way more than they make, and now they’re running low on essential hardware. Is the AI bubble about to pop? What would happen if it did?
Photo Credit: Donald Trump: AP Photo/Alex Brandon, Alex Karp (Palantir CEO): Reuters, Sam Altman: Getty Images
Over the last few years, there has been a sudden explosion of AI data centers across the country. Most are being planned in small towns, with some being home to as few as 2,000 people. Despite the fact that most of the locals fighting back are totally new to organizing and activism, they’re putting up one hell of a fight - in some areas, they have actually managed to beat the tech giants! However, as we explained in our previous article titled “The Fight Against AI Data Centers,” even when we win impressive victories, these companies are not just giving up and going home - they are relocating to other areas where they think they have a better chance of beating any resistance, especially in rural areas of southern states.
The threat AI poses to working people in the U.S. goes beyond data centers driving up electricity bills. The U.S. economy is sick and unstable, and facing a deep crisis of profitability, which has forced investors to seek out big returns on riskier and riskier investments. This includes the highly volatile cryptocurrency markets, high-risk high-reward venture capital startups, betting on market movements, high-yield “junk bonds,” and of course, AI speculation. Some Wall Street tycoons have even been speaking about AI in terms like “magic cure.”
Despite this hyperbolic level of optimism, some AI investors are beginning to sweat. To lay it out as simply as possible: the problem is not that these companies are not turning a profit, but rather that they are spending significantly more than they make to expand the current AI infrastructure and try to dominate the market. Furthermore, there is a concern that their soaring stocks are overvalued compared to the current profits being generated - investors are placing bets on what AI will deliver at some point in the future more than they are betting on what it will deliver this year. Everything is all fine and dandy until too many quarterly reports reflecting overspending and delayed profitability come out in a row and investors start to panic. If too many get scared and dump their stock all at once, the AI investment bubble could pop and the companies could suffer sudden and catastrophic decreases in valuation. Given that these are some of the biggest companies in the world, including Google and Amazon, a collapse like this could cripple the entire economy and trigger a massive recession.
Even still, this is not a guarantee of an AI-fueled economic collapse by itself. When trying to predict whether the AI bubble is going to pop, there are a number of moving pieces that we need to understand.
Why is the Billionaire Class All-In on AI?
One of the main reasons the billionaire class is in love with AI is because of the massive potential it has to affect their profits. Their hope is that AI will create a massive boom in productivity that will allow them to ramp up the speed of work while also downsizing their workforce and cutting their labor costs. Corporations in the U.S. have already laid off hundreds of thousands of workers this year, with many companies citing AI as a cause and shifting work on to their remaining employees while they wait for AI to be able to fill in the gaps. Some of this is certainly “AI washing,” which is to say using AI as an excuse for layoffs these companies already want to enact, but it doesn’t change the fact that companies are looking to use AI to make cuts to the number of entry-level positions for white collar workers.
This isn’t to say all of these white collar jobs will disappear - they won’t. However, if these companies get their way they will make cuts to their workforces, which will create a spike in the already-considerable number of unemployed and under-employed workers in the U.S. as part of a broader effort to drive down wages and bust unions through exploiting desperate workers willing to work for poverty wages, and possibly scab in the instance of a workplace strike. To the billionaire class, this is just another reason to go all-in on AI.
The idea that AI will primarily, or even only, affect white collar workers is gaining some popularity, and it is incorrect. AI will undoubtedly be used to enhance management surveillance, especially in factories and at companies like Amazon, for example using facial recognition to track workers throughout a workplace and record how long it takes them to complete various tasks, how long they rest between tasks, how often they go to the bathroom and how long they take there, so management can force everyone to work faster. Even beyond that, AI will greatly aid management union busting by using facial recognition software to identify workplace activists and agitating troublemakers who talk to their coworkers too much, or who pass out flyers on the job.
However, it’s not just about AI in the workplace. Among the most terrifying applications for AI that the capitalist class has found is in domestic surveillance by police departments, which you may know by the name of the company dominating the industry: Flock. Flock is a company that makes cameras that track license plates of cars - at least in theory. What started as a driver surveillance program has now rapidly expanded into something much more impactful. Far from just recording traffic violations, these cameras, which are springing up in municipalities across the country, are being used for much wider purposes. Local police departments are reportedly making searches on the Flock database to track down undocumented immigrants on ICE’s behalf, and police in Texas even used it to track down a woman they suspected of self administering an abortion. In the case of the Texas police, they were able to use the Flock database to instantly do a nationwide search of more than 83,000 automatic license plate reader cameras in their search, including cameras in states like Washington and Illinois where abortion is legal.
One of Flock’s main limitations is that it tracks license plates, not people. To solve this, Flock is now also partnering with commercial data brokers, which means that the police will be able to look up not just license plates, but people, using the same shady methods that scam callers do. The ACLU had this to say about this development:
…
Flock is also expanding through a new program called “Flock Business Network” which would make it easier for companies across the country to create blacklists of troublemakers as a way to suppress labor organizing and radical political activity, all under the guise of “preventing crime.” It’s not hard to imagine how this “business network” could be used to help private corporations partner with police departments and government agencies to retaliate against organizers, which is a bone-chilling prospect that the American left should take very seriously.
However, even with all this, we still don’t have the complete picture. The capitalist class, especially in the U.S., is also hoping that AI will be able to give them the edge they need to come out on top in future wars. It is an often cited statistic that the U.S. military is the largest and most expensive standing army in the world, and while this is true, it hasn’t helped U.S. imperialism actually win armed conflicts in the modern era. Trump has essentially lost his war on Iran, and the Iranian government is likely to come out stronger than it was before because of its de facto control of the Strait of Hormuz. Even before that, aid from the U.S. military has been enough to prop up Ukraine in its war against Russia, but it hasn’t been enough to deliver an outright victory to Ukraine, even after almost half a decade of fighting. The U.S. military was able to kidnap Nicholas Maduro through a surgical capture and extraction, but it is suspected that they had help from inside Maduro’s government, and it wasn’t really a test of the full strength of the U.S. military. Even beyond that, there was the botched withdrawal from Afghanistan, which showed the whole world how weakened the U.S. had become. The United States has a large, and extremely expensive military that is proving itself to be profoundly inefficient, ineffective, and cumbersome, especially in an era where comparably cheap ($1,000-$2,500 each) remote-controlled drones are capable of intercepting and destroying Tomahawk missiles, which cost over $2 million each with another $4 million to $6 million for the launchers.
As the U.S. military approaches future wars, which are undoubtedly coming, it will be increasingly looking to AI powered drones as a cheap solution, both defensively and offensively. In fact, it was recently revealed that in 2024 the Ukrainian military successfully tested fully autonomous AI-powered attack drones that were capable of identifying and killing targets with no input from human beings at all. These terrifying and monstrous creations were never widely implemented because the Ukrainian military currently bans AI from being used at the final (lethal) stage of combat instead of a human being. However, some within the Ukrainian government are trying to change that - and even if they didn’t, that rule wouldn’t prevent the U.S. military from implementing autonomous drones with “terminator mode” (that’s actually what they’re calling it) that they could buy from Ukraine. In fact, this would likely be a very appealing option to Trump and co, who likely would want to avoid the mass social unrest that would accompany a reinstatement of the draft. If cheap, autonomous drones can kill more effectively than a human soldier, U.S. imperialism could use fewer soldiers, suffer fewer casualties, and likely have a better chance of preventing a mass anti-war movement from breaking out.
From this standpoint, it becomes clear that the capitalist class doesn’t just see AI as a source of quick profits from increased productivity alone - it also sees it as a vital tool that will be useful for creating a large pool of desperate workers ready to work for poverty wages, union busting, domestic surveillance by police, and military conquest abroad. With this view, we can see much closer to the full picture of how vital AI is for the billionaire class. Unfortunately for them, they have run into a serious bottleneck around vital AI hardware that will not be easily fixed.
AI Data Centers and the Chip Choke Point
All across the country, tech giants are rushing to build massive AI data centers. The reason for this is simple: the rapid increase of demand for AI has outpaced the supply of “tokens” (snippets of generated text generated by Large Language Models). A common misconception is that AI is a fad with no real uses. Unfortunately, companies are absolutely finding applications for AI. To be clear: the AI demand is not being driven by individual use, but by large corporations. In a recent poll, about 40% of people in the U.S. age 18-64 said they had used generative AI before. This is compared to the 71% of companies currently using generative AI - and while we couldn’t find exact figures, it stands to reason that companies are using generative AI far more often and more widely than individuals outside of work do, which means that large corporations are the ones using AI the most, and providing the bulk of the demand for AI tokens. Recently, that demand has rapidly outpaced the available supply of AI computing power.
Over the 3 month period between January and March 2026, weekly token usage quadrupled, and AI companies can’t meet the high demand. Some companies like Anthropic, the maker of Claude, have been forced to start rationing “token usage” (amount of AI generated text) during peak usage hours, preventing heavy use. Amazon has said that “capacity constraints” have limited the growth of its AI, and OpenAI, the developer of ChatGTP, has said that they are unable to pursue some opportunities because they don’t have the available processing power to get the job done. Open AI even recently abandoned a video-generation model due to a lack of available processing power.
This is the reason these AI giants are rushing at a breakneck pace to build data centers anywhere they can: because right now AI isn’t primarily limited by its software, but by its processing power and hardware. These AI giants lack the computing power necessary even for the demand they are currently seeing, and they definitely don’t have enough to expand their operations enough to dominate the market.
However, scaling up their processing power wouldn’t be a simple matter for these AI giants even if they weren’t facing resistance in many places they were trying to build. Increasing their capacity on the scale required is not something that can be done easily or quickly - there are already shortages of transformers, switchgear and gas turbines, and most importantly computer chips. These are long term delays - in some cases equipment delivery can take an expected 2 to 5 years.
The chip bottleneck is significant, and causing big problems. Many AI companies have begun making due with things like video chips not designed for AI because they can’t get the real thing, and are grafting them onto their systems in expensive and inefficient ways because it’s all that is on offer. Many companies are even being forced to repurpose obsolete technology to save money and find a way around chip shortages - chips that are 2 or 3 years old may not sound ancient, but in tech terms they might as well be. Even still, in some cases it’s all these companies can get, even if they’re willing to pay top dollar prices. The price of some chips, such as RAM, has doubled or even tripled in just the last six months because 70% of the global supply of these RAM chips is going to AI companies - this has caused steep price increases in things like computers, gaming systems, phones, TVs, and smart home appliances.
Simply making more chips isn’t a simple matter either - there are only a handful of companies, including Nvidia, which make the chips these AI data centers run on. It is possible to have custom chips made for about half the price of what Nvidia charges, but there are serious roadblocks for those seeking to do it. So far, only Google has been able to create a viable alternative to Nvidia’s chips in large volumes, and even then it took them over a decade to do it. Elon Musk has talked about plans to create a facility that would rival TSMC, a Taiwanese company that dominates the advanced chip-making industry, but the plans would take as much as $13 trillion to make real, which calls into question whether or not it will ever happen. Even if it does somehow get built, it would be many years until a replacement for TSMC in the U.S. would be able to enter the market and increase the supply.
What Happens if the Bubble Pops?
AI giants like Anthropic and OpenAI are expected to lose billions in the next few years - when you combine this gloomy outlook for investors with the massive bubble that has formed, it has led some to predict a coming economic collapse driven by AI. If the AI bubble did burst, it would likely be similar to the telecom bubble crash in the early 2000’s, although in terms of severity it would likely be much worse than not only the telecom bubble, but also the 2008 crash. It’s worth briefly outlining why the similarities to the telecom bubble matter.
In the 1990’s, technological advancements in optical transmission technology made it possible to send huge amounts of data over a single cable by using fiber-optic instead of copper cable. This occurred at the same time as the internet was becoming mainstream and mobile phones and data networks were making massive expansions, which led to huge optimism about the pace of technological advancement and its potential for generating immense profits. The result was one of the largest speculation booms in modern history as investors, companies, and governments all across the world pumped hundreds of billions of dollars into building fiber-optic networks and wireless infrastructure on the belief that demand for bandwidth would grow endlessly. By the late 1990’s, capacity had gone far beyond demand - much of the cable laid was not even being lit and was going unused. The telecom bubble eventually burst around the same time as the dot-com bubble - investors became skittish as interest rates rose and the overbuilt capacity came to light. When the funds dried up, profits also evaporated and global telecom stocks lost more than $2 trillion in market value. Some telecom giants like WorldCom and Global Crossing declared bankruptcy, and tens of thousands of workers lost their jobs. To this point, it was very similar to many other bubbles that popped throughout history. However, the telecom bubble was unique because the massive fiber-optic networks remained in the ground after the crash. This infrastructure was eventually bought up by the companies that remained, and today it’s what makes the modern internet, especially broadband, cloud computing, and video streaming possible.
The AI bubble popping isn’t the outcome anyone wants, but if it did come to pass then whatever companies that were able to survive the crash would make every attempt to buy up the already-built (or even half-built) data centers at bargain-bin prices. The biggest financial firms fully expect to be able to survive these kinds of crashes, and although it may make life difficult for a time, they are prepared to do whatever it takes to come out the other end of a potential collapse with a firm hold on important markets and an eye to the future.
If the AI bubble did pop, it would likely have a temporary chilling effect on workplace organizing, as is almost always the case during the beginning of a big economic downturn. However, the pendulum would eventually swing back with a vengeance when the economy improves or conditions become intolerable for too long. This isn’t to say the class struggle would be totally quiet, even for a time - the movement against Trump’s racist, far-right agenda is far from defeated, and many people are once again looking for ways to fight back at the ballot box while we build a movement in the streets. In some places, there are campaigns like the one for Francesca Hong, a DSA-backed candidate for Governor in Wisconsin, that are running on working class demands with an eye towards building a movement around the campaign. These campaigns, which were inspired by the victory of Zohran Mamdani in NYC, are showing that socialists can enact positive change in office if they have a powerful movement behind them. They are one clear and actionable way we can wage a fight to improve the social safety net and fully fund our gutted social programs, and socialists should fully support them wherever they arise.
Is the Bubble Guaranteed to Pop?
However, an economic collapse is not the only possible outcome. The possibility of a “soft landing” for this bubble is not completely ruled out yet - although it seems less likely at the moment, it’s still possible. Even beyond that, if you assume the bubble will necessarily pop, there are still complicating factors.
The most important viewpoint to consider when trying to make these predictions is actually the perspective of the ruling class, which is to say understanding how AI fits into their plans long term, especially in regard to military conflicts. We are living in a period of history where massive crises like war and poverty are baked in, and for the capitalist class the goal is not necessarily to avoid inevitable crises, but to manage them when they break out. For the capitalist class, a war isn’t a bad thing, as long as they win, and an economic collapse that causes working people to lose their jobs and homes is just another investment opportunity. While some individual capitalists may be ruined in the process of any given crisis, the class as a whole survives and the wealthiest firms and families tend to be fine in the end, barring a revolutionary upheaval. From this viewpoint, AI is not only an important tool, but potentially a decisive tool for any capitalist class trying to manage fightbacks and rebellions in the modern era. Given the applications Generative AI has in the workplace, on the streets, and on the battlefield, it’s not a surprise that some within the capitalist class, including Trump, view developing Generative AI as a strategic necessity.
As we said above, it’s true that AI companies are currently spending more than they are making. It is also true that, under the logic of the capitalist free market, that is normally a recipe for disaster. Individual capitalists mostly act out of their own self interest, not out of the collective interest of their class, and if a company is not delivering a return on their investment, they’ll pull their money. If too many do this, especially at once, the bubble will pop.
However, it is not true that, in this special case where AI represents such a long-term strategic necessity for the capitalist class, they will let it fail. This is not to suggest that the biggest capitalists would endlessly take a loss on AI, but rather that they may turn to other methods of keeping these AI firms alive long enough to ramp up their infrastructure and scale up the technology that already exists.
The most likely avenue that this could take would be through nationalizations, even if they were only carried out partially, and the operation of these AI firms as public-private partnerships or even State-Owned Enterprises similar to what is seen in China today. The capitalist class of China has had success with their particular brand of capitalism, often called State Capitalism, in the modern era. For the capitalist class, one of the major advantages State Capitalism on the model of China provides would be to allow funds to be directed to long term strategic priorities with less risk being assumed by any individual capitalist or investment firm. Instead, the state itself assumes the risk, and the capitalist class enjoys the results. This is not to suggest that State Capitalism has somehow “solved” the crises of capitalism in China - it certainly has not. However, aspects of it would offer the U.S. capitalist a potential solution to the particular problems they could face around AI if the investment bubble pops. It’s no secret that Trump has a history of eyeing aspects of Xi Jing Ping’s regime with envy, especially the large military parades. It is possible, however not guaranteed, that if the AI bubble pops, or comes close to popping, that the capitalist class in the U.S. could resort to nationalizations to keep the AI industry from collapsing and failing to deliver technology they believe will give them an edge in the class struggle at home and imperialist conquest abroad.
We typically like to end articles with actionable advice on what the working class movements in the United States and around the world can do to fight back against the attacks we are seeing today, to drive the movement forward. However, as this article is already very long, we will instead direct readers to a previous article, the entirety of which deals with the question of how we can fight back against AI Data Center construction projects and win, which we hope will be helpful in adding to the conversation.
Nothing to Lose News is committed to accuracy and journalistic integrity.
If you find a factual error, broken link, or other problem in one of our articles, please contact us directly via Substack or email our editor at sophieschollwrites@gmail.com and let us know so we can look into it.
If you want to assist in the effort to build Nothing to Lose News, please get in touch and let us know.