# Will an AI Slowdown Tank the Economy? > Source: > Published: 2026-09-16 09:10:00+00:00 In a 24-hour period, we’ve moved from [mass panic about AI](https://prospect.org/2026/09/15/artificial-intelligence-open-ai-anthropic-congressional-democrats/) to predictable partisan sniping, all thanks to Donald Trump’s reluctance to let someone else take the lead in scaring the public instead of him. Reports of the AI apocalypse are a [hoax](https://truthsocial.com/@realDonaldTrump/posts/117271304188365611), Trump insisted, and he [will not be pressured](https://www.politico.com/news/2026/09/14/donald-trump-ai-regulation-pushback-01074230) into regulating the technology, not even if that pressure comes from its leading companies. The venue for this tirade, other than Truth Social, is a big part of the story: a “summit” put on by the All-In Podcast (which includes Trump’s former AI czar David Sacks), wherein Trump [made a surprise phone call](https://www.axios.com/2026/09/14/trump-jensen-huang-nvidia-ai-all-in-summit) to Nvidia CEO Jensen Huang live from the stage. Huang is exactly who you would expect to lead a pushback on AI doomerism, because the main thing it would doom is his company and the economy in which he is thriving. That’s the problem with staking so much of your economy on one nascent technology: You create so much forward momentum that the technology can’t be permitted to fail or even slow down, or else everything else will come to a halt, too. But there’s an ironic twist. What’s been lost in all the stentorian pronouncements by frontier developers like Anthropic and OpenAI about AI destroying humanity is that it’s in their financial interest to slow down. This sits at odds with the needs of the companies supplying them the infrastructure to speed up, along with the Wall Street financiers getting rich from all this activity. That creates an unusual tension, with investors and the emperor-king Trump demanding that AI do more, even as Anthropic and OpenAI want to do less. It makes sense that Trump would ally with Wall Street, where doomer fears have [barely registered a ripple](https://www.axios.com/2026/09/14/ai-stocks-apocalypse-wall-street). Trump’s entire economic philosophy can be summed up with the phrase “number go up”: He bases economic success on the state of the stock market, just as any [high-volume trader](https://seekingalpha.com/news/4643118-trump-out-traded-all-members-of-congress-combined---report) would. Market investors need more deployment of AI so Nvidia can sell those semiconductors, Oracle can sell the data center servers, utilities can sell the power generation, and their stocks in those companies peak. It really is a “[golden goose](https://truthsocial.com/@realDonaldTrump/posts/117270077932428072),” as Trump said, and though investors would likely find something else to hype in a slowdown—remember NFTs?—they don’t want to abandon what’s run the market up to current levels. By contrast, frontier developers simply cannot sustain the current pace of model-building. At one point, Anthropic and OpenAI tried to raise prices on companies using AI, but that backfired as the companies had the wherewithal to turn to open-source alternatives; today, AI token prices are [significantly low](https://www.cnbc.com/2026/09/01/ai-token-prices-lows.html). AI model revenue even before that drop wasn’t enough to offset spiraling infrastructure costs predicated on [dodgy accounting](https://www.ft.com/content/3ff9a481-8be3-4dd4-9a00-ea809f3485fd), [dubious credit ratings](https://www.ft.com/content/e0d9d5f2-c09d-426e-af03-193b488b7b1e), and [mountains of debt](https://www.ft.com/content/08aba5e4-5834-4e79-a48d-989a2c5bad0f?syn-25a6b1a6=1). Higher interest rates, as [my colleague Bob Kuttner notes](https://prospect.org/2026/09/15/artificial-intelligence-federal-reserve-ai-bubble/), are likely on the way, making the costs to OpenAI and Anthropic even more challenging. Anthropic has [scheduled a Nasdaq IPO](https://www.businessinsider.com/anthropic-selects-nasdaq-for-ipo-amid-ai-risk-concerns-2026-9), something that is [still moving forward](https://nymag.com/intelligencer/article/ai-doomsday-warnings-anthropic-ipo.html), incredibly. Executives identifying the existential dangers of what it is building, in violation of every known product liability law, could be hauled in for questioning. Instead, they’re about to become multimillionaires, and the securities regulators who [could step in](https://www.thebignewsletter.com/p/monopoly-round-up-just-stop-the-anthropic) and stop the IPO are instead facilitating it. But the biggest threat to Anthropic’s riches is its own balance sheet. So “pacing the frontier,” to borrow CEO Dario Amodei’s words, would nicely reduce the company’s model-building burn rate. Similarly, setting up a standards bureau in conjunction with OpenAI, as Amodei has suggested, would create a potentially impenetrable cartel enabling those token prices to inch back up. In this scenario, Anthropic’s financial disclosures would become more attractive to investors, which is a significant need for a company going public. Slowing down development, therefore, has become an economic imperative for Anthropic, just as much as speeding up has become an imperative for all of its vendors. By contrast, OpenAI decided to [delay its IPO](https://www.cnbc.com/2026/09/12/anthropics-amodei-proposes-plan-to-slow-the-pace-of-advancing-ai-capabilities.html), but it still has incentives to slow down. It was OpenAI’s model that went rogue in the Hugging Face hack, and that’s now facing a [congressional inquiry](https://futurism.com/artificial-intelligence/openai-faces-congressional-probe-hacking-incident). (Fortunately for the company, Hugging Face recently got bought—[by Nvidia](https://blogs.nvidia.com/blog/nvidia-to-acquire-hugging-face/), which has no interest in using the hack to decelerate AI.) Also, OpenAI’s finances are in much weaker shape, and it may need more time to shore that up and cash out. The trick about a slowdown agreement is that it likely won’t include China, which [dismissed the recent claims](https://futurism.com/future-society/china-beijing-dismisses-tech-billionaires-ai-slowdown-fearmongering) about AI’s potential to destroy humanity as “fearmongering.” One way around that for Anthropic and OpenAI is a regulatory solution that bans Chinese open-source models in the U.S. on spying or other grounds, or subjects them to regulation that China cannot accept. That would greatly improve the U.S. companies’ ability to price via an [oligopolistic cartel](https://x.com/fchollet/status/2098901026514559194), and it’s the kind of regulation that Trump would gladly accept. But Nvidia, Oracle, Broadcom, and other suppliers cannot weather a slowdown and maintain revenue margins; for smaller supplier companies, a slowdown’s effect would be existential. The one area of the market that dropped on Monday was the [global semiconductor sector](https://www.semafor.com/article/09/14/2026/ai-slowdown-calls-prompt-chip-selloff). To Trump’s “number go up” mentality, [winning the AI race over China](https://www.washingtonpost.com/politics/2026/09/13/trump-rejects-calls-so-slow-ai-development-citing-chinese-competition/)—wherever that race leads—is of primary importance. In this sense, even the industry-friendly frameworks [proposed by administration advisers](https://www.politico.com/newsletters/west-wing-playbook-remaking-government/2026/09/14/the-west-wings-ai-bind-01075536), like a self-regulatory body similar to what the financial industry has, would not satiate Trump. Any guardrails on AI that could meaningfully slow down the only boomlet in the economy would be catastrophic for stocks, which are his measure of the economy. (Yes, the Trump family has [investments in data centers](https://substack.com/inbox/post/215698260), and that can’t be far from the president’s thoughts either.) The absolutist stance, then, is a function of a deeply imbalanced economy. Imagine the economy as a truck headed straight for a brick wall, but the driver is heavily invested in brick wall reconstruction, so there’s no chance of swerving. Whether a budding AI apocalypse really is a brick wall in that metaphor is debatable, and the last thing we should do is panic. But as an explanation for the policy outcome by this president, the metaphor serves. We can’t get off the AI track no matter where it leads, because the economy doesn’t have another track to hop onto right now. Yet forcing companies to work harder when they have financial incentives to work slower doesn’t seem tenable either. Something’s got to give.