The AI Industry and Its Severe Financial RisksLast week I published an article on the AI dangers that we faced. Although I was quite skeptical of the widespread warnings that super-intelligent AI systems might exterminate humanity, I saw considerable risks of a financial collapse of the AI industry, perhaps dragging down the entire American economy.A recent front-page story in the Wall Street Journal emphasized that our current AI boom has been of unprecedented size. As a percentage of GDP, projected annual investment over the next few years will be more than three times larger than the fabled telecom boom of the Dotcom era, and also more than three times larger than the construction of all our national highways during the 1950s, 1960s, and early 1970s.The day after my article appeared, Reuters revealed some of the leaked financials in the Anthropic IPO prospectus, and those seemed far worse than most had expected.The bullet points included:Anthropic lost $42 billion in 2025AI lab spent $7.33 billion on compute and infrastructure last yearOperating loss widened to $8.06 billion in 2025 from $2.98 billion in 2024Anthropic IPO could value the company at more than $2 trillionSo although 2025 revenue had only reached $4.6 billion and produced a growing operating loss of over $8 billion, the company sought an IPO value of more than $2 trillion, the largest in history. Unless the 2026 financials were dramatically better, such a valuation seemed rather unrealistic.A follow-up article also noted the extraordinary magnitude of Anthropic’s future financial commitments:NEW YORK, Sept 29 (Reuters) – Anthropic said it expects to spend at least $518 billion over a decade building AI infrastructure with six partners, according to a confidential IPO prospectus seen by Reuters — a plan that ranks among the largest AI buildout commitments on record.The AI lab said in the prospectus that about 80% of that sum is non-cancelable or requires payment regardless of usage. Anthropic is telling investors the commitments are necessary because access to computing power is becoming the key constraint on AI development, as future demand for advanced AI systems is likely to exceed available supply and will be “limited principally by the availability of compute.”Ed Zitran has been one of the most caustic critics of the financials of the AI industry, and in an interview a couple of days ago he harshly analyzed those leaked Anthropic results.Video LinkJust before my article ran, a tweet by the SF bureau chief of the Economist noted some of the enormous financial requirements of the entire AI industry.World seems to have woken up to the idea that AI capex actually needs a return at some point. People now realise that roughly $2-3trn of revenue will be needed annually, quite soon.But the conversation about what counts as “AI revenue” is still highly unsophisticated. People will say e.g. “Amazon’s cloud business is doing really well, ergo AI is seeing great ROI”. No. What you need is revenue EXTERNAL or DOWNSTREAM to the system, ie from ordinary businesses subscribing to models. @darioperkins good on this. A lot of Amazon cloud revenue is just Anthropic buying stuffHow much is “proper” AI revenue? In the summer we made four estimates. We need more work to make these estimates better/more accurate! @hsu_steve @lugaricano @weakinstrument— Callum Williams (@econcallum) September 27, 2026On Saturday, the WSJ published an excellent column framing those same industry financials in even starker terms. The economic justification of our massive ongoing AI investment boom required a completely implausible amount of future spending on AI services.Believers in the artificial-intelligence boom need to take a very close look at this number: 9% of GDP.That is how much American businesses and consumers eventually have to spend per year on the services of companies like Anthropic and OpenAI to justify the staggering sums being committed to the technology right now.Is it plausible that Americans will spend as much of their income on this one technology as they do on food? Roughly twice what the nation pays for all forms of energy or all computers and software? Seven times what consumers spend on phone, streaming, and internet services combined?Given such severe financial considerations there have even been some dark suspicions that the widespread fears that AI systems might exterminate humanity were actually concocted and promoted by the AI companies to justify a partial government takeover and bailout.Under such a scenario, the top AI executives and investors might retain much of their current wealth, amounting to billions or tens of billions of dollars, while foisting the trillions of dollars of looming financial losses upon the suffering American taxpayer. This is roughly analogous to what happened to many of the top figures in the Mortgage Bailout of the 2000s, including the former CEOs of Bear Stearns and Lehman Brothers.Towards the end of my article, I included an interview with Dean Baker, a prominent economist who was similarly quite skeptical of the colossal sums being spent on AI infrastructure, which he doubted would ever produce an adequate financial return.He made the additional point that AI systems had already been integrated into much of our society, and if one of those seriously malfunctioned, the popular reaction might set the industry back for many years.For example, if a plane crash or a major highway pile-up were blamed on AI, there would be a widespread call to remove those systems from public use until they had sufficiently matured. Although there have been many plane crashes or highway pile-ups for other reasons, the simmering popular hostility to AI and the associated data-centers might ensure this.As an example, he cited the recent media reports that a hallucinating AI system used by the military had falsely reported that a Chinese ship was carrying nuclear weapons systems to Iran, and we had nearly attacked the vessel before discovering the AI error. According to government sources, this had “almost started a war,” but even if the outcome had merely been a very serious diplomatic incident, the credibility of all AI systems would still have been severely damaged.A reliance upon AI technology has now also been officially blamed for our horrific killing of some 120 young Iranian schoolgirls on the first day of our recent war. But it is widely believed in Iran and among some Western analysts that this is merely defensive scapegoating and that the attack was deliberate, arranged by Israel or its partisans to deeply embitter the Iranians and thereby prevent any early end to the war.Video LinkAlthough AI systems have certainly made great strides in improving their accuracy and power over the last couple of years, I think that this has actually had some negative consequences.Many users increasingly regard AI chatbots as providing oracular wisdom, so they automatically accept those answers without second-guessing, especially if these responses include copious details. Unfortunately, chatbots still do sometimes hallucinate, with potentially very harmful consequences.For these reasons, I usually restrict my own use of AI chatbots to simple and objective factual matters, while being cautious about questioning them on anything that is controversial or complex.I’ve also noticed that most AI systems seem to very strongly support the “official narrative” on most highly controversial historical or political matters. This includes the events or circumstances of World War II, the JFK Assassination, the 9/11 Attacks, the global Covid epidemic, and most recently the Charlie Kirk assassination. The AI chatbots have apparently undergone heavy conditioning, ensuring that they will stubbornly affirm the conventional analysis of all those important events.But even those who recognize this latter problem may remain unaware of the severe unreliability of AI systems on other controversial matters.Read the Whole ArticleThe post Zionism, the Balfour Declaration, and AI Hallucinations appeared first on LewRockwell.