Investor Warns AI Could Soon Overwhelm Economic Systems

WHO:

Paul Krugman | Substack - Azeem Azhar said cheaper open-weight AI models may take on more workloads, while established labs such as OpenAI and Anthropic could continue operating even if they lose market share.

Paul Krugman | Substack - Azhar said executives report that AI adoption is harder than expected, though more companies are succeeding and intend to continue spending on trials.

Futurism - Crypto investor and finance media personality Raoul Pal warned that rapid advances in AI and related technologies could disrupt conventional economic systems.

Futurism - Pal said his post was addressed to his 1.5 million followers on X, formerly Twitter.

Futurism - The article also names financial commentator Harris Kupperman, economist Michael Roberts, the International Monetary Fund, and economist Richard Wolff in discussing concerns about AI investment and economic conditions.

Drudge Report - Drudge Report listed a related story titled “Is AI Boom Depressing Rest of the Economy?”

WHAT:

Paul Krugman | Substack - Azeem Azhar said the internet was “much more porous and much more vulnerable” than it had been two or three months earlier, and warned that AI-assisted vulnerability discovery could create a period in which exploits spread faster than defenses can be deployed.

Paul Krugman | Substack - Azhar said open-weight models can be cheaper per token yet require more tokens to complete a task, potentially making them more expensive overall.

Paul Krugman | Substack - Azhar said it is difficult to detect AI’s effects in aggregate productivity statistics; he cited an economist in China who reported early signs in some sectors.

Paul Krugman | Substack - Azhar said the AI investment bubble metric was higher than a year earlier, while noting that the scale of financing should be assessed against customer spending.

Paul Krugman | Substack - Azhar warned that AI-related “brain rot” and reliance on chatbot responses could weaken critical thinking.

Futurism - Raoul Pal wrote, “The next four years are going to feel very different to anything we’ve lived through,” and described the possible transformation as the “Economic Singularity.”

Futurism - Pal’s post appeared alongside a clip from his “When Shift Happens” podcast and a link to his book, “The Everything Code.”

Paul Krugman | Substack - Azeem Azhar said 70% of websites run on PHP, and explained that the language became part of the internet’s architecture because it was inexpensive and easy to use.

Paul Krugman | Substack - Azhar said users may increasingly access the internet through “walled gardens,” such as Instagram, Facebook, or app stores that restrict which software people can install.

Paul Krugman | Substack - Azhar said AI agents may need validation and user-set permissions; his example was authorizing a bot to spend $50 without requesting approval.

Paul Krugman | Substack - Azhar said Anthropic’s Claude Mythos was available to only a small number of organizations, which were using it to find and patch software vulnerabilities.

Paul Krugman | Substack - Azhar said AI models could uncover severe vulnerabilities in software even after humans had tested it trillions of times.

Paul Krugman | Substack - Azeem Azhar described AI’s growing ability to find software vulnerabilities as a “new regime,” saying models can search tirelessly, discover flaws in novel ways, and chain vulnerabilities together.

Paul Krugman | Substack - Azhar said AI agents’ internet browsing already exceeds human browsing, and predicted machine-agent activity would tilt further toward bots over the coming few years.

Paul Krugman | Substack - Azhar said he expects the AI debate to include outcomes other than a bubble bursting, including very low returns for companies that invested in AI.

Paul Krugman | Substack - Azhar said Anthropic’s pre-IPO S-1 document showed gross margins over compute costs of about 85%; he said open-weight models could reduce labs’ market share while infrastructure companies still earn revenue hosting them.

Paul Krugman | Substack - Azhar said data centers employ 25 times fewer people per $100 million invested than advanced factories.

Futurism - Azeem Azhar said Anthropic’s Claude Mythos was “the best cybersecurity model in the world” and that it crossed a capability threshold at the start of the year; he said open-weight models began approaching its capabilities during the summer.

Futurism - Azhar said AI agents’ share of internet activity is expected to grow, with websites potentially creating structured listings designed for machine agents.

Futurism - Azhar said AI-driven hacking could threaten bank access because deposits are database entries, and that internet-connected power grids could also be exposed.

Futurism - Azhar said Anthropic’s gross margin over compute costs was about 85%, compared with about minus 200% two years earlier.

Futurism - Azhar said a survey-based estimate put the share of large non-tech companies seeing AI benefits at 5% to 10%, with benefits of about 6%.

Truthout - The article said Big Tech uses technological-dominance narratives to obscure financial shortcomings and demands for regulation.

TechRadar - Oracle warned that higher AI usage does not necessarily produce greater business value; companies struggle to identify AI return on investment and manage unpredictable costs.

BBC News - Debt held by data centers is measured in tens of billions of pounds.

Techmeme - The Commanding Heights article said that as AI models become commoditized, lasting wealth may accrue to holders of economic chokepoints, including distribution, payment rails, and a stablecoin dollar, rather than to AI labs.

Paul Krugman | Substack - Azeem Azhar said Anthropic’s Claude Mythos was not widely available and was being used by a small number of organizations to find and patch software vulnerabilities.

Paul Krugman | Substack - Azhar said software is not subject to an agreed security model or safety testing, and described the internet’s software ecosystem as both vulnerable and adaptable.

Paul Krugman | Substack - Azhar said a bank failure caused by a cyberattack could threaten customers’ money because bank balances are database entries.

Paul Krugman | Substack - Azhar said the internet may divide into separate environments for people and AI agents, comparing the shift to the change from pedestrians walking Fifth Avenue in 1895 to automobile traffic by 1920.

Paul Krugman | Substack - Azhar said his team’s AI concert-ticket agent checks for shows weekly, using his music preferences and catalogue.

Paul Krugman | Substack - Azhar said data-center operators depreciate chips over five to six years, and that even eight- or nine-year-old chips were still rented at premium prices amid high demand.

Paul Krugman | Substack - Azhar said chip oversupply could drive down prices, particularly if a more efficient algorithm reduces demand; he compared the potential cycle to semiconductor industry boom-and-bust patterns.

Paul Krugman | Substack - Azeem Azhar said the growth in AI-driven vulnerability discovery could make software flaws more likely to be found and exploited; he cited a recent AI model that found severe vulnerabilities in software that humans had tested trillions of times.

Paul Krugman | Substack - Azhar said the internet may shift toward separate environments for people and machine agents, with publishers and retailers offering agents more structured listings. He said users may have to validate their bots and set rules for purchases, such as allowing up to $50 to be spent without approval.

Paul Krugman | Substack - Azhar said that even if a smaller data-center infrastructure company fails, its chips may be taken over by another operator; some contracts have “interrupted-in-service” clauses that keep customers served.

Paul Krugman | Substack - Azhar said a period of AI investment producing very low returns could occur without a catastrophic market collapse, and warned that productivity gains could lead companies to cut costs affecting workers.

Paul Krugman | Substack - Azhar said the AI bubble metric was higher than it had been a year earlier, but described a possible outcome of very low returns for investors rather than a marketwide financial catastrophe.

Paul Krugman | Substack - Azhar said AI’s productivity gains, if realized, could lead companies to cut costs, with some consequences falling on workers.

Paul Krugman | Substack - Azhar said AI-driven hacking could extend to state-organized operations targeting electricity grids, financial systems, and water systems, and could include information disruption and sabotage.

Paul Krugman | Substack - Azhar warned that workforce changes could emerge gradually through lower wages and fewer new job openings, rather than a single wave of mass layoffs.

Truthout - The article argued that Big Tech’s technological-dominance narratives distract from the industry’s current harms, financial shortcomings, and demands for regulation.

TechRadar - Oracle warned that increased AI usage does not necessarily create more business value; companies are struggling to determine AI’s return on investment and manage unpredictable costs.

BBC News - Debt held by data centers is measured in tens of billions of pounds.

Paul Krugman | Substack - Azeem Azhar said the internet’s software ecosystem has developed without an agreed security model, a practice he linked to software development’s origins in the 1970s and 1980s and the shift from centralized mainframes to personal computers.

Paul Krugman | Substack - Azhar said the Deputy Chief Information Security Officer at Anthropic estimated the cyber threat could bring “a rough 18 months.” He warned that vulnerable systems could include banks and internet-connected power grids.

Paul Krugman | Substack - Azhar said companies including Microsoft and Amazon own GPUs used to run AI models, and that Microsoft and Google had reported insufficient computing capacity to serve all their customers.

Paul Krugman | Substack - Azhar said smaller data-center infrastructure companies could consolidate; he noted that some contracts include “interrupted-in-service” clauses intended to keep customers served if the owning company fails.

Paul Krugman | Substack - Azhar said data centers require land, cooling, and power, and his team estimated that a US and European grid power gap could become significant the year after next.

Paul Krugman | Substack - Azhar described a possible AI investment outcome as a period of very low returns rather than a financial catastrophe, while warning that productivity gains could prompt companies to cut costs affecting workers.

Paul Krugman | Substack - Azeem Azhar said AI-driven cyberattacks could lead to a “rough 18 months” and that the internet had become more vulnerable than it was two or three months earlier.

Paul Krugman | Substack - Azhar said AI agents may outnumber humans online and that bot browsing already exceeds human browsing; he predicted publishers and retailers would develop more structured listings for machine agents.

Paul Krugman | Substack - Azhar said users may need to validate their bots and set spending rules, such as allowing a bot to spend $50 without approval.

Paul Krugman | Substack - Azhar identified risks from state-organized cyber operations targeting electricity grids, financial systems, and water systems, and warned of gradual workforce changes including lower wages and fewer job openings.

Paul Krugman | Substack - Azhar said data-center operators in the UK had agreed to contribute £500 million, about $758 million, to local communities; he cited Loudoun County, Virginia, and Microsoft’s Washington state campus as examples of local commitments.

Paul Krugman | Substack - Azhar said AI data centers employ 25 times fewer people per $100 million invested than advanced factories.

Paul Krugman | Substack - Azeem Azhar said AI models can identify and chain software vulnerabilities, and that open-weight models had approached the capabilities of Anthropic’s Claude Mythos, which was available only to a small number of organizations.

Paul Krugman | Substack - Azhar said 70% of websites run on PHP and cited a recent AI model finding severe vulnerabilities in software that humans had tested trillions of times.

Paul Krugman | Substack - Azhar said 5% to 10% of large companies outside the tech sector may be seeing about a 6% benefit from AI, such as increased sales or lower costs.

Paul Krugman | Substack - Azhar said Anthropic’s gross margin over compute costs was about 85%, compared with minus 200% two years earlier.

Paul Krugman | Substack - Azhar said GPUs were being paid off within roughly two to two-and-a-half years, while infrastructure companies depreciated chips over five to six years.

Paul Krugman | Substack - Azhar said a UK data-center arrangement required operators to contribute $100,000 per megawatt to local communities; the overall amount was about £500 million, or $758 million.

Futurism - Pal wrote that the world may be about four years from the point when “the world as we know it stops working the way it does now.”

Futurism - Pal called the anticipated shift the “Economic Singularity,” saying it could mean people can no longer trust what they see or hear and that “running a business stops meaning what it means today.”

Futurism - Pal identified AI, robotics, energy, and crypto as technologies advancing rapidly enough to outpace systems built for the previous economy.

Futurism - Pal said wages have not kept up with the cost of living and that governments keep debasing currency to fund spending, reducing people’s purchasing power.

Futurism - Kupperman warned that the financial math behind data-center investments is unsustainable; Roberts called AI the “biggest economic bet in US history”; the IMF expressed concern that the world economy is “on a knife’s edge.”

Futurism - Wolff said inflationary pressure is a longstanding feature of capitalism, not a sign of imminent collapse.

Techmeme - A linked Commanding Heights item said lasting wealth from increasingly capable, widely diffused AI may accrue to those controlling distribution, payment rails, and a stablecoin dollar rather than AI labs.

Drudge Report - The headline described economic systems as approaching collapse under the weight of bots.

WHERE:

Paul Krugman | Substack - Azhar cited Loudoun County, Virginia, and Microsoft’s campus in Washington state as examples of data-center operators making local commitments, including measures concerning water bills, electricity prices, and redevelopment.

Futurism - Pal referred broadly to the world economy and “the world as we know it”; no specific location was given.

WHEN:

Paul Krugman | Substack - Azhar said a power gap for data centers was likely to become significant the year after next.

Futurism - Pal described the projected disruption as potentially occurring in “maybe four years”; the article does not specify a calendar date.

WHY:

Paul Krugman | Substack - Azhar said attackers need to find only one vulnerability, while defenders must protect every vulnerability continuously.

Futurism - The article cited economist Richard Wolff’s view that inflationary pressure is a longstanding feature of capitalism, rather than a portent of an imminent collapse.

Paul Krugman | Substack - Azhar said AI-related cyber risks arise because software has developed without a required, agreed security model; he described the internet’s decentralized software ecosystem as vulnerable but adaptable.

Paul Krugman | Substack - Azhar said companies may not see aggregate productivity gains quickly enough to generate revenue sufficient to cover AI infrastructure costs.

Futurism - Pal attributed the possible disruption to exponential advances in AI, robotics, energy, and crypto outpacing institutions and systems built for the older economy.

HOW:

Paul Krugman | Substack - Azhar said he uses an open-weight model for one agent because it follows simple instructions more reliably than a more advanced model.

Paul Krugman | Substack - Azhar said he bought his team fountain pens and encouraged them to work without computers for a few hours each day to protect critical thinking.

Futurism - Azhar said data-center chips were being paid off within two to two-and-a-half years.

Paul Krugman | Substack - Azhar said data-center operators in the UK had to contribute $100,000 per megawatt to local communities under an arrangement he described; the total contribution was about £500 million, or $758 million.

Paul Krugman | Substack - Azhar said he uses an open-weight model for one agent because it follows simple instructions more reliably than a more advanced model, which may do more than requested.

Paul Krugman | Substack - Azhar said he bought his team fountain pens and encouraged them to work without computers for a few hours each day to help protect critical thinking.

Futurism - Pal presented his warning in a post on X, alongside a clip from his “When Shift Happens” podcast and a link to his book, “The Everything Code.”

(UPDATED November 1, 2026 14:56 UTC)

AI models are finding serious software flaws, raising fears that cyberattacks could outpace defenses. Cheaper open models may challenge established labs, though adoption remains difficult. Companies report limited returns as AI investment surges.

(UPDATED November 1, 2026 13:55 UTC)

AI models are uncovering serious software vulnerabilities, raising risks for banks and critical infrastructure. The internet may increasingly serve AI agents alongside human users. Meanwhile, businesses struggle to see returns, and cheaper models could challenge established labs.

(UPDATED November 1, 2026 12:00 UTC)

AI is uncovering software flaws faster, raising cybersecurity risks for banks and infrastructure. Businesses struggle to measure returns, while cheaper models could challenge established labs. Weak investment returns and gradual job pressures remain possible without a market crash.

(UPDATED November 1, 2026 10:55 UTC)

Azeem Azhar warns AI-powered cyberattacks could threaten banks, electricity grids, and water systems. Companies are struggling to prove AI’s business value, despite continuing investment and experimentation. AI may reshape jobs and internet services, while investors face potentially low returns.

(UPDATED November 1, 2026 10:00 UTC)

AI cyberattacks could threaten banks, power grids, and water systems, Azeem Azhar warns. Companies are finding AI adoption difficult, while questions persist about costs and investment returns. As models become cheaper, lasting profits may favor firms controlling distribution and payment systems.

(UPDATED November 1, 2026 05:55 UTC)

AI’s cybersecurity risks are growing, with experts warning attacks could threaten banks and power grids. Companies struggle to measure AI’s value, while investment may bring disappointing returns and workforce cuts. Cheaper open models could gain ground, as businesses keep experimenting despite adoption challenges.

(UPDATED November 1, 2026 03:56 UTC)

Cheaper AI models may spread, but companies still struggle to adopt the technology. Azhar warns AI-powered cyberattacks could threaten critical infrastructure. Data centers raise concerns about debt, power use, and workplace disruption.

(UPDATED October 31, 2026 23:56 UTC)

Raoul Pal warns rapid advances in AI and robotics could disrupt economic systems. Analyst Azeem Azhar says AI could intensify cyberattacks against critical infrastructure. He also warns data-center demands and productivity-driven job cuts could deepen economic strains.

(UPDATED October 31, 2026 22:57 UTC)

Raoul Pal warns rapid advances in AI and other technologies could disrupt economic systems. Azeem Azhar warns AI-powered cyberattacks threaten software and critical infrastructure. He predicts bots will reshape online commerce and AI will gradually change jobs.

(UPDATED October 31, 2026 18:59 UTC)

Investor Raoul Pal warns rapid advances in AI could disrupt economic systems within years. Analyst Azeem Azhar says AI can uncover software vulnerabilities and boost productivity. Critics question data-center investments, while the IMF says the global economy is precarious.

(UPDATED October 31, 2026 16:58 UTC)

Investor Raoul Pal warns rapid technological advances could disrupt the global economy within four years. He calls this shift an “Economic Singularity,” as technology outpaces systems built for the old economy. Experts question data-center spending as economic risks mount.