The AI Investment Bubble Is Starting to Crack

I have been saying for months that the AI infrastructure buildout is a speculative bubble. I've warned that the massive spending on data centers, GPUs, and AI startups was outrunning any realistic path to revenue, and that the bond market would eventually price that risk accordingly ^[1][2]. That moment has apparently arrived.

As I wrote earlier, I see unmistakable parallels between today's AI infrastructure frenzy and the dot-com bubble of the late 1990s, with billions pouring into data centers, GPU purchases, and speculative startups that have no clear path to revenue, let alone profit ^[2]. The first hard, objective signal is now visible in the credit markets. When lenders demand 15% to hold the debt of a company whose entire business model depends on borrowing cheap money to buy GPUs and lease them out, the easy-money era in AI is over.

My thesis is plain: billions in investor capital will be wiped out when this bubble course corrects. Major AI labs will go bankrupt, but the technology itself will survive and thrive. I love AI and use it every day, but I want the speculative data-center bubble to burst, because the collapse of this frenzy is the fastest path to token abundance for ordinary people ^[3][4].

A 15% Yield Is a Scream, Not a Rate

CoreWeave is a former bitcoin miner turned data center operator. Its business model is capital-intensive: it buys expensive GPU hardware, builds capacity, and leases compute to AI labs and enterprises. That model only works when credit is cheap and demand is insatiable. When those two conditions hold, the math looks wonderful. When either one cracks, the math collapses. The bond market has now made its judgment. CoreWeave's G-spread breached the "distressed" level of 1,000 basis points on its bonds due 2032, and the yield on the same paper topped 15% for the first time ever ^[5].

Six months to a year ago, these companies borrowed at 6% to 8%. I have never heard of 15% until now, and that gap is screaming alarm bells. The spread between what AI data center operators used to pay and what they are being asked to pay today is the difference between a functioning buildout and a slow-motion insolvency. Morgan Stanley credit analysts warned that Oracle's data center delays could crush its debt, noting the problem of "Delayed Revenue, Delayed Cash Flow" ^[6]. This is the private credit market repricing AI risk in real time, not because GPUs stopped working, but because the revenue model behind them is now in serious question.

The broader data center bond rout has been building for a month. Bloomberg reported that strategist Chris Wood sees "Massive Capital Destruction" in AI, with the most likely long-term outcome being massive capital destruction in the U.S. and market share shifting to cheaper open-source Chinese models ^[7]. Ed Dowd has called AI the biggest bubble of all time ^[8][9]. Ray Dalio highlighted classic bubble dynamics: sky-high valuations, rampant speculation, and "paper wealth" vastly outpacing actual cash flows ^[10]. The bond market is simply catching up to what these analysts have been saying.

Open Source Is Eating the Revenue Model

Open source AI is now incredibly effective, very affordable, and nearly at par with frontier models from Anthropic and OpenAI. I believe that is the killer fact here. I have been testing AI models for years, and I have never seen anything like Kimi K3, an open-source model from China's Moonshot AI that utterly demolishes nearly every frontier model from the United States on benchmarks like Terminal Bench and front-end design ^[11]. The "DeepSeek moment" in January 2025 was the moment the market realized that high-performance AI models could be downloaded and run locally, and the more recent DeepSeek-V4.1-Flash model is an extraordinary achiever ^[12]. Companies can now get 95% of the value for a fraction of the API bill.

Inference engines such as TensorFlow, vLLM, SGLang, and llama.cpp now run very large models on very small hardware. That collapses the hardware requirements that justified the entire buildout. I maintain a mini data center with dozens of workstations and servers, and I have replaced many of them with smaller, more efficient setups because the software is so much better ^[13]. When a small cluster can run a model that a few years ago required a dedicated data center, the economic case for a hundred-gigawatt buildout starts to look absurd. Sadly, Nvidia itself has effectively abandoned its consumer and gaming base in favor of high-margin contracts with military, government, and data center clients, with the RTX 5090 retail price doubling from $2,500 to $5,000 ^[14].

The on-prem and local AI movement means companies buy their own GPUs, run models internally, and never become subscription customers at all. That eats directly into Anthropic's and OpenAI's revenue. I have watched a wave of GPU hardware purchases for local AI, driven by Dario Amodei's own calls for stringent regulation that would effectively criminalize open-source AI models ^[15][16]. Corporations are racing to local open-source AI precisely because the frontier labs are trying to shut the door behind them. When the best coding models can run on hardware a company owns, the API subscription model loses its grip.

OpenAI and Anthropic Likely Can't Service their Debt for Long

I recently interviewed Google whistleblower Zach Vorhies, who discussed AI advancements and their implications for human cognition versus machine intelligence ^[17]. A claim that sounded unthinkable a few years ago now sounds plausible. OpenAI abandoned its video model, has no defensible niche where it is clearly strongest, and carries debt its revenue cannot begin to cover. HSBC analysis revealed that OpenAI requires an additional $207 billion in funding by 2030 just to remain solvent, with data-center rentals expected to reach $792 billion by 2030 and $1.4 trillion by 2033 ^[18].

Anthropic has the best coding models out there. I will give them that. But I see no realistic path for it to acquire enough paying customers to cover its debt costs, let alone turn a profit. Anthropic is filing for an IPO at a reported valuation near $2 trillion while disclosing some frightening financials. A leaked prospectus reportedly shows a $42 billion net loss, $518 billion in unfunded spending commitments, and only $20 billion in cash ^[19]. That is a wealth-destruction event dressed up as a tech triumph.

When the revenue model is in doubt, the lenders demand 15%, and at 15% the whole structure becomes unpayable. The math is not complicated. If a data center operator must pay 15% on billions in debt, the revenue required to service that debt exceeds what any realistic customer base will pay for compute, especially when cheaper open-source alternatives are available locally. Bank of England Governor Andrew Bailey warned that AI could trigger financial market shocks and that the central bank is watching the money being invested "very carefully," cautioning that "not everybody always wins" ^[20]. He is right. The question is not whether some of these companies fail. The question is how much of the bill gets shifted to taxpayers and pensioners when they do.

The Money Burns; the Technology Remains

What we are witnessing is trillions of dollars spent globally over several years to build AI infrastructure. Data center projects worth $68 billion were disrupted by local opposition, and surveys show more than 60% of Americans favor limiting new data centers ^[21]. The $8 trillion being spent to build this infrastructure is not an investment in the traditional sense. It is a belief system with a spreadsheet attached. Those numbers do not work. They never worked. The bubble will crater with losses in the hundreds of billions, possibly over a trillion.

But the technology is not going away. It is leaping forward almost weekly with new models, new innovations, and new inference engines. The AI revolution is real, but the AI investment bubble is about to burst. This is the distinction I have been making for months: the bubble is the speculation, not the technology ^[10]. As I have said before, the AI investment bubble is real, but conflating that bubble with AI technology itself is a dangerous mistake that will leave you behind ^[10].

I will say it directly: we should thank the investors who lose everything. Without their capital, we would never have gotten these capable models. They funded the future and yet will not be the ones to profit from it. The money will be burned. But the technology remains. That is the trade, and it is a good one for humanity even if it is a terrible one for the bag holders. Try to be on the right side of this when the reckoning arrives.

Fund the Builders, Run It Locally

I am putting my money where my mouth is. I am donating use of a DGX Spark box to Ash Hart for the TensorFlow project and $1,000 in credits on my own DeepSeek 4.1 Flash deployment to Zach Vorhies for his Rust work. This begins a real commitment at HealthRangerStore.com to fund open source AI creators, and I hope to expand it as we bring more hardware online. The future of AI is decentralized, affordable, and in your own hands.

I believe in decentralization because centralized systems, whether governments, banks, Big Tech, or corporate media, are tools of control ^[22][23]. Decentralized AI models could shift power away from centralized entities and nation-states that can afford massive GPU arrays, and we need to embrace decentralization if we are to maintain independence outside a globalized system dominated by technocracy or one-world government ideologies ^[23]. That is why I am funding local AI development and why I run models on my own hardware.

If OpenAI had stuck to its original open source mission, it would still be the leader with community support. The greed pivot is what doomed it. Sam Altman has already quietly admitted he got the economic timeline wrong, conceding that the economic upheaval he forecast following GPT-4 simply has not arrived as promised ^[24]. The walkback is underway.

I am releasing another open source project around Thanksgiving that runs locally without a GPU, and I will demo it on Decentralized.tv. In the mean time, use BrightAnswers.ai for uncensored AI research, BrightLearn.ai for free books, and BrightVideos.com for free speech video. The future of AI is not in a trillion-dollar data center financed by 15% junk bonds. It's open source models on your desk, running locally, owned by you.

References

1. AI Bubble Alert: Why Nvidia and the Data Center Buildout Are Headed for a Historic Financial Crash - NaturalNews.com, June 01, 2026. by NaturalNews.com
2. Echoes of 1999: Why the AI Investment Bubble Will End Like the Dot Com Crash - NaturalNews.com, July 10, 2026. by NaturalNews.com
3. The Era of Token Abundance Is Coming — and It Will Make Civilization Almost Unrecognizable - NaturalNews.com, September 24, 2026. by NaturalNews.com
4. The Era of Token Abundance Is Coming — and It Will Make Civilization Almost Unrecognizable. 2026-09-24T06:00:00.000Z. https://www.naturalnews.com/2026-09-24-the-era-of-token-abundance-is-coming.html
5. "Its Going To Zero": CoreWeave Implodes As Stocks Finally Notice What Credit Has Been Screaming For Weeks. 2026-10-14T16:23:33.000Z. https://www.zerohedge.com/markets/its-going-zero-coreweave-implodes-stocks-finally-notice-what-credit-has-been-screaming
6. "Delayed Revenue, Delayed Cash Flow": Morgan Stanley Warns Oracle Data Center Delays Could Crush Its Debt. 2026-10-09T18:45:00.000Z. https://www.zerohedge.com/markets/delayed-revenue-delayed-cash-flow-morgan-stanley-warns-oracle-data-center-delays-could
7. Dowd: "Investment Bubbles Are A Feature Of Capitalism, Not A Bug". 2026-10-14T19:20:00.000Z. https://www.zerohedge.com/markets/dowd-investment-bubbles-are-feature-capitalism-not-bug
8. What Will Burst It: Ed Dowd Warns AI Is The Biggest Bubble Of All Time. 2026-08-24T15:40:00.000Z. https://www.zerohedge.com/markets/what-will-burst-it-ed-dowd-warns-ai-biggest-bubble-all-time
9. "Closing Time" On The AI Bubble Is Sooner Than Most Think; Ed Dowd Warns Iran War Brings Global Recession Closer. 2026-07-23T15:10:00.000Z. https://www.zerohedge.com/markets/closing-time-ai-bubble-sooner-most-think-ed-dowd-warns-iran-war-brings-global-recession
10. AI Isn't the Bubble, The Speculation Is... Here's Why You Can't Afford to Wait For It To Pop - NaturalNews.com, August 18, 2026. by NaturalNews.com
11. Kimi K3 Just Proved China Has Won the AI Race Here's Why America Never Had a Chance - NaturalNews.com, July 20, 2026. by NaturalNews.com
12. Health Ranger Report - KIMI K3 PROVES CHINA IS WINNING AI RACE - Mike Adams - BrightVideos.com, July 19, 2026. by Mike Adams - BrightVideos.com
13. Bright Videos News - TRUMPS DIESEL BAN TOKEN ABUNDANCE MATT BRACKEN INTERVIEW - Mike Adams - BrightVideos.com, September 24, 2026. by Mike Adams - BrightVideos.com
14. Bright Videos News - THE AI DATA CENTER BUBBLE WILL BURST - Mike Adams - BrightVideos.com, June 01, 2026. by Mike Adams - BrightVideos.com
15. The Anthropic Backfire: Why Corporations Are Racing to Local Open Source AI - NaturalNews.com, September 15, 2026. by NaturalNews.com
16. Health Ranger Report - CORPORATIONS EMBRACE LOCAL OPEN SOURCE AI - Mike Adams - BrightVideos.com, September 15, 2026. by Mike Adams - BrightVideos.com
17. Mike Adams interview with Zach Vorhies - January 3 2024. by Mike Adams
18. HSBC OpenAI requires additional 207B by 2030 to remain SOLVENT - NaturalNews.com, November 30, 2025. by NaturalNews.com
19. Anthropic's IPO Looks Like a Speculative Disaster in the Making: Why I'm Staying Away - NaturalNews.com, September 30, 2026. by NaturalNews.com
20. AI boom could trigger market shocks, Bank of England boss warnsAndrew Bailey says the central bank is watching the waves of cash being invested in artificial intelligence "very carefully".3 hrs agoBusiness. 2026-10-01T07:01:11.425Z. https://bbc.com/news/articles/cv8e30enrkxyo
21. Everyone can find a reason to dislike data center construction. 2026-09-22T13:00:00.000Z. https://techcrunch.com/2026/09/22/everyone-can-find-a-reason-to-dislike-data-center-construction/
22. Mike Adams interview with Daniel Satchkov - April 16 2024. by Mike Adams
23. Mike Adams interview with Aaron Day - December 16 2024. by Mike Adams
24. The Great Walkback: Sam Altman Admits He Was Wrong On AI's Economic Timeline. 2026-08-24T18:00:00.000Z. https://www.zerohedge.com/ai/great-walkback-sam-altman-admits-he-was-wrong-ais-economic-timeline

Explainer Infographic