Why the Host Says the Real Economic Warning Is Not AI but the Labor Market

The host argues that the greatest economic warning sign, whether the stock market plunges or the economy slides into a global recession, is not artificial intelligence. He says the video will use GPUs and AI as a lens, while discussing SpaceX as well, to look at the real red flag for the economy.

He dismisses the financing problems of a small company such as Coreweave as a distraction, calling them old news. The video is instead built around a piece by TS Lombard, which the host says he disagrees with on many points.

According to the host, TS Lombard argues that AI spending is recycled spending and that the economy is not yet showing major earnings growth. Instead, the firm describes what it sees as recycled capex, with money circulating among the same companies.

The host illustrates the chain: Google spends $100 billion on chips from Nvidia, Nvidia invests in neoclouds, the neoclouds buy more chips from Nvidia, and the cycle continues into AI startups and hiring. The question TS Lombard raises is where the fundamental expansion of the economy is coming from.

The host partly agrees that the point has merit. Capex has been a major contributor to gross domestic product in the United States and, to some extent, globally, though Europe is not holding up as well. Still, the core issue, in TS Lombard's framing, is where the revenues are coming from.

The host says many people share a similar view that AI is a large bubble. Those who hold it argue that the underlying system looks dead, that neoclouds are unprofitable, and that companies such as Oracle are risky.

They also suspect that there is double and triple ordering of chips. Once that normalizes, they expect a depreciation cliff, and then a capex slowdown because there is not even room to plug in all the available GPUs, which would become an economic problem.

The host says this is a classic bear narrative from one point of view and that it is useful to begin dispelling it, starting with Oracle. He notes that Oracle's net income has risen from $2.9 billion to $4.7 billion.

He argues that while it was previously fair to say Oracle lacked enough money to pay its bills, its margins are now stable. As a result, he says, Oracle is not really on the brink of default. He adds that he does not own Oracle and is not promoting it.

The host says Nebius is not as bad as it appears. He reports that its revenue has grown to about $582 million. The only reason it shows an operating loss, he says, is a $259 million write-off on chips.

He argues that if depreciation is added back, Nebius is actually profitable. He again states that he has no position in Nebius. His conclusion is that neither Oracle nor Nebius is likely to trigger a collapse of the AI bubble.

The host argues that the real indicator for AI is not Oracle or Nebius but what happens to chips roughly six and a half years old and to the depreciation cliff. He notes that Michael Burry and others, including himself, have long said chips should lose value over time, as planes do.

However, he says that with the recent surge in popularity of open-weight models, chips that are 6.5, 4.5 and 2.5 years old are becoming more valuable rather than less. If Oracle and Nebius are not falling off a cliff and older chips are appreciating, then the depreciation cliff is not the issue, he argues. He concludes that the depreciation cliff thesis is premature.

The host presents a sponsored segment for Gigastar, which he says lets creators sell a slice of their channel's future YouTube revenue. The shares are called channel revenue tokens, or CRTs, and the host describes them as securities rather than coins or crypto.

He says Gigastar is registered with the SEC and FINRA and has built its own trading platform, so buyers can trade after 12 months. The host notes that the company itself is raising money on Wefunder, that this is an investment in the startup rather than in any channel, and that every investment carries risk. He states that his company received $4,000 for the placement.

The host cites an interview in which a Wall Street finance figure said that if AI were going to take people's jobs, it would have already done so. The host says he disagrees and argues that such a person is blind to how powerful agentic AI has become.

He points to GPU rental rates at neoclouds and at SpaceX. He says SpaceX's AI-related revenue from renting out chips roughly tripled, and that its prepayment deposits rose a further 35%, which he presents as evidence of hyperscaler demand.

He also argues that a 6.5-year-old chip should be declining in value, but its rental rate rose in 2026. He attributes this to open-weight models enabling more agentic compute on older chips, and to a lack of enough chips for demand. According to his figures, rental rates for these chips rose from about 135 to about 152 since December, an increase of 12.6%.

The host notes that the H100 came out in September 2022 and declined substantially as newer technology arrived. He calls this the normal curve. He says he agrees with Michael Burry that depreciation pain is coming, but then asks what is happening in 2026 that is pushing up the rental rates of older chips.

He says those rates are now up 15% from the fourth quarter of 2025. He attributes the rise to productivity gains from AI that are only now being realized. He also mentions chips that are 4.5 and 2.5 years old, and says that Burry's timing has been wrong.

The host says that rental rates for GB200 Frontier Blackwell chips, which exclude the GB300 and Vera Rubin, are up 39% from the end of 2025. He presents this as further evidence that AI is not slowing down.

He cites ASML, a leading indicator for chip buildout, saying its backlog is exploding and has nearly tripled for machines that cost about half a billion dollars each. The host concludes that this is not a sign that AI is rolling over, and that TS Lombard is wrong to say AI is what will kill the market.

The host turns to unemployment, citing an argument by the New York Times and USA Today that a rise in the unemployment rate for Black workers can be an early signal of trouble for everyone. The Black unemployment rate has just recorded its fastest increase since 2012.

He notes that this is the largest one-month increase since 2020, excluding COVID. He explains that Black workers are often the last people hired and the first fired, so pressure on their employment can signal broader weakness.

The host says Netflix announced layoffs of 6% and then 9% of its workforce over the past couple of days. He notes that there are about 153 million workers in the labor market, so layoffs always happen.

His concern is that at some point the economy's ability to absorb these layoffs will disappear. He says that, in his opinion, labor productivity from AI is about to skyrocket, which would be bad for many workers.

The host cites a research firm whose name he says he does not know how to pronounce. The firm argues that the September jobs report disappointed, with payrolls rising just 29 thousand. It says the labor market is shakier than the headline numbers suggest, though things are broadly okay for now.

The firm also argues that the Federal Reserve's current hiking cycle will likely see fewer rate hikes because of the labor market. The host links this to comments by Kevin Warsh about tail risks, and notes that the ADP report showed pain in professional and business services, while construction, driven by data centers, is the main source of growth.

The host says his analysis leads him to increase his bet that the Fed may deliver only one rate hike this year. He argues that the Fed is keenly aware the labor market is not showing wage inflation.

He says the Fed cannot hike to stop the AI or oil buildout, and can only control inflation expectations, which he says are already stable after the last hike. He says he holds a few hundred dollars on exactly one hike, and that a deal with Iran could double the odds of that outcome to 50%. He also expects softer labor market data in November or December to reduce the likelihood of another hike.

The host describes using AI to cut a family video. He says about 95% of the cuts in the video were made by AI. A task he estimates would have taken about an hour by hand took about three minutes.

His workflow was to upload the footage with a prompt, return about 20 minutes later, make small tweaks, and export. He says this means such jobs no longer require a video editor.

The host says the change in editing he described did not exist three months ago. He argues that the damage to the labor market is not happening now but is coming, and that the labor market is the real canary in the coal mine to watch.

He stresses that he is not saying to sell everything or that the economy will crash. He expects AI investment to keep booming and predicts a euphoric stock rally after an Iran deal, with labor market data rolling over later. He says early signs of this are already showing and will likely get worse, which is his greatest concern.

The host says the key question is whether AI revenue comes from businesses and end users outside the system to pay for productivity-enhancing tools, or from capex recycling. He thinks it has mostly been capex recycling but is turning into productivity tools.

He argues that these tools will hurt the labor market at first with layoffs, though in the long term the labor market will create new jobs as in past industrial revolutions. He says part of his ReinvestOne launch will include agentic AI modules for businesses and consumers, available at the beginning of 2027.

The host promotes ReinvestOne lifetime access, which he says is available as a single payment. In the closing exchange, someone asks why he does not advertise these points more, and he says the team will try a little advertising to see how it goes.