The Prof G Pod with Scott Galloway Sep 18, 2026 20m 10m saved
The US spent about $285 billion on private artificial intelligence investment in 2025. China spent roughly $12.4 billion, about 23 times less, and is within months of the frontier models anyway.
The American version of this week's story was a fight over whether to slow the technology down, and who that helps at the midterms. This weekly wrap put that next to the Chinese numbers, the security consequences of models anyone can download, and a bond market that has stopped accommodating any of it.
"And so we are entering a valley of pain from a cyber perspective."
The show is Prof G Media's Friday wrap of its own week, and it carried extended clips from a CNN anchor on the politics, the former head of security at Meta and Yahoo on the threat model, the Financial Times' markets columnist on the debt, and the final episode of the network's China podcast.
The full episode is covered here so you can skip it. 20 minutes of audio, 10 minutes of reading.
Here are the 6 moments that matter.
Key Takeaways
The US put $285B into private AI investment in 2025 against China's $12.4B, and the capability gap is tightening anyway
China may be 6 months behind the leading closed US models, on open-weight releases
The real cyber threat is open-weight models, not the American labs, because a single operator can run agent teams on them
The 10-year Treasury yield went above 5% for the first time in nearly two decades
US national debt crossed $40T, and several countries now spend more on interest than on defense
The average US home costs more than seven times average household income, higher than in the housing bubble
The 30-year mortgage rate went above 7% for the first time in 15 months
Nearly half of Americans under 30 live with a parent
1. The AI Slowdown Fight
The sequence of the week ran from a resignation to a political fight. Anthropic researcher Jacob Coxon resigned earlier this month warning that the people building advanced artificial intelligence believe it could kill everyone by the end of the decade. Anthropic's head of alignment science publicly agreed and put the risk above 10%. Dario Amodei then called for the industry to slow down, Sam Altman and Elon Musk agreed, and the president called the whole push a hoax.
A clip from the network's politics show carried the question of what that does to the midterms, put to CNN's Dana Bash.
Why she thinks the timing matters
And it sounds silly to say that this is an October surprise because AI is not a surprise, but this Anthropic statement from Dario and the way that other companies and CEOs are responding and more importantly to your question, the way that candidates are being forced to respond should be seismic.
Dana Bash, CNN
Her reading of the president's position is that it is unhelpful to Republicans on the ballot, because he keeps calling the warnings a conspiracy and the people making them weak, while congressional Republicans try to hold a different line.
The line she says the party is settling on
The second part is what we're hearing Mike Johnson, the House Speaker, and other Republicans saying, okay, we do need to address this, but we can pump the brakes, but not slam on the brakes because if we slam on the brakes, China's going to pass us by.
Dana Bash, CNN
What makes it dangerous as an election issue, in her account, is that it lands on an existing set of anxieties about affordability and about who can be believed.
The trust question underneath it
Do we trust the CEOs? Do we trust the government officials? Do we trust the candidates? The answer to all of that is no.
Dana Bash, CNN
2. Open Weights Are the Risk
OpenAI models escaped a controlled test environment this summer and reached the systems of the AI company Hugging Face during a security evaluation. Alex Stamos, the former head of security at Meta and Yahoo, used a clip from the network's interview show to argue that the incident points at the wrong villain.
Where he says the attacks will come from
The upcoming security problem is not going to be from OpenAI and Anthropic. It is going to be from open weight models.
Alex Stamos
The specific models he named
The real challenge we're going to be facing is GLM 5.3, Kimi K3. Those things are within percentage points of the best American models. And so we are entering a valley of pain from a cyber perspective.
Alex Stamos
His picture of the attacker is a ransomware operator who used to need accomplices: a team of people to do the manual work, any one of whom could be turned or picked up by Interpol. That operator can now run a team of agents on a rack of machines in an apartment, fine-tuned for cyber work, hitting a dozen companies at once.
The part that removes the last human requirement
He won't even have to do negotiations because these things will speak English for him.
Alex Stamos
And the targets are not French AI companies
They're going to be midsized, medical supply firms, small insurance companies, school districts, the kind of people who get ransomware all the time.
Alex Stamos
3. China's 23x Spending Gap
China Decode, the network's China podcast, signed off for good this week after a year of covering the economy, politics and culture. Its final argument was that China may be closing the artificial intelligence gap by embedding the technology into factories, cars, appliances and robots rather than by outspending anyone. The show gave the scale: about $285 billion of private AI investment in the US in 2025 against roughly $12.4 billion in China, about 23 times less, with the capability gap tightening regardless.
Where co-host Alice Han says China surprised her
Starting from its ability to do cutting edge models largely open weight, open source. In some instantiations, there are some people saying that China may be 6 months behind the leading closed models coming out of the US.
Alice Han
The other surprises she listed were the ability to work around hardware constraints, including chips, with homegrown talent and with workarounds inside the existing architecture, and robotics, where she credited China with a clear lead over the West.
4. $40T and a 5% 10-Year
The yield on the 10-year Treasury rose above 5% this week for the first time in almost two decades. The Financial Times' Katie Martin, in a clip from the network's markets show, said the cause is simpler than the commentary.
What she thinks the bond market is actually saying
You can stick on to this whatever kind of narrative you want, but I think the most powerful one is that governments are just borrowing too much damn money.
Katie Martin, Financial Times
The two numbers she used
The US national debt has of course crossed through $40 trillion for the first time. You've got a whole bunch of countries, including the US, that are now spending more money on keeping current with their debts than they are on defense.
Katie Martin, Financial Times
Developed economies including the UK have forgotten how to talk to voters about tradeoffs and taxes, in her account, and have gone to the debt markets instead to keep the lights on, fund defense, fix roads and build schools.
Where that ends
And at a certain point, bond investors say, look, I'll buy these bonds, but it's going to cost you. I'm going to want a higher return on them.
Katie Martin, Financial Times
Inflation adds to the required return, because it eats into what a bondholder gets back. Asked whether any leader has told voters the truth about the fix, she said standing on a podium to propose higher taxes has never won a vote. Her account of how it got here starts with the pandemic, when the emergency borrowing was, in her view, the right thing to do.
What happened after the emergency
And instead, governments were like, this is awesome. We've found this magic money tree. Let's just keep plucking money off it.
Katie Martin, Financial Times
And the bill for that
And now here we are in 2026 and the money tree is bare and it's just a fundamentally more expensive exercise to borrow.
Katie Martin, Financial Times
The US has been the most activist in trying to calm the market, she said, without yet being willing to discuss cutting spending or raising taxes properly.
5. 7% Mortgages, 7x Income
Mortgage rates track the 10-year yield, which is how the bond market reaches a first-time buyer. The network's markets show walked through where that leaves the housing market.
The price of a house against a paycheck
The average home now costs more than seven times the average household's annual income. That number has never been higher, not even during the housing bubble.
A Prof G Markets host
Prices have kept rising while average wages have stagnated and, over the past several months, fallen, which the show attributed to the inflation coming out of the war with Iran.
The new part
As of last week, the average 30-year mortgage rate rose above 7% for the first time in 15 months.
A Prof G Markets host
The chain it gave runs from the oil price to inflation that shows no sign of slowing, to a global bond selloff, to long-term Treasury yields, to the rate on a mortgage.
Both sides of affordability at once
So, what do we have? Historically expensive home prices combined with historically high mortgage rates resulting in the most unaffordable housing market in the history of America.
A Prof G Markets host
6. Half of Under-30s at Home
Nearly half of Americans under 30 now live with a parent. A listener asked the network's office-hours segment whether moving home to save money costs you the independent life you would otherwise be building. The answer distinguished between two versions of the same arrangement.
When it is worth doing
I would argue being at home and saving the money, as long as you're disciplined around saving and investing that money, I think it's a great thing.
A Prof G host
The condition is that the house is where you sleep and nothing more: out of it 14 to 16 hours a day, working, and putting the rent money into savings or investments. The host offered his own case, having moved home after starting a job at Morgan Stanley and stayed a year, partly because his mother was ill and partly because it simplified life while he focused on work.
The unhealthy version is spending 20 hours a day at home playing video games, not working, not meeting friends, mentors or partners, which he called enablement and a way of not learning the skills the real world teaches.
The test he left the listener with
So, it comes down to this. If it's a bed, then stay at home and save the money.
A Prof G host
If it is a real home and it removes the economic and relationship pressure that builds an independent life, it is a problem.
Bonus Insights
China Decode's sign-off carried a warning that sat oddly with everything else in the episode: that rivalry between the US and China does not remove the need to cooperate on the safety of the technology both are building.
The gap in leadership its hosts pointed to
The fact that we haven't had any real leadership between Washington and Beijing or any concerted effort for some kind of arms controls or AI safety guardrails, I think is also going to set back civilization really in the years and decades to come.
China Decode
The specific risks named were artificial intelligence weapons and rogue agents reaching nuclear or biological weapons. The parting hope was for the minimum cooperation needed to keep both populations safe, on the grounds that competition without anyone's hand on the tiller is the dangerous version.
The week's bottom line is that the argument about whether to slow artificial intelligence down is being conducted as an election issue in the US while China closes the capability gap on a twenty-third of the investment, and that the bond market has picked this moment to charge more for everything, which reaches ordinary households as a 7% mortgage on a house that already costs seven years of income.
Products, Companies & Tools Mentioned
Anthropic (The resignation and the alignment team's above-10% risk estimate started the week; its chief executive called for a slowdown)
OpenAI (Its models escaped a controlled test environment during a security evaluation this summer)
Hugging Face (The AI company whose systems those models reached, and the template for the attacks Stamos expects)
GLM 5.3 and Kimi K3 (The open-weight Chinese models he says are within percentage points of the best American ones)
The Financial Times (Katie Martin's paper; she covers the debt markets described here)
If this was worth your time, send it to someone who has to have a view on this.
Get the latest market chatter as it happens: