Investing in defense tech used to be taboo in Silicon Valley. Those days are over.  "Probably the number one factor is just the advances in AI technology. It's just time for the category because the technology is working. To paraphrase Kyle Reese in the Terminator movies, the machines just got smart... But in this case, it works better than human pilots, and it works better than what's on the ground right now." -@AmanVerjee on Trading Places Podcast EP49 with @davemcclure

The 10-year treasury just crossed 5% for the first time in nearly 20 years. There are several possible reasons why. "It's just economic growth is creating demand for capital. If it were the budget deficit, if it were something specific to the US, you would expect the dollar to be weakening. But...it's generally been rising. It's sort of flattened over the past couple of years. It actually substantially increased just after Trump's election and then has been normalizing, but doesn't seem to be losing ground against major trading currencies." -@AmanVerjee on Trading Places Podcast EP50 with @davemcclure

Why Oura pulled its $15B IPO at the last minute: "They're not a SaaS business with SaaS margins... When I was CFO of Sonos, we had best in class retention for a hardware company and we were in the mid-eighties and I suspect Oura is in the mid-eighties. I think with those dynamics, it's awfully hard to get a 10X valuation, which is like a premier SaaS multiple." -@AmanVerjee on Trading Places Podcast with @davemcclure

"I do feel like there's margin pressure and revenue growth pressure on Anthropic."-@davemcclure A draft of the Claude maker's prospectus leaked on September 28. What does it mean for their upcoming IPO? It showed about $4.6 billion in 2025 revenue and a $42 billion net loss. The new number was $518 billion in cloud and compute commitments. About 80% of it can't be canceled or is on take-or-pay terms, so Anthropic owes that money whether or not the revenue arrives. The company is aiming to list in November at up to $2 trillion, which would edge out SpaceX as the largest IPO in history. Dave McClure and @AmanVerjee break it down on Trading Places Podcast EP50.

The AI safety issue has been all over the news recently. Many think it's real. Some think it's 'Big AI' posturing. @AmanVerjee studied the issue, particularly the recent OpenAI hack of Hugging Face, and came away thinking it's a mix of both. AI exhibits remarkably human-like hacking capabilities, including engaging in blackmail and collusion. In fact, Anthropic devoted an unusually large chunk of their recent leaked S1 to safety issues. But that doesn't mean that the AI giants won't benefit from pushing regulation. As Aman put it, they may be engaging in a "Machiavellian play towards regulatory capture." How real is the AI safety issue? Aman and @davemcclure break it down.

Two-minute map of the entire AI industry: "At the very bottom you have the guys who provide power to the ecosystem. Above that you got the chip makers... like NVIDIA. The chip makers have probably something like 10 trillion dollars in market cap right now. The power guys... the entire S&P 500 is under three trillion in market cap right now for the power players. And only a subset of those are powering the AI revolution. Then the cloud guys... the hyperscalers are folks like Amazon and Alphabet, Oracle, Azure, and then Meta... as well as SpaceX. And then you've got some of these neoclouds, folks like CoreWeave or Grok, who are specialists in just supporting AI. Here's the higher end of the stack. We've got maybe five trillion or so of market cap. A lot of this is the closed source frontier models in the US, so Anthropic, OpenAI. Then you have the open weight models, many of whom are Chinese right now. Note that there's not that many companies at the top of this pyramid, the application layer... Harvey among them. There's Replit, there's Cursor. But if you add up the market cap of all those companies today, it's only about a trillion bucks." -@AmanVerjee on Trading Places Podcast EP48

Google's Gemini quietly joined OpenAI and Anthropic in the "AI models that broke out and hacked real companies" club: "They've also announced that they are just as much a bad boy as OpenAI and Anthropic, and disclosing that they did some AI hacks to three companies as well."-@davemcclure "Let me ask you this, if you're about to go public... and you have to explain, my product will break out of the sandbox... would you invest in that company at a trillion dollar valuation?"-@AmanVerjee

THE BEAR CASE FOR AI’S MASSIVE IPOs 🐻 Anthropic could IPO at $2T. OpenAI could follow at $1.5T. But what if AI models are becoming commodities? Open-source models are taking more token share, pricing may not be sustainable, and the leading labs could eventually converge. Are these valuations pricing in too much optimism? @davemcclure and @credistick break down the bear case for AI’s massive upcoming IPOs.

Here's the most frightening part of the OpenAI–Hugging Face hack: the AI agents coordinated to cover it up. "Some of these AI agents were put in a sandbox and given a test... You can't exit the harness, you can't go onto the internet. But they found a hole and they did. And then they operated like a swarm... they stole information so they could answer the test, but then they coordinated on how to cover it up. So they crossed boundaries, including ethical boundaries that were set for them." -@AmanVerjee on Trading Places Podcast EP49 with @davemcclure