AI versus Marshy - The AI Bubble and Nvidia

I’m writing this article because I’m fascinated by the rapid growth of AI and its impact on the tech industry. As we approach the 3-year anniversary of ChatGPT, I’ve been thinking about the potential risks and consequences of this technology. In this article, I’ll be focusing on the AI bubble and its effects on companies like Nvidia. Let’s dive in and explore the situation. The second groundswell has (obviously) been generative AI. This technology also requires GPUs (with exceptions - check out Groq ) in order to work and currently demand largely outstrips supply. This makes (at the time of writing) Nvidia the most valuable company of all-time. big number goes brr Palantir is lesser known and more secretive. It sells fraud detection, pattern recognition, and works with governments, police, and security forces around the world. It’s been going since 2003 and has enjoyed its highest ever valuation in 2025. brr brr Palantir is also the name of one of the tools of evil in The Lord of the Rings . Not satire It’s founded by Peter Thiel. He’s kind of Silicon Valley royalty - famous for things like co-founding PayPal, destroying Gawker (also covered in Ryan Holiday’s Conspiracy ), co-writing a great read called Zero to One , and more recently - recreationally opining on the advent of The Antichrist . The reason he gets a mention here (other than that being an impressive collection of facts above) is that he has joined Michael Burry in offloading Nvidia stocks. see you later These moves in isolation still don’t mean much - the counterpoints to this are they are just two people who have regularly got things wrong, and the company is worth much more than any withdrawn stock. On the other hand, one of the people who predicted the 2008 crisis and literally one of the most powerful men in Silicon Valley are backing away… But I’m not an investor (my beloved Sobah went into liquidation!) ​ So let’s look at things through another lens, such as people’s sentiments . I’m seeing a lot of hate for AI out there. Vince Gilligan hates it. “ AI is the world’s most expensive and energy intensive plagiarism machine ”. The Australian government refuses to make adjustments to Copyright law for training. “And this is fundamental to their right as people who are generating works to ensure that they are fairly remunerated for that and that there are fair terms of use.” It’s become the topic of worn out comic artists - “ Every advert now ” Very appeasing My friend Davor calls it a growing dependency . And I love this take from a teacher named Jeffrey : Stop making so much sense! The people are getting frustrated. I think some of this backlash points to a desire for something more human and connected. But I also believe that - like a self-absorbed “friend” who behaves anything like one - people will simply stop engaging them and let it drift away. This will take longer than makes sense - but it will happen. But what about the gorilla in the room? OpenAI will go down in history for putting a version of AI on the map that fundamentally changed the direction of technology investment. But it will also go down. I loved this line “ OpenAI is Speedrunning the 2008 Financial Crisis ”. They lose $3 for every $1 they make. They’ve done $1.5T worth of deals without legal advice (paywalled). This snippet is an entertaining summary . I think this still is mid-Eff-word - sorry Ed It’s incredibly over-leveraged and will require people paying multiple orders of magnitude more for it than they currently do. Which is not what people want to do, and harder to justify when there are multiple other companies offering very similar products that are ready to step in ( open source models , Google , Groq ). Since starting this newsletter I’ve “got to know” the public face of Sam Altman. I saw him in Melbourne when he did his tour in June 2023. I remember being in the room and hearing the words “ harness the power of capitalism ” and wondering what he meant. Compare that with a couple of weeks ago… I don’t like this line of questioning ​ Sam doesn’t like the nature of Brad’s question. ​ I want to zero in on this bit: “We do plan for revenue to grow steeply. Revenue is growing steeply.” Finance isn’t my strong suit - but I have managed sales targets and want to know - how steeply is steeply? So I ran some envelope math with Perplexity : ​ I interpreted “our revenue is much higher than that” to be no larger than 5X To grow steeply - the company would need to exceed a triple-triple-double-double (a classic term in SaaS for hockeystick growth). Originally appeared in newsletter : AI versus Marshy 73 - OpenAI is this generation’s MySpace