Jay Cuthrell marks his birthday by examining AI infrastructure economics: Anthropic's rising valuation, the race for custom silicon and compute, and a looming memory shortage driven by AI demand. He covers hyperscaler capacity constraints, emerging token pricing models, security risks around advanced AI capabilities, and the workforce implications of rapid AI scaling.
Chapters
Welcome to What the Fudge podcast for May 31st, 2026, the I am one year older edition. This week I'm going to again try to keep it under 5 minutes. Let's see how we do. First up, it's the ever-growing AI infrastructure economic tug of war. Today, what we're seeing is that Anthropic is now officially surpassing OpenAI in valuation, meaning that it's in the hitting many, many billions of dollars following a bigger financing round than we've probably seen in some time. The competition is intense for compute power. That means that compute effectively is that primary bottleneck. We also see other players like Snowflake committing to other investments that similar size for AWS. Also, that means the AWS line of custom silicon. So if you're familiar with the Gravitons, that family, all things related to Trainium, Inferentia, etc. And of course, SpaceX still building data centers to handle all the other demands that they conveniently put into their press release in anticipation of their IPO. So if you think about this in terms of the hardware realities, we're seeing this concept of what I think the pundits are calling a RAMageddon. And so if you've ever seen Micron stock lately, you would understand where that might come into play. Again, it's AI-driven demand for memory, plus it affects everything else that's in that value chain for memory and any device that uses or consumes some form of, uh, the chips that form the basis of memory. This also occurs with storage. So if you ever remember a time when storage became expressly expensive, I've also talked about my cheapest chips theory and what happens when we come out of this and on the other side of it, what happens to the price of things like RAM, things like storage. But I digress. If you look at what's also going on around the hyperscalers, what they're building on in terms of capacity, there are some companies that are going to be facing somewhat of a sticker shock if they want to be part of this story. And I believe that this tokenomics story of rationing token usage will probably come up. We've already seen some internal references allegedly at AWS where they're taking down some of those burn as many tokens as you can as part of your how you're measured as an AWS employee. But if you really want to dig into it, there's also this concept of what was previously the Anthropic Mythos story of also, again, fine marketing if you can get ahead of it, meaning that there is bugs aplenty, there is security issues aplenty.. And so you've got everything from Glassworm botnets, CrowdStrike concepts, all the way through to what the complexity would be of some having access to something like a Mythos or what now Anthropic is calling a Mythos-like capability so that you can use that in your CI/CD pipeline for increasingly shift-left notions of DevSecOps, which I talked about before. The other element, of course, is that there was all the fallout from things related to Google I/O. I think really great coverage from MG Siegler as always around all of the kind of forward-looking as opposed to shipping-related things for Google I/O. So a promise, you know, I would gladly pay you Tuesday for a fully functional device story today. If you think of this as where else we might see, there's even concepts around where we have encyclicals coming from Pope Leo around AI ethics and that story, but Ultimately, this means that humans are contending with the possible ethical and future outcomes, whether they are dystopian or not, as it relates to all things AI, automating away things that might be considered part of the human experience. If you believe that you work to live or live to work, these are some of the philosophical things that are being done. If you look at what happens in terms of the CIO/CFO view of the world, there are definitely instances where granular governance around AI are getting more story coverage. As well as I, I would say from a cubicle-heavy or knowledge-work-heavy job role, there's probably going to be some impact on AI in those areas around what it is that you do. So imagine again your job description, if it has 20 bullet points associated with those last 4 or 5 bullet points, maybe something that becomes frankly done by machine learning or artificial intelligence. And so thinking about what those next 4 to 5 data points might be, it could be in fact that you're essentially managing a digital employee. So you're almost picking up managerial responsibilities, even if you're an individual contributor. So with that in mind, I think it's important to also understand that if you look at what's happening with the story around AI, as it relates to the ethics, as it relates to the economics, it's fair to say that this was a week of a lot of hand-wringing, a lot of ink being spilled. But I want to keep this under 5 minutes. So thank you again for listening to What the Fudge for May 31st, 2026. Take care out there.
