What The Fudge for June 28, 2026: Cost Crunch, IPO Chill, and Agentic Future

And welcome to What the Fudge podcast for June 28th, 2026, the I Didn't Travel Again, Really? That's True edition. And I'm going to try to again try to fit into 5 minutes. I think I was able to get that just under the wire last week. So let's get started. First up is all things related to where AI gets powered, gets energized on top of hardware and how that hardware and the infrastructure associated with it is always present in the news cycle. So we've got everything from IBM talking about major breakthroughs getting to a 0.7 nanometer chip fab process that's doing 3D stacking, uh, in this case nanoscale stacking for the transistors. The next piece of that would be— is beyond miniaturization— would be overall data center efficiency. So we'll see some of these stories play out as that becomes commercialized. But for now, it's one of those kinds of popular science announcements. It's, it's absolutely scintillating, but you have to see product shipping. Meanwhile, back at the farm, OpenAI and Broadcom have gotten their first custom inference chip. Be, be very clear on this— inference, not training, inference chip. It's named Jalapeño, which is amusing. But the goal, of course, with all custom silicon is to slash the costs associated with and reduce the dependence on some of the industry incumbents, which you would say are, for example, NVIDIA. And so in this case, it's OpenAI working with Broadcom, as Broadcom does allegedly with other hyperscale providers, to help them build their own custom chips that are not related to or part of NVIDIA. And so now you've got this kind of discourse around where we begin to say, is there— should there be data centers in space. And there's even some contrarian research now being published where, hey, engineering is hard to put things into space. And they go into everything from radiation hardening to the concept of what does it mean to truly cool in space when if you're in the sun, you're baking, if you're not in the sun, you're freezing. Really interesting stuff. But I think what we're trying to kind of guard against here is the unbridled enthusiasm of suddenly overnight there's just going to be AI data centers in space. It will happen. The question is, is when? And in that race, is it more important than getting to moon— to the moon, getting to Mars, setting up or establishing manufacturing in space? Interesting things to think or, you know, ponder. But those are the big stories. Uh, people are starting to try to deflate, uh, some of the rhetoric around that. Uh, next up, if we just look at the regulatory IPO markets, it looks like the— I think OpenAI is going to delay. So everything I'm reading says OpenAI is going to delay, and that means that they won't do an IPO. What does that mean? And so everyone's trying to figure out what that apparently will mean for everyone else, as opposed to it being OpenAI, again based on the news, saying they're going to delay their IPO. The other part of this is if you do have a story around where training and inference costs are being challenged. There's that training part in the creation of, or the use of, so-called distillation techniques. And so there's another one of these distillation accusations that's entered the market. This time I believe it's Anthropic making the distillation claim against Alibaba. So we'll see how that plays out as well. We also now start to see the, uh, Mythos and Similaro It's too dangerous, you can't hold on to it, so I'm going to stagger the release. And so the same thing now is happening for a GPT-5.6, which of course is an OpenAI story, so they probably want some of that same sweet, sweet marketing hype associated with it. And then just kind of rounding this out, if you weren't already pricing out things like new hardware, you're noticing that the cost of that hardware is going up. And I talked last week about, yes, there will be impacts to iPhone pricing, but now it's also going to be things like Macs, iPads, and that, that thinking around if you bought now, is it the best price it's ever going to be, or do I wait? There's a long timeline before our supply chains for this type of component and subcomponent, even some of the base materials, before any of this comes back into some semblance of what we were experiencing, say, several years ago, you know, in a post-pandemic world or in a pre-pandemic world where you kind of had a steady state and gradual reduction of costs. Right now, AI is just making that big, huge sucking sound, just like I mentioned last week. And so with that, I'm going to tighten it up and say that we are just now under the 5-minute mark. So keep an eye on things like AI coworkers, agents. We're seeing some interesting challenges around how companies are adapting and adopting that into their organizations. But thank you again for listening to What the Fudge podcast for June 28th, 2026. Take care.