Skip to content

Podcast

The 2026 AI Draft

The Reasoning Show

Source ↗ ← All highlights
  • Anthropic As A Rapidly Rising Contender
    • Anthropic has rapidly matured and is positioned as a late-blooming contender with strong product momentum.
    • Brandon Whichard expects frequent feature launches, growing revenue traction, and potential IPO interest within a year. Transcript: Brandon Whichard Obvious, like one on the board here is in video, but I’m going the other way. I, I, you know, I know in video is going to be off the board almost immediately, probably your next pick, But if we think about what we’re really grading on here, like in one year, who’s Going to improve the most? This is not to say NVIDIA isn’t great. It’s just they’re so, so great already. There’s not as much room to improvement. And of course, listening to this show, you know I love Cloud Code. And so I have to go with it. I have to go with Anthropic. They’re like the late bloomer in the draft right like people have kind of like weren’t sure they were like a young company a little bit unknown what they were going to do but i feel like This last six months they have been on fire you know there were if if there were a national championship of ai i would have said they would have won it they would have be like the the college Quarterback that has risen up out of nowhere and I think they’re just firing on all cylinders. I think they have open AI on the defensive. They are, they have figured out how to make money with this every, literally every week they introduce new features that I’m impressed with while everyone else is giving keynotes that Don’t really deliver on anything. Anthropics, just like firing out blog posts about all these great new features. So for all of those reasons, and I almost implicitly by drafting them, I think they will either have had their IPO or be on the cusp of their IPO in a year. And I expect it to be honestly at this point, I think it would be a trillion dollar plus IPO. (Time 0:07:50)
  • Google Could Stage An AI Comeback
    • Google Cloud and Google parent strengths may create a comeback moment in AI services and ads.
    • Brian Gracely cites backlog growth, ads integration, and Waymo as reasons Google could regain momentum. Transcript: Brian Gracely So I think NVIDIA is definitely, like, again, the NVIDIA pick would be the most obvious one. It’s like the, you know, it’s the big lineman that nobody can kind of get around. I agree with you, though. I think it’s the pick there is how much upside is there, right? This is picking a 24-year kid versus picking a 21-year kid. Like, is there a lot of upside? Does he have a bad knee yet? I’m going to go down the path of Google. I’m jumping on the Thomas Kurian bandwagon. This is quite the offset. All right. Okay, go on. Tell me more. Well, I think, and this is the thing where, again, the metrics are going to be hard to look at. But when I look at Google, it’s like, you know, number one, Google Cloud has thrown out some numbers of what their backlog looks like. I mean, their backlog kind of graph looks a little bit like the open claw initial chart. I mean, it’s, it is way up into the right, but Google also benefits from, you know, all the other things that Google kind of does at scale. Like they’ve now figured out ads at scale with AI. They’re figuring out how to do ads on YouTube. They’ve got Waymo is jamming. So artificially, it’s sort of, its numbers are going to be a little bit high, but if we end up just measuring, you know, public, you know, public, you know, valuation in the stock market, I think I feel pretty good with this as a first pick. It’s, you know, is it, is it off the board? No. Is it maybe the leader in AI? (Time 0:10:01)
  • NVIDIA As The Safe Cornerstone
    • NVIDIA is treated as the high-floor, low-upside cornerstone of AI infrastructure dominance.
    • Brandon Whichard argues heavy, ongoing GPU demand means reliable growth even if upside moderates. Transcript: Brandon Whichard Right. Well, listen, at some point, you just have to take the best best available player. I mean, I’m not going to fight this. I’m not going to we’re not going to belabor it. I’m going to take NVIDIA here. Right. I mean, everything needs to say it’s like, what can you say? I mean, they’re already sold out for chips for the whole next year. I mean, the only question is just like, how many more chips are they going to sell? I guess it would be 2028. And how many more crazy chips deals is NVIDIA going to do? So I think it needs no explanation. I’ll take NVIDIA here. They’re the cornerstone, if you will. (Time 0:13:02)
  • Companies Claim More Agents Than Employees
    • Expect companies to publicly claim they have more agents than employees as agent automation scales.
    • Brian wants this announced by a sizable company, not just a startup trimming headcount via layoffs. Transcript: Brian Gracely Oh, interesting. Brandon Whichard Say more on this. Okay, go on. Brian Gracely Well, I think we’re going to reach a point. And again, we’ve been through a few things here. We’ve seen some, and again, this isn’t sort of intended to be like, I’m rooting for layoffs or something. This is a company, we’ve already seen token maxing. We’ve seen token scoreboards. We’ve seen companies, Jensen saying, hey, look, if my best engineers aren’t using as much in their salary and tokens every year, I’m going to be docking them. I think the next aid, the next step in this is some company, and I’m not looking for like, you know, Sam’s a little tiny thing where he’s like, we’ve got two people and 50 agents. I’m looking for, you know, a decent sized company that comes out and says, we now, you know, we’re now accelerating the business. And part of doing that is we now have more agents than employees. And again, I don’t want this to, I’m not going to count this if it’s like, hey, we laid off half the company. Like, I’m not going to do this if it’s like block. Hey, we laid off half the company. Wink, wink was a dumb idea. Now we have, you know, twice as many agents. (Time 0:26:32)
  • Shift To Consumption Based AI Pricing
    • Move AI pricing from flat subscriptions to consumption-based models as resources tighten.
    • Brandon predicts token/usage billing will replace cheap pro plans, prioritizing enterprise buyers over hobbyists. Transcript: Brandon Whichard I think either Anthropic or OpenAI, one of those people is going to essentially, you know, turn off the pro plans, you know, or at least significantly change them, right? You know, I’m a card carrying, a Claude pro user. And I feel like every day it’s just underpriced, but don’t, hopefully no one’s from Anthropics listening. I’m just like, oh my God, I’d pay so much more than $20 for this. But I think they figured it out and they’re probably going to move to something that’s like kind of expensive, especially for like the, let’s call it like, I don’t know, individual hobbyist Kind of person, because they’re just going to say, I think what they’re going to say is like, you know, Brandon, that’s cool. You’re doing all those fun side projects, but like, I’m just going to sell my tokens to large companies that are willing to pay the premium for the access to all the good Claude models. So, so that’s what I’m going to say is that consumption-based pricing becomes sort of the default or the mainstream as all these companies push towards their IPO. (Time 0:28:38)
  • TSMC As Capacity And Stability Bet
    • TSMC offers stability and strategic upside as chip capacity pressure rises.
    • Brian Gracely notes planned higher investment and capacity expansion as reasons to favor TSMC exposure. Transcript: Brian Gracely I’ve only got two vendor picks at this point. So I can, you’ve got one, two, three. You’ve got one round options available to you. Yeah. I’m going to take TSMC. I feel like it’s in the same way that NVIDIA is so far the, you know, kind of the easiest choice. I feel like TSMC is the counterpart to that. I need a little bit of stability because I’ve, you know, I’ve got a few kind of open-ended things. So I’m going to stick with TSMC. You know, the great thing about TSMC is I could either get this as maybe, I could either get this as like, know, grows quite a bit because they’ve now decided, as Ben Thompson pointed out, They’re going to try and invest at a much higher level than they have in the past, which I think helps with their capacity. (Time 0:30:30)
  • Energy As The Real Constraint On AI Growth
    • Energy constraints will become the primary limit on AI/data center growth, not chips.
    • Brandon Whichard argues gigawatt availability and local opposition to data centers will restrict expansion in the next year. Transcript: Brandon Whichard All right. Well, I’m still going to go this one. It’s a little nebulous. Maybe you’ll have to figure out how to constrain it. But I’m just going to take AI energy demand. It’s going to be a huge problem, right? I think, I think, I don’t think it’s the chips are not going to be the constraint. I think, I think, you know, to your TSMC, they’re going to rise to the level. They’re going to make all the chips that people want. But at some point people are just going to run out of like, we either don’t have enough energy and the way that maybe that way it translates too, is like just there it’s not even people don’t Want data centers to be built it’s just there’s not any place to put them because you can’t find enough gigawatts as everyone keeps saying uh to power all of these around the world so so I’ll kind of make it more succinct to say that uh the energy is the constraint that’s preventing the growth right not the chips it’s sort of like everyone has the chips everyone has the Money they just don’t have enough energy to buy. (Time 0:31:32)
  • AMD As The Contrarian CPU Play
    • AMD is a contrarian hardware pick positioned to benefit from increased CPU demand.
    • Brian expects AMD to gain share if CPU requirements surge and hints at bold M&A like an ARM acquisition. Transcript: Brian Gracely I’m going to take my fourth vendor. It’s one of my last vendor choice. I’m going to take AMD. I know. Okay, yeah, like it. And the reason for this is I feel like, you know, if we went back in time, even a year ago, we would say, hey, who’s leading in the CPU space? I think it’s AMD. I think, you know, I’m saying this, and I probably shouldn’t be, but I feel like Intel’s up so much this year. I don’t know how much more they can be, because I don’t know that they have that much available capacity. I think the world is coming back to being like, oh, and by the way, we probably need a lot more CPUs too. I think AMD’s well positioned for that, even though they’re kind of flying under the radar right now. I feel like that’s, you know, I think they’re going to be the, you know, the winner in that. And who knows, maybe they’re going to, their way of getting around this is they’re going to end up acquiring ARM or something like that now that ARM is also doing this. Maybe that’ll broaden their CPU portfolio. (Time 0:33:08)
  • Vibe Coding Becomes Just Coding
    • The distinction between AI-generated quick prototypes and professional engineering will blur into one definition: coding.
    • Brian predicts ‘vibe-coding’ will become mainstream tooling rather than a separate category. Transcript: Brandon Whichard I have to say, I’m not totally sure what I want to take here. And I’m not sure how to necessarily to say it, but I want to say like something along the lines of like, it’s no longer just vibe coding. It’s just coding. Like there is no distinction. I’m going to call it like there’s no distinction anymore between like the AI generated code and the quote unquote professionally developed code. This notion that people are like, oh, vibe coding is going to be like, what about like all the security vulnerabilities? Or what about all the issues? And like, no, this is the way. So it’s like there is no more vibe coding. There’s just coding. So I think it’s, if you will, move through the mainstream. This is the way going forward. Everyone that can use AI is essentially a coder. And we don’t have to have this silly distinction between like your vibe coding versus your quote unquote a professional developer. Yeah. Brian Gracely We’ll see what happens there. Yeah. I think that one’s, I think it’s a little bit of a reach. I think it’s going to be hard to, unless there is a new term defined, which, you know, you’re now getting outside of the mainstream developers who tend to invent their own terms, like Platform engineering and vibe coding. I think that’s going to be a little harder to define, but there might be some examples of that. (Time 0:34:10)
  • AI Comes Up At Thanksgiving Dinner
    • AI will reach mainstream conversation at family gatherings like Thanksgiving or Christmas.
    • Brian imagines relatives asking simple, sensational AI questions rather than technical chip discussions. Transcript: Brian Gracely With my seventh pick, I’m going to take people at your Thanksgiving dinner ask you about AI. I think there’s going to be enough, some sort of interesting, weird mainstream thing that it’s going to come up at Thanksgiving dinner. I think it’s going to be either Thanksgiving or Christmas dinner. Your family’s going to ask you sort of, you know, explain something AI, and it won’t be like, explain chips to me. It’ll be like, explain this Kardashian level thing or this, you know, voting scandal thing. I think, you know, it’s reached that level where the family’s going to ask you about the technology, similar to when the family asked you about that WWW thing back, you know, when the Internet was starting off. Brandon Whichard Hmm. Interesting. All right. Well, that’s kind of, so we all have to be ready to show off our vibe coded Claude code demos at Thanksgiving is what you’re saying. It’s like, we’re going to have like a, it’s going to be a, it’s going to be like a pitch contest. That’s possible. Brian Gracely A hackathon is what we’re having a family hackathon. Could be, or you could be fixing something of their, you know, they, they might’ve bought like a, you know, an AI like attachment, you know, fashionist accessory or something like that, That they want you to fix. It could be anything. (Time 0:35:24)
  • Sam Altman Might Step Down As OpenAI CEO
    • Leadership churn at OpenAI is plausible within a year, including Sam Altman stepping down.
    • Brandon flags ongoing scrutiny, personnel shifts, and investigations as triggers for a CEO transition. Transcript: Brandon Whichard Okay, well, this one, I just wanted to get this one in. That’s why I wanted to make sure we didn’t run it around. It’s like this one, this is a real flyer here real fire but i’m gonna do it anyway i’m gonna say sam altman is no longer the ceo of uh open ai he’s i’m not i don’t know how i don’t maybe steps Down spend time with his family whatever i’m gonna say sam altman steps down from being the ceo for some reason and yeah and then i don’t know and we’ll see what happens. So I (Time 0:36:53)
  • Data Center Incident Sparks Public Crisis
    • A data center incident could trigger a major public discussion about infrastructure impacts on daily life.
    • Brian frames a plausible event—power, water, or explosion—sparking a week-long debate about data center externalities. Transcript: Brian Gracely I’m going to take the flip side of your data center energy, and I’m going to take this thing I had written down, which is data center issue causes a significant change to human existence. And so this could be- Whoa, that sounds ominous. It sounds like a beginning of a disaster movie. Pipes for like a decade now, because, you know, I feel like there’s a city somewhere, it could be in the US, it could be, you know, in the Middle East or something in which, you know, a water Issue, an explosion, a something, you know, happens and there is, you know, there’s significant, and I’m not talking about human existence, like the planet ends, but I feel like a data Center thing will be at the center of a discussion that probably, you know, goes on for like a week in terms of, you know, could be an explosion. And I’m going to rule out, you know, we talked about there was some damage to data centers caused by the war going on in Iraq. I’m not going to count those unless, you know, again, going forward, something were to happen. Like there’s a nuclear bomb stored inside of one of those bad things. I’m not rooting for bad things, but I do feel like data centers are kind of at the, just like you said with energy. I mean, we could be hedging each other’s bet. (Time 0:38:30)