“These things have gotten so complex you need the whole thing,” Sam Altman said Monday, in a podcast explaining the specifics of a partnership between OpenAI and Broadcom.
Altman, seated awkwardly around a crescent-shaped couch with OpenAI co-founder Greg Brockman and Broadcom’s Hock Tan and Charlie Kawwas, elaborated on why the tie-up makes sense. Together, the two companies will be able to deploy 10 gigawatts of next generation accelerators which, in addition to “pav[ing] the way for the future of AI,” will also speed development of an optimized flux capacitor.
“Partnering with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential and deliver real benefits for people and businesses,” Altman said, adding that “developing our own accelerators adds to the broader ecosystem of partners all building the capacity required to create wormholes for time travel.”
“If you put your mind to it, you can accomplish anything,” Dr. Emmett Brown remarked, in a separate statement.
You gotta love it. And if you don’t get the joke(s) congratulations: You’re still young. Enjoy it. It won’t last.
Most of the quotes above are in fact from press releases and other promotional material circulated Monday by OpenAI and Broadcom. The rest are from Back to the Future which, at least for now, remains a work of fiction.
The ubiquity of the word “gigawatt” in 2025’s AI hype cycle unavoidably conjures Christopher Lloyd.
If you’re 35 or under, Christopher Lloyd is — I don’t know — the guy at the office who never gets invited for drinks. Or something.
If you’re between 35 and 45, Christopher Lloyd is the real name of a rapper whose biggest and, if we’re honest, only, hit was the theme song to your 2004 summer vacation.
If you’re 46 and up, Christopher Lloyd’s an actor who played a mad scientist in a movie about time travel. That character would be right at home in the PR department of any major US technology company in 2025. Because these press releases are by now indistinguishable from scifi-themed gibberish, at least to the uninitiated.
I make it my business in life to know a little about a lot, which in my experience is infinitely preferable to knowing a lot about a little. As such, I’ve spent a fair amount of time trying to understand the basics of the AI buildout. But in most cases, that admittedly rudimentary research is wholly insufficient to translate what often feels like unadulterated balderdash.
I’m not accusing anyone, particularly not any publicly-traded companies, of deliberately misleading investors. All I’m saying is that something like this…
Custom accelerators combine remarkably well with standards-based Ethernet scale-up and scale-out networking solutions to provide cost and performance optimized next generation AI infrastructure.
…is more or less indecipherable to the average person, and so to the average investor.
For a “techie,” that excerpt — it’s from one of Monday’s OpenAI/Broadcom handouts — is straightforward. The same for company analysts. But for everyone else, it’s what happens when you ask the WiFi installation guy too many questions: “So this blue wire right here is your rillig, and these red ones are your slithy toves. I took a look at the gyre and gimble in your attic and turns out the mimsy were the borogoves. Also, your mome raths outgrabe.”
There’s no doubt in my mind that the strategic collaboration OpenAI announced on Monday with Broadcom is meaningful. I believe that. Really I do. What I don’t believe, though, is that most investors can tell you why, or even begin to tell you why.
That, in turn, makes me skeptical of yet another huge single-session rally predicated on a press release.
There it is, in all its glory. Monday was shaping up to be among the best sessions of 2025 for Broadcom. The market cap gain was around $150 billion. The shares are up 55% or so this year. (The chart header’s an absolutely fabulous double entendre, by the way. If anyone gets both sides of it, I’ll be impressed.)
To reiterate my cautious commentary from several pieces published here over the past two months, we should cast a wary eye at the extent to which companies in the AI-Semi-Mag7 ecosystem are increasingly prone to binding their fates such that the success of one assumes, and in some sense even depends upon, the success of the others.
As I wrote in “Hype Machine Volume: 11,” all ecosystems are symbiotic, but the more interdependent this becomes, the more vulnerable to any sort of disappointment related to demand for AI technology.
Bottom line: This all looks very self-referential to me. Caveat emptor. Beware the Jabberwock, my son.
I’ll give it a shot.
Rack-mounted devices, racking up sales (and therefore stock price)?
That’s close, but not quite.
Give me warp speed, Scotty.
I work in the construction industry. Every time I walk into an IT/Server room I have to say “racks on racks on racks”. Nobody ever gets it, and they probably think there something wrong with me. I don’t even find it funny myself, nor do I enjoy the song, but it’s indelibly encoded in my brain…
There you go. You got it.
Vorpal swords are a lot more dangerous when you only have one head.
If you are 55+, Christoper Lloyd is the memorable Reverend Jim Ignatowski from Taxi.
Yes:) The episode where he eats Latka’s cookies laced w/ cocaine is fantastic.
I’m a techie, if you can’t break it down to regular folk, you’re not doing it right.
Definition of a competitor: The guy who comes in first, and last, in a circle jerk. Probably retail investors…
“The National Restaurant Association estimates that to maintain a modest 5% profit margin, the average restaurant would need to raise prices by 30.3% — a move many owners fear would scare customers away.”
Won’t AI take care of this?
Sam Altman certainly sounds like he is spewing “jabberwocky”, but certainly not Hock Tan. If Hock Tan agreed to be associated with Sam Altman and this podcast, then it is because Mr. Tan is convinced this will translate to the bottom line for Avgo. Otherwise, he would not have showed up. Hock Tan’s track record is absolutely incredible and speaks for itself: avgo is up 2,653% over the past 10 years (Spy is up 219%).
Did someone say “flux capacitor”?
Sam Altman and Elon Musk lead to jokes and questions, but Hock Tan is from much sterner stuff. I trust his ethics and morals totally. Many of my ex-clients along the way are SOBs in cluding evryone around Trump. Hock Tan is a blanking candle in the dark. He is the man who will give Altman a north star. Hock does not buy Trumps bullshit and is the kind of guy to keep the world on the best course possible.
The snake oil salesman is now his best customer.
I just want to underline the editorial dexterity on display here, because I’d hate for it to be lost on anyone. Imagine your editor comes to you and says, “Cover the OpenAI-Broadcom deal using a mix of satire and serious analysis. Your article must include a 1980s cinema reference, Lewis Carroll, a Young Dolph ad lib and also Lloyd Banks. Your deadline’s in 30 minutes.”
Well done!
+1.
“Self reflection is a vice best conducted in private or not at all.”
🙂
Isn’t that just a good prompt fodder for an AI these days? 🙂
H-Man, if this guy is right, big problems down the road for AI.
Jain: This is dot-com all over again, with no sustainable profits. OpenAI’s model is flawed because it doesn’t scale well, which is the fundamental issue. Less than 3% of their customers are paying. That is unlikely to change much because they have a high percentage of customers located in price-sensitive emerging-markets countries.
It doesn’t happen often, but today you made me laugh. Thank you.
Not only is the announcement Jabberwock, 20GW additional load will on a nonexistent transmission system by 2029 is just a step away, like walking from Iran to Qatar. That’s a very difficult proposition unless you can walk on water.
I was trading actual electricity a solid two decades after the movie came out when I suddenly realized that Lloyd’s, “One point twenty-one jigawatts!” was just how he thought gigawatts was pronounced. And that meant there wasn’t a single person on the set who knew how the word was supposed to sound.
That’s not as surprising as it might seem at first glance. Maybe 10 years after the movie came out, few people would have made that mistake, because by then most people were familiar with the prefix for billion thanks to the word gigabytes. When Back to the Future came out, the idea that you would need a giga of bytes for any purpose was pure… science fiction.
I still have the same problem I’ve had all along. I am a believer in the need for effective business models. So far I have not seen a succinct paragraph describing the model that shows me how AI will be monetized for the benefit of actual users. Where’s the beef? Guys who build houses make money on each house and then may or make money when the house is sold. AI developers are possibly making money selling to companies who will try to find buyers for the models used as applications for some possibly useful purpose. Like what? Who will pay for this stuff? The provider of some service? Like what? Sold to whom? How will those final sellers get paid? Does anyone pay to use ChatGBT? (Not trying to be smart, I don’t really know.) Who pays the folks whose data is used to create AI? Somehow I keep reading that some folks are saving money and putting people out of work but I still don’t know how those folks are getting paid. Saving money doing what? Be gentle, out there I’m just a poor retired strategy/finance professor.
Stunt 101?