More or Less · Topics · Episode #163
The Coming AI Correction: Too Much Infrastructure, Not Enough Use Cases -- and Pensioners Left Holding the Bag
Clip from the full recording, 54:15–60:20 · download mp4 · watch at this point on YouTube
Dave predicts a 2027-28 pullback once the hundreds of billions in AI infrastructure outrun real enterprise adoption, while Sam warns the eventual losers won't be tech insiders but pension funds -- setting up a populist backlash he half-jokingly says could make AOC president.
Dave’s core thesis is that human adoption of new technology is always much slower than the hype cycle assumes – true of electricity, cars, and the early internet (pets.com “did end up happening,” just twenty years later than promised). He argues AI is deep in that same hype phase: trillions of dollars of CapEx have gone into infrastructure while, from what he’s hearing directly from the front lines, even the best enterprise sales teams are struggling to get organizations to actually adopt the tools. His prediction: sometime around 2027-28, after the current wave of IPOs, the market corrects because the infrastructure buildout has badly outrun real use cases.
Sam agrees the mismatch is real but focuses on who absorbs the loss when the market corrects: not the tech insiders capturing IPO wealth, but pension funds – “the pensioners always get fucked” – the same dynamic he sees in Nvidia/BlackRock-style infrastructure deals. He distinguishes this AI wealth event from Google or Meta’s, where founders got rich but also built something with permanent financial value; here, he worries a lot of people are about to get extremely wealthy without having delivered any value in a durable sense, which he predicts creates a populist backlash beyond Silicon Valley once ordinary people connect AI’s uselessness to their own economic pain. He caps it with the running joke that this is “how AOC becomes president… and then we all get guillotined.”
Key points
- Dave: technology adoption is always slower than hype suggests, citing electricity, cars, and pets.com (correct thesis, ~20 years early on timing).
- Dave: trillions in AI CapEx have outrun real enterprise adoption; even top sales teams are struggling to get large organizations to actually use the tools.
- Dave predicts a market correction around 2027-28, after the current wave of AI IPOs, once the infrastructure/use-case mismatch becomes undeniable.
- Sam: draws a parallel to Nvidia/BlackRock-style deals where "pensioners always get fucked" holding the downside of speculative infrastructure bets.
- Sam distinguishes this cycle from Google/Meta's wealth creation: those founders built durable financial value, while today's AI wealth may be created without equivalent lasting value delivered.
- Sam predicts a populist backlash once the broader public connects sudden, undelivered AI wealth to their own economic squeeze.
- Running joke: Sam predicts this dynamic is "how AOC becomes president... and then we all get guillotined."
Where they landed
Quotes
“The human metabolism of technology is much slower than we intuitively think, especially when we're on the rise of one of these new technologies.”— Dave
“We've built out a dramatic amount of infrastructure... and the use cases and the things that are gonna metabolize this further into society have not appeared yet.”— Dave
“This is the whole -- the pensioners always get fucked, right? This is like the NVIDIA-BlackRock deals, where the bag holders are the pension funds.”— Sam
Suggested tweets
Dave Morin's prediction: a real AI market correction lands around 2027-28, once the infrastructure buildout finally outruns actual enterprise adoption.
Tweet this →“The pensioners always get fucked.” Sam Lessin on who actually eats the losses when the AI infrastructure bet doesn't pay off.
Tweet this →Sam Lessin, only half-joking: this is how AOC becomes president -- “and then we all get guillotined.”
Tweet this →All topics · Full episode #163 + transcript · Subscribe on YouTube · Spotify