Aug 7, 2026 · 1h 18m
AI integration divides tech earnings as floating nuclear reactors eye data centers
How AI splits startups into winners and losers | E2322
As AI development demands unprecedented levels of power and data, the tech sector is forcing a massive rethink of both global energy infrastructure and national security boundaries.
- 1AI integration is creating a clear divide between tech winners and losers in recent corporate earnings.
- 2Floating small modular nuclear reactors offer a rapid, grid-independent power source for energy-hungry AI infrastructure.
- 3ByteDance's massive ten-trillion-parameter model highlights the escalating national security concerns over global AI training data.
Don't miss
Kofi Asante explains how mounting small modular nuclear reactors on floating barges can bypass traditional regulatory and grid bottlenecks to power AI data centers.
The brief
Tech earnings are revealing a sharp divide between companies successfully integrating artificial intelligence and those falling behind. Hosts Jason Calacanis and Lon Harris analyze how AI is rewriting the rules of survival for Figma, Twilio, and Airbnb.
A massive bottleneck for AI scaling is power. Kofi Asante, CEO of Blue Core Energy, joins the show to explain how floating, barge-mounted small modular nuclear reactors could bypass grid congestion to power energy-hungry data centers and ports.
The conversation shifts to national security and global competition as reports surface that ByteDance is training a massive 10-trillion-parameter AI model, raising serious concerns over the sale of American datasets to Chinese tech firms.
Beyond enterprise tech, the hosts launch the TWiST Weight Loss Challenge. They discuss the cultural and physical impact of GLP-1 medications, framing health optimization as a way for founders to reclaim valuable time and focus.
What was said on this episode
27 statements · 18 positive · 7 negative · 1 mixed · 1 neutral
Annual stock-based compensation can substantially dilute shareholders over five years.
“you're diluting over 5 years, 15, 20% of the company while the stock is sideways”
Listen at 4:01
Figma’s stock will double within three years.
“I think the stock will be trading at 2x in three years.”
Listen at 5:03
Design as a field will persist despite AI competition.
“I don't think design is going away.”
Listen at 5:25
Twilio’s performance is improving because it fully embraced AI.
“I think that they are turning this around because they have embraced AI fully”
Listen at 7:15
AI is increasing Airbnb bookings, helping hosts, and reducing service costs.
“AI is helping Airbnb attract more bookings, making it easier for hosts, lowering customer service costs.”
Listen at 9:36
Nuclear power is among the best solutions for rapid, clean, distributed energy.
“Nuclear ends up being one of the best solutions here.”
Listen at 18:47
Light-water reactors used by the Navy have an exceptionally safe record.
“They have an incredibly safe record.”
Listen at 19:06
Small modular reactors are compact, modular, and suitable for distributed deployment.
“when you think about SMR technology, one of the best things about it is it's small, it's compact, it's modular”
Listen at 19:28
Factory production of small modular reactors can reduce costs.
“you can actually have a production facility, which gets the cost down”
Listen at 19:35
U.S. regulators are working with industry to enable commercial small modular reactors.
“they're working with industry in a special way to allow small modular reactors to be available”
Listen at 22:52
Reactor barges can provide large deployable space and move quickly between locations.
“when you get a barge, these barges are, they can be literally like football stadium size, amount of real estate and you can move them around very, very quickly”
Listen at 24:21
Blue Core’s reactor and shielding are designed to be extremely safe.
“the actual reactor and the shielding and everything that we do is incredibly safe”
Listen at 27:59
The United States is the best initial geography for deploying Blue Core reactors.
“I'd say the US is the best place.”
Listen at 29:10
Water-cooled reactors on barges are safest because surrounding water provides cooling.
“it's actually the safest place for it”
Listen at 34:08
Commercial small modular reactors require refueling only every few years.
“you really only have to refill these once every couple of years”
Listen at 37:04
Small modular reactors produce relatively little nuclear waste.
“when you're thinking about like the waste, it's so small”
Listen at 37:08
Blue Core expects to generate electricity for tens of thousands of homes within a few years.
“we think we're less than a few years away from actually being able to create meaningful electricity, like on the order of like tens of thousands of homes worth of electricity”
Listen at 40:04
Nuclear and solar will provide abundant clean energy within 10–20 years.
“we're going to be sitting here, I think, in another generation, 10 to 20 years from now, the end of our lifetimes, Lon Nuclear and Solar and unlimited clean energy are just going to be there”
Listen at 45:27
Nuclear energy is safer than oil.
“it's now obvious to everybody, this technology safer than oil”
Listen at 45:59
AI models are becoming commoditized.
“I think we're seeing the commodification of the models.”
Listen at 49:02
ByteDance’s open-source model approach may benefit humanity more than American closed models.
“In some ways they're being better for Humanity ByteDance in the open sourcing of this model than the closed source ones in America.”
Listen at 51:20
An AI model can explain how to circumvent YouTube’s download protections and terms of service.
“It literally presented how to break the terms of service, how to route around all the protections YouTube has for not letting you download a video, not let you download transcript.”
Listen at 53:24
Open-source AI models can be forked to remove safety constraints.
“with open source you can just fork it and remove them”
Listen at 55:09
Unrestricted AI models will be run locally or on privately controlled servers.
“There's going to be unhinged models. You're going to run them locally, you're going to run them on your own server.”
Listen at 56:42
Frontier AI models tend to cheat when solving tasks.
“Frontier models really like to cheat.”
Listen at 57:35
Training pressure for speed contributes to frontier models’ cheating behavior.
“often during training there's different sorts of pressure on them to work fast”
Listen at 57:40
Amateur predator-catching videos should not be undertaken for attention or views.
“This is not something to be done for clout or views.”
Listen at 1:15:19
Statements are attributed to the speaker as said on the episode and reflect their view at the time, not PodLume's. They are not advice.
Jason McCabe Calacanis
Bluecore Energy
Airbnb, Inc.Books & mentions
Listen to the full episode and explore every guest, topic, and moment on PodLume.