
Aug 29, 2026 · 2h 27m
Listen from 1:33:59
Listen at 1:33:59
AI’s boom collides with surveillance, jobs and human intimacy
NVIDIA's $96.2B Quarter, China's 200,000 Fake Accounts, & OpenAI's New Chip | EP #284
The episode connects AI’s extraordinary commercial growth to harder questions about geopolitical power, privacy, work, health and human relationships.
- 1NVIDIA’s record quarter fuels debate over durable AI demand versus a circular financing bubble.
- 2US–China competition is expanding from model development into influence operations, local chips, surveillance and AI companions.
- 3Emerging technologies promise gains in cancer treatment, tooth regrowth, energy and lunar infrastructure while reshaping social life.
Don't miss
The episode links a newly approved pancreatic-cancer treatment and tooth-regrowth research to a broader argument that biotechnology may be entering an abundance phase.
The brief
Peter Diamandis, Salim Ismail, Dave Blundin and Alexander Wissner-Gross examine NVIDIA’s $96.2 billion quarter, asking whether AI demand is durable growth or a circular financing bubble.
The panel frames US–China AI competition as a contest over models, video generation, chips and influence operations, while China restricts AI companions over emotional dependence.
Flock Safety’s license-plate network becomes a test case for the surveillance bargain: police gain investigative power, but ordinary movement becomes increasingly legible.
The discussion rejects an inevitable AI job apocalypse, then turns to local processing, energy abundance, cloud seeding and the infrastructure needed to scale intelligence.
The episode’s most striking turn is biological: a newly approved pancreatic-cancer treatment and tooth-regrowth research sit alongside lunar compute, robotaxis and intimate robots.
What was said on this episode
66 statements · 46 positive · 16 negative · 1 mixed · 3 neutral
Nvidia will remain sold out for the foreseeable future.
“they're going to be sold out for as far as the eye can see, no matter what.”
Listen at 6:40
Nvidia’s revenues carry 80–85% gross margins.
“These are 80 to 85% gross margin revenues.”
Listen at 6:54
Nvidia-financed customer demand would make its growth look bubbly.
“If it turns out that a disproportionate amount of this demand is directly or indirectly Backstopped or financed or credited by Nvidia through one or more intermediaries, then I think that starts to look a little bit too bubbly”
Listen at 9:54
Any Nvidia-related downturn will be limited to a mini AI winter.
“my forecast is at most a mini winter, as it were”
Listen at 10:46
Any Nvidia demand correction would produce at most a mini winter.
“my forecast is at most a mini winter”
Listen at 10:46
Nvidia is expanding from chips toward an intelligence operating system.
“Nvidia is trying to move out of just being a chip company to try and become like literally an operating system for the whole of intelligence”
Listen at 12:33
“I think there is a path to that, but I think it's a highly nonlinear multidimensional path.”
Listen at 15:29
Tesla Optimus could become a multitrillion-dollar annual business.
“I think Optimus alone, not financial advice, could be a multitrillion dollar per year business”
Listen at 15:48
AI training requires large numbers of coherent GPUs in one location.
“for training time you need all the GPUs to be coherent in one location”
Listen at 18:39
Several AI-discovered drugs will generate more revenue than the legacy economy.
“If it comes up with five or 10 more drug discoveries, which it inevitably will, very soon, the amount of revenue flowing back into the AI economy from just that will dwarf the legacy world.”
Listen at 21:07
American AI labs prioritize revenue more than Chinese AI labs.
“American AI labs are revenue maxing and Chinese AI labs are not.”
Listen at 26:53
Omnimodal models will ultimately be the strongest model architecture.
“I think in the end omnimodal models are the best.”
Listen at 28:33
Omnimodal models will improve within six to twelve months.
“in the medium term, which is like six to 12 months”
Listen at 28:46
Chinese AI models face enterprise adoption barriers because of trust concerns.
“the Chinese, they don't have the ability to sell Kimmy or Quinn into the enterprise use case because nobody knows if they can trust it”
Listen at 29:24
Chinese AI labs can globally monetize video models.
“the Chinese are brilliant by going after video because they can just launch it on the world”
Listen at 29:59
AI that models and acts in the physical world will prevail.
“the intelligence that can model and then manipulate the world and act inside the world is going to win.”
Listen at 32:27
China may have an advantage in embodied AI.
“where China's heading may have a slightly better advantage”
Listen at 32:35
AI agents will become increasingly intelligent and personalized.
“these AI agents are so good. I mean, they're going to get smarter and smarter and smarter and user specific.”
Listen at 34:44
AI companions may benefit loneliness, disability, elderly care, therapy, and social skills.
“There are some really solid benefits around loneliness, disability, elderly companions, therapeutics and coaching social skills”
Listen at 39:21
The CCP will reverse course and embrace AI companions.
“I think the Chinese Communist Party is actually going to do an about face sometime in the next year or two or three and decide that they actually love AI companions.”
Listen at 39:49
The CCP may eventually use controlled AI companions for youth influence.
“I think the party decides actually having everyone, or maybe not everyone, but a sizable fraction of the youth of China deeply engaged with Party. Deeply controlled AI companions is actually, from the Party's perspective, net helpful, not net harmful.”
Listen at 40:21
AI romantic companionship will remain legal in some US jurisdictions.
“my bet is that at least in some places in the US AI companionship, including romantic companionship, including robotic romantic companionship, will be legal.”
Listen at 41:49
Earth can support tens of billions of people.
“We have a carrying capacity in the tens of billions.”
Listen at 43:04
Declining human population will not be a major existential risk.
“I don't think it's going to be a major X risk at all.”
Listen at 43:33
Flock Safety represents municipal AI’s current leading form.
“I think this, honestly, I think this is the face of AI at the municipal level at this point.”
Listen at 46:41
AI has reduced the marginal cost of mass surveillance to nearly zero.
“the cost of following everybody has gone to zero”
Listen at 49:38
AI has reduced the marginal cost of mass surveillance to nearly zero.
“the cost of following everybody has gone to zero.”
Listen at 49:38
AI-enabled mass surveillance is unlikely to end well.
“I don't think it ends well.”
Listen at 50:55
Benevolent dictatorships typically cease being benevolent over time.
“the problem with benevolent dictatorships is that they typically don't stay benevolent very long.”
Listen at 55:10
Salim rejects Goldman Sachs’s professional-services risk assessment.
“Deeply disagree.”
Listen at 1:10:29
31% of surveyed companies expect AI to increase staffing and wages.
“31% expect increasing staffing and increasing wages.”
Listen at 1:12:14
AI will move accounting workers from rote entry to higher-level judgment.
“The human being is above the loop and is categorizing the transactions, looking for problems, categorizing vendors, how do you improve income?”
Listen at 1:15:43
AI-era economies will reward human agency.
“high agency and agency in general is One of the few human traits that is actively rewarded in an era of superintelligence.”
Listen at 1:18:58
OpenAI’s Jalapeno inference chip uses dramatically less power than Nvidia’s GB300.
“Jalapeno runs on 7 watts against the GB300. That runs at 1400 watts”
Listen at 1:27:27
AI inference workloads are definitively shifting away from Nvidia hardware.
“inference is moving off of Nvidia for sure”
Listen at 1:28:03
AI inference workloads are shifting away from Nvidia hardware.
“inference is moving off of Nvidia for sure.”
Listen at 1:28:03
AI training will remain on Nvidia and stay capacity-constrained.
“training is staying on Nvidia and training is infinitely sold out”
Listen at 1:28:13
OpenAI may become a hyperscaler using its own chips.
“OpenAI might actually do this to themselves even though they are competitive with anthropic, they're one of the two top tier frontier firms left. Whether they might actually choose to become a hyperscaler themselves with their own chips in house.”
Listen at 1:29:49
OpenAI may launch a cloud hosted on its own chips.
“OpenAI offers its own cloud. We'll call it OpenAI Compute. It'll be hosted on OpenAI chips”
Listen at 1:30:36
CUDA is no longer a moat for inference workloads.
“CUDA as the inference time mode is already dead.”
Listen at 1:33:07
CUDA’s training advantage probably has a limited lifespan.
“CUDA as the training time mode is probably limited lifespan”
Listen at 1:33:10
Nvidia’s interconnect technology is becoming its new competitive moat.
“the interconnect is the new mode.”
Listen at 1:33:15
Apple’s 512GB unified-memory Macs enable substantial local AI capability.
“The 512 gig Unified Memory puts a massive amount of intelligence under your desk.”
Listen at 1:33:59
Local Apple AI may especially benefit law firms and HIPAA-constrained healthcare companies.
“the big winners here are law firms, healthcare companies that have HIPAA issues and can't put their stuff on the cloud.”
Listen at 1:34:21
Apple currently lacks an adequate AI strategy.
“they have totally missed the boat and it's just embarrassing that a company with that much cash Flow has no AI strategy.”
Listen at 1:35:20
Modernized uranium enrichment could create an energy boom.
“I think it'll create an energy boom.”
Listen at 1:45:30
Solar growth is systematically underestimated by focusing on current capacity instead of deployment growth.
“The problem is solar keeps getting underestimated because people look at installed capacity today rather than the vertical slope of the deployment curve.”
Listen at 1:47:46
Solar growth is underestimated when judged by current installed capacity.
“solar keeps getting underestimated because people look at installed capacity today rather than the vertical slope of the deployment curve.”
Listen at 1:47:47
Drone-based cloud seeding could eventually enable total weather control.
“we could have total weather control.”
Listen at 1:51:56
Weather modification creates cross-border conflicts because weather ignores political boundaries.
“Weather. The challenge is weather typically doesn't respect weather boundaries.”
Listen at 1:53:02
Daraxonrasib nearly doubled median survival in previously treated metastatic cancer.
“Dirax and Rasib nearly doubled median survival from 6.7 to 13.2 months.”
Listen at 1:57:21
Daraxonrasib increased tumor response rates from 11% to 32%.
“It also tripled tumor response rates from 11% to 32%.”
Listen at 1:57:27
AI-driven biology will produce many new cancer targets and treatment approaches.
“I would expect lots of new targets, lots of new approaches to emerge from both the AlphaFold3 style of protein folding problem and structural biology being solved, as well as virtual cell models”
Listen at 1:58:50
Biomaterial advances developed for enamel may also support bone regrowth.
“similar biomaterial discoveries and inventions will also be helpful for bone regrowth”
Listen at 2:03:35
AI-enabled regenerative medicine may eventually regrow teeth, limbs, and ears.
“I can grow a tooth, I can grow an arm, I can grow an ear, I can grow anything.”
Listen at 2:04:33
Consulting firms’ technology forecasts should generally be multiplied ten- to twentyfold.
“take any of their estimates and multiply it by 10 or 20 times.”
Listen at 2:08:47
A lunar semiconductor fabrication facility is the Moon’s most valuable application.
“I do think the pedofab is the killer app for the moon.”
Listen at 2:09:52
Lunar economic development depends on mining, chip fabrication, and data centers.
“the critical path is mining chips fabbing data centers on the moon”
Listen at 2:10:32
One million Tesla robotaxis could generate about $100 billion in rideshare revenue.
“that's probably about $100 billion of revenue in rideshare revenue for Tesla.”
Listen at 2:13:20
Tesla’s latest Full Self-Driving Model Y substantially improves users’ driving experience.
“all my friends that are driving the new full self driving Model Y's are like, oh my God, this is life changing.”
Listen at 2:14:33
Entrepreneurs should deploy fleets of Cybercabs to generate rideshare revenue.
“when the cybercap comes out, you know, buy 100 of them, put them out to work and we'll have a fleet of Cyber cabs generating revenue for you.”
Listen at 2:15:05
Robotaxis will shift vehicle economics from ownership toward autonomous revenue generation.
“this, this changes from purchasing a car to autonomous revenue miles.”
Listen at 2:16:40
Robotaxis shift vehicle economics from ownership toward autonomous revenue miles.
“this changes from purchasing a car to autonomous revenue miles.”
Listen at 2:16:40
Electric autonomous vehicles could become the cheapest transportation mode for poor populations.
“the poorest people in the planet will be driven around by electric autonomous vehicles. It will be the cheapest mode of transportation.”
Listen at 2:17:48
Somnia Lab built a robot designed for intimate relationships.
“Somnia Lab has built a robot for intimate”
Listen at 2:20:09
Internet-video pretraining scales better than human motion-capture data.
“it doesn't scale nearly as well as just pre training off of Internet video is my guess.”
Listen at 2:23:58
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.
Books & mentions
Some links are affiliate links — PodLume may earn a commission if you buy.
Weekly email
State of the Pod, every Monday
The week's most-discussed names and stories from podcasts, with the moments behind them. Free, five minutes, unsubscribe in one click.

