TBPN
TBPN

Sep 14, 2026 · 2h 50m

Listen from 2:15:42

Listen at 2:15:42

AI’s race raises questions about safety, supply chains and durable businesses

The AI Slowdown Debate | Nico Wittenborn, Scott Keogh, Mitchell Green, David Rosenthal & Ben Gilbert, Faraj Aalaei

The episode connects frontier-AI governance to semiconductor bottlenecks, cybersecurity, consumer software, industrial strategy and the economics of consolidation.

3 key takeaways
  1. 1AI safety debates increasingly turn on practical levers across chips, energy, regulation and cybersecurity rather than abstract principles alone.
  2. 2Solo-GP investing, Bending Spoons-style acquisitions and specialized AI reveal how software markets are reorganizing around speed and focus.
  3. 3Scout, Home Depot and semiconductor design show how durable advantages still depend on physical systems, industry knowledge and customer behavior.

Don't miss

The episode’s sharpest turn comes when the abstract AI slowdown debate is tested against cybersecurity attacks and the physical supply chains behind advanced computing.

The brief

The hosts frame the central dispute: frontier AI may need safety constraints, but slowing development could also weaken innovation and strategic advantage.

Nico Wiesnet describes Adjacent’s solo-GP model, consumer-software thesis and interest in a market where AI is reshaping products, incumbents and acquisition cycles.

Scott Keogh makes the AI debate tangible through Scout’s clean-sheet electric vehicles, range-extender strategy, physical controls and plan to earn trust through ownership.

Mitchell Green connects enterprise-software durability, cybersecurity risk and Bending Spoons-style consolidation, arguing that viable but overcapitalized companies may become targets.

The episode closes on Home Depot’s scale and Cognit Chip’s specialized AI opportunity, suggesting that durable businesses and domain-specific data may matter more than general hype.

What was said on this episode

114 statements · 59 positive · 39 negative · 2 mixed · 14 neutral

  1. John Cooganon RhabdomyolysisNegative0:51

    Rhabdomyolysis can result from excessive exercise and dehydration and is avoidable.

    “It comes from working out too much and not being hydrated. Very avoidable.”

    Listen at 0:51

  2. John Cooganon AI prevention of rhabdomyolysisPositive0:58

    Any risk is worthwhile if AI has a one-percent chance of preventing rhabdomyolysis.

    “If there's even a 1% chance that AI can find a prevention for rhabdo, I think it's worth any amount of risk.”

    Listen at 0:58

  3. John Cooganon Government approval for frontier AI coordinationNeutral10:10

    Formal coordination among frontier AI labs requires government approval.

    “We need government approval.”

    Listen at 10:10

  4. AI will produce substantially more benefits than harms.

    “But it's going to be more good than bad, but by a lot.”

    Listen at 17:30

  5. John Cooganon AI leadershipPositive17:37

    The country or company that wins AI will prevail broadly.

    “Whoever wins AI, and we're leading by a lot, whoever wins AI wins.”

    Listen at 17:37

  6. John Cooganon Alternative AI APIsNegative25:33

    Open-source and alternative AI APIs are less mature than frontier-lab offerings.

    “none of them are as mature as the frontier labs”

    Listen at 25:33

  7. Nico Wittenbornon Solo-GP venture fundsNeutral34:37

    Solo-GP funds constituted a majority of new funds last year.

    “last year, the majority of new funds were SoloGP funds.”

    Listen at 34:37

  8. Nico Wittenbornon AI sourcing systemsPositive37:06

    AI systems identify companies once they generate public signals and data.

    “the AI picks it up”

    Listen at 37:06

  9. Nico Wittenbornon AI in venture sourcingNeutral37:20

    Founders currently prefer human relationships over working directly with computers.

    “you still want a human-to-human interaction with a founder because they don't want to work with a computer. Not yet, at least.”

    Listen at 37:20

  10. Nico Wittenbornon Founder-investor relationshipsNegative37:20

    Founders still prefer human relationships over computer-mediated investor relationships.

    “you still want a human-to-human interaction with a founder because they don't want to work with a computer. Not yet, at least.”

    Listen at 37:20

  11. Nico Wittenbornon Bending SpoonsNegative40:11

    Bending Spoons acquires companies at roughly ten percent of peak market capitalization.

    “Benning Spoons buys for 10% of the peak market cap”

    Listen at 40:11

  12. Nico Wittenbornon Adjacent fundsNeutral41:09

    Adjacent’s first two funds primarily focused on consumer subscription companies.

    “the first fund and the second fund to the most degree were very focused on consumer subscription.”

    Listen at 41:09

  13. Nico Wittenbornon Consumer subscription companiesNegative41:21

    Consumer subscription companies experience more churn and eventually hit market ceilings.

    “they're much more churny, right? So at one point, depending on how deep the market is, you'll hit a ceiling”

    Listen at 41:21

  14. Nico Wittenbornon Consumer subscription company aggregationPositive42:14

    Aggregating consumer subscription companies offers a significant opportunity.

    “there's actually a really big play here to aggregate companies”

    Listen at 42:14

  15. Nico Wittenbornon Consumer AI applicationsNegative48:11

    AI has increased consumer-app competition, making venture-scale outcomes harder.

    “the competition increases significantly because so many more new apps. So getting to a venture scale outcome is even more. Even harder.”

    Listen at 48:11

  16. Nico Wittenbornon Consumer AI startupsNegative48:20

    Consumer AI has made venture-scale outcomes harder to achieve.

    “the venture outcomes are more difficult”

    Listen at 48:20

  17. Nico Wittenbornon Simple consumer AI applicationsNegative48:23

    Simple consumer AI applications are unlikely to justify venture investment.

    “if it's very simple, it doesn't make sense.”

    Listen at 48:23

  18. Nico Wittenbornon Personal AI agentsPositive50:11

    Personal-agent products are expected to produce many viable applications.

    “I do expect that there will be a lot of things that work there.”

    Listen at 50:11

  19. Nico Wittenbornon Personal-agent productsPositive50:11

    Personal-agent products will generate many viable consumer applications.

    “I do expect that there will be a lot of things that work there.”

    Listen at 50:11

  20. Nico Wittenbornon Personal-agent companiesNegative50:35

    Early investment in personal-agent companies is currently difficult.

    “it's very difficult for me to play there right now.”

    Listen at 50:35

  21. Nico Wittenbornon MetaMusePositive51:06

    Meta moved rapidly to launch Muse in response to the personal-agent trend.

    “the speed at which Meta moved on this personal agent trend and launched Muse is really impressive.”

    Listen at 51:06

  22. Nico Wittenbornon MetaMusePositive51:06

    Meta moved rapidly to launch Muse in response to personal-agent trends.

    “the speed at which Meta moved on this personal agent trend and launched Muse is really impressive.”

    Listen at 51:06

  23. Nico Wittenbornon Europe’s innovation ecosystemNegative53:54

    Europe has insufficiently supported innovation.

    “Europe, we have not done enough to support innovation.”

    Listen at 53:54

  24. Nico Wittenbornon EuropeNegative53:54

    Europe has insufficiently supported innovation through policy and institutions.

    “Europe, we have not done enough to support innovation.”

    Listen at 53:54

  25. John Cooganon European technology ecosystemNegative54:57

    Europe has not created conditions for a European-scale Google competitor.

    “Europe hasn't gone far enough to actually truly create an area where you could have a European Google”

    Listen at 54:57

  26. Nico Wittenbornon Europe in AINegative55:24

    Europe will find it difficult to catch up in AI.

    “I think it's very hard to catch up in AI.”

    Listen at 55:24

  27. Nico Wittenbornon European technology companiesPositive55:43

    European technology companies should focus on differentiated regional strengths.

    “It should be differentiated into their own strengths.”

    Listen at 55:43

  28. Nico Wittenbornon Europe’s technology strategyPositive55:43

    Europe should focus on areas matching its distinctive strengths.

    “It should be differentiated into their own strengths.”

    Listen at 55:43

  29. Nico Wittenbornon Venture investmentsNegative58:49

    Most venture investments fail to produce successful outcomes.

    “Most of the venture bets do not turn out.”

    Listen at 58:49

  30. Nico Wittenbornon Venture investmentsNegative58:49

    Most venture investments fail to produce successful outcomes.

    “Most of the venture bets do not turn out.”

    Listen at 58:49

  31. Scott Keoghon Scout MotorsPositive1:02:25

    Scout Motors has grown to 1,600 employees and is industrializing a South Carolina factory.

    “now fast forward, we have 1,600 employees. We're industrializing a factory in Blythwood, South Carolina.”

    Listen at 1:02:25

  32. Scott Keoghon Scout MotorsPositive1:02:28

    Scout Motors is industrializing a factory in South Carolina.

    “we're industrializing a factory in Blythwood, South Carolina.”

    Listen at 1:02:28

  33. Scott Keoghon American-made rugged SUVsPositive1:04:50

    Most rugged-SUV customers want products made in America.

    “the majority of the segment says, I want to buy things that are American. I want to buy things that are made in America.”

    Listen at 1:04:50

  34. Scott Keoghon Scout MotorsPositive1:04:56

    Scout Motors was always premised on building vehicles in America.

    “Scout's whole premise was always going to be that.”

    Listen at 1:04:56

  35. Scott Keoghon Scout MotorsPositive1:06:12

    Scout Motors has engineered its vehicle for high durability.

    “we've fundamentally engineered the car to be super robust.”

    Listen at 1:06:12

  36. Scott Keoghon Electric-vehicle batteriesPositive1:06:51

    Electric-vehicle batteries are retaining power, charge, and durability.

    “these batteries are holding their power, they're holding their charge, and they're holding their robustness.”

    Listen at 1:06:51

  37. Scott Keoghon Scout MotorsPositive1:07:52

    Scout vehicles will offer Level 2-plus driver assistance.

    “the Scout will have level two plus”

    Listen at 1:07:52

  38. Scott Keoghon Scout MotorsPositive1:08:59

    Scout Motors plans first customer deliveries in 2028.

    “In 2028, exactly.”

    Listen at 1:08:59

  39. Scott Keoghon Scout MotorsPositive1:09:19
    “we'll be roughly in the zone of 3,800 to 4,000”

    Listen at 1:09:19

  40. Scott Keoghon Scout MotorsPositive1:10:33

    Scout Motors will sell vehicles directly to consumers.

    “we are going to go direct to consumer”

    Listen at 1:10:33

  41. Scott Keoghon Scout MotorsPositive1:12:04

    Scout Motors plans to build 100 retail stores over time.

    “We will build 100 stores over time.”

    Listen at 1:12:04

  42. Scott Keoghon Scout MotorsPositive1:12:33

    Scout Motors will allocate 85% of investment to service infrastructure.

    “85% of the investment will go into service since that's where the most critical function is.”

    Listen at 1:12:33

  43. Scott Keoghon Scout MotorsNeutral1:14:31

    Scott Keogh expects 85% of early customers to test-drive Scout vehicles.

    “at this early phase, I would say 85% is going to want to get in this vehicle, drive it, test it out.”

    Listen at 1:14:31

  44. Scott Keoghon Scout MotorsNeutral1:14:33

    About 85% of early Scout customers will want a test drive.

    “85% is going to want to get in this vehicle, drive it, test it out.”

    Listen at 1:14:33

  45. Scott Keoghon Scout vehicleNeutral1:22:18

    The Scout vehicle uses body-on-frame construction and a solid rear axle.

    “it's body on frame. which means you have a frame you put the hat on top of the vehicle it's got a solid rear axle”

    Listen at 1:22:18

  46. Scott Keoghon Scout vehiclePositive1:22:31

    The Scout prototype has 1,000 pound-feet torque, 800 horsepower, and four-second zero-to-60 acceleration.

    “this is a thousand pound feet of torque 800 horsepower this vehicle zero 60 in four seconds”

    Listen at 1:22:31

  47. Scott Keoghon Scout MotorsPositive1:28:16
    “the vehicle is going to start in the high 50s.”

    Listen at 1:28:16

  48. Mitchell Greenon AI-driven race carPositive1:36:15

    A sufficiently funded AI-driven race car could outperform every professional driver.

    “I think it would be faster than any pro.”

    Listen at 1:36:15

  49. Mitchell Greenon AI-driven race carsPositive1:36:15

    A sufficiently funded AI-driven race car could outperform every professional driver.

    “I think it would be faster than any pro.”

    Listen at 1:36:15

  50. Mitchell Greenon Public software companiesPositive1:37:34

    Public software companies have recently reported strong earnings.

    “Public company software earnings have been pretty strong.”

    Listen at 1:37:34

  51. Mitchell Greenon AI and software agentsPositive1:38:08

    AI and agents will accelerate the pace of innovation.

    “the pace of innovation, because of AI and agents and all this stuff, is only going to increase.”

    Listen at 1:38:08

  52. Mitchell Greenon AI and software agentsPositive1:38:08

    AI and agents will accelerate the pace of software innovation.

    “the pace of innovation, because of AI and agents and all this stuff, is only going to increase.”

    Listen at 1:38:08

  53. Mitchell Greenon Companies failing to innovateNegative1:38:41

    Companies that fail to innovate will lose competitive position.

    “the people that are not innovating are going to get left behind.”

    Listen at 1:38:41

  54. Mitchell Greenon Companies failing to innovateNegative1:38:41

    Companies that fail to innovate will fall behind competitors.

    “the people that are not innovating are going to get left behind.”

    Listen at 1:38:41

  55. Mitchell Greenon Highly leveraged companiesNegative1:40:04

    Highly leveraged companies will struggle more to self-disrupt and innovate.

    “companies that are very very highly levered will find it harder to try to disrupt themselves and continue to innovate”

    Listen at 1:40:04

  56. Mitchell Greenon Highly leveraged companiesNegative1:40:04

    Highly leveraged companies will struggle more to innovate and disrupt themselves.

    “companies that are very very highly levered will find it harder to try to disrupt themselves and continue to innovate”

    Listen at 1:40:04

  57. Mitchell Greenon Frontier AI leadersNegative1:49:14

    Frontier AI leaders are genuinely concerned about excessively rapid AI progress.

    “those guys are like genuinely. concerned that if this stuff advances too fast”

    Listen at 1:49:14

  58. Mitchell Greenon Frontier AI developmentPositive1:49:26

    Frontier AI development needs regulation.

    “this stuff needs to be regulated”

    Listen at 1:49:26

  59. Mitchell Greenon Open-source artificial intelligencePositive1:49:28

    Open-source AI models are valuable for avoiding concentration of power.

    “I also think it's valuable that we have open source models”

    Listen at 1:49:28

  60. Mitchell Greenon AI regulationNeutral1:50:38

    AI will be regulated.

    “this stuff is going to be regulated.”

    Listen at 1:50:38

  61. Mitchell Greenon AI developmentNeutral1:50:38

    AI development will be regulated.

    “this stuff is going to be regulated.”

    Listen at 1:50:38

  62. Mitchell Greenon Government AI regulationNegative1:50:45

    Government may regulate AI excessively.

    “government over-regulate things”

    Listen at 1:50:45

  63. Mitchell Greenon AI-enabled voice fraudNegative1:53:15

    Voice fraud will increase dramatically as voice recreation becomes easier.

    “voice fraud is going to explode”

    Listen at 1:53:15

  64. Mitchell Greenon AI-enabled voice fraudNegative1:53:15

    Voice fraud will increase dramatically as voice cloning becomes easier.

    “voice fraud is going to explode”

    Listen at 1:53:15

  65. Mitchell Greenon Cybersecurity attacksNegative1:54:18

    A cybersecurity attack may be the world’s largest current risk.

    “the biggest risk today might actually be a cybersecurity attack, like the biggest global risk.”

    Listen at 1:54:18

  66. Mitchell Greenon Cybersecurity attacksNegative1:54:18

    A cybersecurity attack may be the world’s largest current global risk.

    “the biggest risk today might actually be a cybersecurity attack, like the biggest global risk.”

    Listen at 1:54:18

  67. Mitchell Greenon Internet outageNegative1:54:33

    A multi-day internet outage would be highly consequential.

    “imagine the internet goes down for like three days.”

    Listen at 1:54:33

  68. Mitchell Greenon Overcapitalized software companiesMixed1:56:19

    Many potential acquisition targets are viable businesses burdened by excess capital.

    “Some of these are good businesses. They're just overcapitalized.”

    Listen at 1:56:19

  69. Mitchell Greenon Bending Spoons-style software acquisitionsPositive1:56:48

    Buying a business at three times revenue can quickly repay itself with $60–70 million EBITDA.

    “if you're basically about a business where you pay three times revenues and you think you can run it for, you know, $100 million our business for $300 million our business. and you think you can run it with like 70 million of EBITDA or 60 million of EBITDA, well, then you basically paid yourself back.”

    Listen at 1:56:48

  70. Mitchell Greenon Bending SpoonsPositive1:56:48

    Buying software at three times revenue can quickly repay the purchase through EBITDA efficiencies.

    “if you're basically about a business where you pay three times revenues and you think you can run it for, you know, $100 million our business for $300 million our business. and you think you can run it with like 70 million of EBITDA or 60 million of EBITDA, well, then you basically paid yourself back.”

    Listen at 1:56:48

  71. Mitchell Greenon Future companiesPositive1:58:33

    Some of the world’s greatest companies will be created over the next twenty years.

    “there's going to be some of the world's greatest companies created over the next 20 years.”

    Listen at 1:58:33

  72. Mitchell Greenon Future technology companiesPositive1:58:33

    Some of the world’s greatest companies will be created over the next 20 years.

    “there's going to be some of the world's greatest companies created over the next 20 years.”

    Listen at 1:58:33

  73. Nico Wittenbornon The Home Depot, Inc.Positive2:02:56

    Home Depot’s IPO-day investment outperformed other major public stocks through the comparison date.

    “you would have if you had bought Home Depot. on the day of its IPO.”

    Listen at 2:02:56

  74. Mitchell Greenon The Home Depot, Inc.Positive2:03:24
    “$17 million.”

    Listen at 2:03:24

  75. Mitchell Greenon The Home Depot, Inc.Positive2:03:24
    “$17 million.”

    Listen at 2:03:24

  76. Mitchell Greenon Home Depot employee strategyPositive2:04:00

    Home Depot recruited tradespeople to provide specialized customer assistance.

    “they recruited former tradespeople. So plumbers, electricians, people that could actually help you with your home project.”

    Listen at 2:04:00

  77. Mitchell Greenon The Home Depot, Inc.Negative2:10:23

    Home Depot nearly failed around 2006–2007 before the housing crisis.

    “the company almost completely died in 2000. Six, seven going into the housing crisis.”

    Listen at 2:10:23

  78. Mitchell Greenon The Home Depot, Inc.Positive2:11:42

    Home Depot built substantial supply-chain and e-commerce capacity before COVID-19.

    “They built up an astonishing amount of supply chain and e-commerce readiness. in the three, four years leading up to it.”

    Listen at 2:11:42

  79. Mitchell Greenon U.S. single-family housing constructionNegative2:12:54

    The United States is underbuilding single-family housing.

    “America's not building enough housing, especially single family homes.”

    Listen at 2:12:54

  80. Mitchell Greenon U.S. single-family housing constructionNegative2:12:54

    The United States is underbuilding single-family housing.

    “America's not building enough housing, especially single family homes.”

    Listen at 2:12:54

  81. Mitchell Greenon The Home Depot, Inc.Positive2:13:07

    America’s housing shortage creates a major long-term tailwind for Home Depot.

    “Home Depot is experiencing like the most ridiculously awesome secular tailwind from that ever”

    Listen at 2:13:07

  82. Mitchell Greenon U.S. housing stockNeutral2:13:21

    The median U.S. home is 15–20 years older than when Home Depot was founded.

    “The median age of a home is up like 15 to 20 years versus when they were getting founded.”

    Listen at 2:13:21

  83. Mitchell Greenon U.S. housing stockNeutral2:13:21

    The median U.S. home is roughly 15–20 years older than when Home Depot was founded.

    “The median age of a home is up like 15 to 20 years versus when they were getting founded.”

    Listen at 2:13:21

  84. Mitchell Greenon Home Depot storesNeutral2:15:23
    “they essentially didn't build another store until two years ago.”

    Listen at 2:15:23

  85. Mitchell Greenon The Home Depot, Inc.Positive2:15:42

    Home Depot shifted expansion toward e-commerce and specialized fulfillment centers.

    “they just focused on e-commerce build out and building more fulfillment centers and specialized fulfillment centers for e-commerce.”

    Listen at 2:15:42

  86. Robotics will provide Home Depot with an additional growth tailwind.

    “I think that robotics are going to be another tailwind for them”

    Listen at 2:17:53

  87. Nico Wittenbornon Home Depot lumber deliveryPositive2:19:21

    Home Depot can deliver 3,000 pounds of lumber to homes within two hours.

    “you can get 3000 pounds of lumber in two hours at your house”

    Listen at 2:19:21

  88. Mitchell Greenon Home Depot China expansionNegative2:20:28

    Home Depot opened multiple Chinese stores, but DIY culture was weak there.

    “They opened a dozen plus stores, maybe dozens of stores. And yeah, interestingly, in China, it's not cool to be working on your house on your own.”

    Listen at 2:20:28

  89. Mitchell Greenon Home Depot warehouse-store modelPositive2:21:38

    Large warehouse stores generated higher ROI than smaller limited-assortment stores.

    “your ROI on a giant store that had everything was much higher than your ROI on a bunch of little stores that had a limited number of things”

    Listen at 2:21:38

  90. Mitchell Greenon The Home Depot, Inc.Neutral2:23:20

    About half of Home Depot’s e-commerce may be in-store pickup.

    “it might be like half of their e-commerce is actually pickup.”

    Listen at 2:23:20

  91. Mitchell Greenon The Home Depot, Inc.Neutral2:23:20

    About half of Home Depot’s e-commerce may consist of in-store pickup.

    “it might be like half of their e-commerce is actually pickup”

    Listen at 2:23:20

  92. Faraj Aalaeion Home Depot contractor businessPositive2:24:15

    Home Depot’s recent growth has primarily come from professional contractors.

    “a huge focus of the company and most of their growth for the past decade or so, COVID excluded, has been from pros, from contractors”

    Listen at 2:24:15

  93. Nico Wittenbornon Home Depot commercial contractor businessPositive2:24:59

    Home Depot has recently expanded further into large-builder and commercial supply workflows.

    “Historically, Home Depot has not been penetrated into that. They've been penetrating a lot more into that.”

    Listen at 2:24:59

  94. Mitchell Greenon Home Depot technology strategyMixed2:25:14

    Home Depot’s current technology focus is modest AI use rather than drone delivery.

    “But it's not drones. Like it's not any of the things you were just saying. It's AI a little bit.”

    Listen at 2:25:14

  95. Mitchell Greenon Semiconductor chip developmentNegative2:34:11

    Developing a modern semiconductor chip now costs several hundred million dollars.

    “And now it takes several hundred million dollars to do a chip. Cost of developing chips has skyrocketed.”

    Listen at 2:34:11

  96. Mitchell Greenon Modern chip developmentNegative2:34:12

    Developing a modern chip costs several hundred million dollars.

    “it takes several hundred million dollars to do a chip.”

    Listen at 2:34:12

  97. Mitchell Greenon chip designNegative2:34:21

    Designing a chip takes two to three years.

    “Time it takes to build a chip, to design a chip, it's two to three years.”

    Listen at 2:34:21

  98. Mitchell Greenon Semiconductor chip designNegative2:34:21

    Designing a chip typically takes two to three years.

    “Time it takes to build a chip, to design a chip, it's two to three years.”

    Listen at 2:34:21

  99. Mitchell Greenon Software and semiconductor developmentNegative2:34:35

    Software advances leave chip development behind by nearly six years.

    “the software is moving so fast that leaves chips behind by almost six years”

    Listen at 2:34:35

  100. Mitchell Greenon AI for semiconductor designPositive2:35:13

    AI can substantially reduce semiconductor design time and cost.

    “There was a great intersection of these two really awesome technologies to be able to solve all that problem, collapse that time by a large factor, reduce the cost it takes to do a chip.”

    Listen at 2:35:13

  101. Mitchell Greenon CognichipPositive2:36:36

    Cognichip’s enterprise product can accelerate chip design.

    “chip designers can use to make the chip design a lot faster”

    Listen at 2:36:36

  102. Mitchell Greenon AI models for chip designPositive2:36:43

    AI models can perform tasks consuming about ninety percent of chip engineers’ time.

    “90% of the time in the chip business, our engineers are spending doing things that can be done by these models now.”

    Listen at 2:36:43

  103. Mitchell Greenon AI semiconductor design modelsPositive2:36:43

    AI models can perform tasks consuming about 90% of semiconductor engineers’ time.

    “90% of the time in the chip business, our engineers are spending doing things that can be done by these models now”

    Listen at 2:36:43

  104. Mitchell Greenon AI-assisted chip designPositive2:39:38

    Shortening chip-development timelines reduces uncertainty about market requirements.

    “By collapsing the time, you know, the collapsing time that it makes to do a chip. you're actually getting rid of all of those unknowns.”

    Listen at 2:39:38

  105. Mitchell Greenon Accelerated chip developmentPositive2:39:49

    Shorter chip-development cycles increase the chance of matching market needs, power, and performance.

    “You're increasing the probability of you hitting the market dynamics that you want at the time you want it with the right kind of power and performance.”

    Listen at 2:39:49

  106. Mitchell Greenon General-purpose LLMs for chip designNegative2:41:54

    General-purpose LLMs perform poorly in chip design because relevant open data is scarce.

    “The reason large-scale LLMs are not good in chip design is because chip design data as an open source is actually not available.”

    Listen at 2:41:54

  107. Mitchell Greenon Large Language ModelNegative2:41:54

    Large-scale LLMs perform poorly in chip design because relevant open-source data is scarce.

    “The reason large-scale LLMs are not good in chip design is because chip design data as an open source is actually not available.”

    Listen at 2:41:54

  108. Mitchell Greenon Industry-specific AI modelsPositive2:42:28

    Industry-specific models trained on relevant data can become highly capable.

    “a specific model trained on a specific technology becomes very, very powerful”

    Listen at 2:42:28

  109. Mitchell Greenon Industry-specific semiconductor AI modelsPositive2:42:28

    Models trained on industry-specific semiconductor data can become highly powerful.

    “a specific model trained on a specific technology becomes very, very powerful”

    Listen at 2:42:28

  110. Mitchell Greenon Cognichip semiconductor datasetPositive2:43:04

    Cognichip believes it has built the semiconductor industry’s largest dataset.

    “for the last two and a half years has been building a data mode that we think today is the largest data set in the industry for semiconductors”

    Listen at 2:43:04

  111. Mitchell Greenon Semiconductor industry chip developmentNegative2:44:26

    The semiconductor industry cannot sustainably maintain four-year, hundreds-million-dollar chip development.

    “Our industry cannot go on taking four years to design a chip and cost hundreds of millions of dollars”

    Listen at 2:44:26

  112. Mitchell Greenon Semiconductor chip development economicsNegative2:44:26

    Current semiconductor development economics are unsustainable and unattractive for investment.

    “Our industry cannot go on taking four years to design a chip and cost hundreds of millions of dollars and not know whether there's going to be a good market for it or not. That's not sustainable. That's not investable.”

    Listen at 2:44:26

  113. Mitchell Greenon Faraj Aalaei’s first semiconductor companyPositive2:45:40

    Faraj’s first semiconductor company reached an IPO in roughly three years.

    “We went from opening the door to having a chip in hand that we were selling and generating revenue. And we took the company public almost three years to the day we opened the door.”

    Listen at 2:45:40

  114. Mitchell Greenon Faraj Aalaei’s first semiconductor companyPositive2:45:46

    Faraj’s first semiconductor company reached an IPO in nearly three years, a record he says still stands.

    “We took the company public almost three years to the day we opened the door. That still holds the fastest record that any productive company has gone from inception to going public.”

    Listen at 2:45:46

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.

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.

Listen to the full episode and explore every guest, topic, and moment on PodLume.

AI’s race raises questions about safety, supply chains and durable businesses · PodLume