Bloomberg Tech
Bloomberg Tech

Aug 18, 2026 · 53 min

AI’s next test: financing, safety and the teen market

Anthropic’s $65 Billion Surge, OpenAI’s Teen Push

The episode connects AI’s commercial expansion to the financial, regulatory and security risks that could shape its next phase.

3 key takeaways
  1. 1Anthropic’s revenue surge intensifies competition with OpenAI while exposing uncertainty around how AI growth is measured.
  2. 2GPU financing and concentrated technology spending are creating new market risks beyond the semiconductor companies driving the cycle.
  3. 3Teen safeguards, frontier-model stress tests and venture strategies show AI’s hardest problems now extend beyond model performance.

Don't miss

Dan Lahav explains why controlled cybersecurity testing is safer than discovering frontier-model failures after deployment.

The brief

Anthropic’s sharply higher annualized revenue raises the stakes for a potential IPO and sharpens comparisons with OpenAI, even as differing measurement methods complicate the growth story.

Denny Fish argues that hyperscaler balance sheets and an unusually strong semiconductor cycle are increasing risk, while software companies such as Microsoft offer a broader way to invest in AI.

OpenAI’s teen-focused ChatGPT is designed around learning, age estimation, parental controls and safeguards against presenting the system as a friend or sentient companion.

The episode widens the lens to Meta’s youth-safety trial, Apple’s position as an anti-AI stock and the financing structures behind NVIDIA’s expanding GPU ecosystem.

Dan Lahav makes the security case for controlled stress tests: frontier models may improve faster than defenses, making the transition period especially difficult to manage.

GV’s Dave Min closes the investment arc by favoring application-layer companies, founder-led teams and AI systems aimed at scientific breakthroughs, despite rising valuations and concentration.

What was said on this episode

33 statements · 22 positive · 8 negative · 3 neutral

  1. Denny Fishon HyperscalersNegative6:04

    Hyperscalers may issue 2–3% equity over the next couple years to fund AI investment.

    “for the next couple of years, if you're a hyperscaler, you'll probably issue 2% or 3% of equity and accept the dilution”

    Listen at 6:04

  2. Denny Fishon Hyperscaler equity issuanceNegative6:06

    Hyperscalers may issue 2–3% equity to finance AI spending.

    “if you're a hyperscaler, you'll probably issue 2% or 3% of equity and accept the dilution”

    Listen at 6:06

  3. Denny Fishon Leading-edge semiconductorsPositive6:59

    Leading-edge semiconductors are in the strongest cycle ever observed.

    “we are in the most pronounced cycle that we've ever had for leading edge semiconductors.”

    Listen at 6:59

  4. Denny Fishon AI supply chainPositive7:19

    The AI supply chain exceeded expectations during the latest earnings season.

    “The AI supply chain, if anything, pretty much exceeded expectations across the board.”

    Listen at 7:19

  5. Denny Fishon AI-related estimatesPositive7:25

    Estimates for AI-related results in 2027 and 2028 increased after earnings.

    “the estimates for 2027 and 28 just went up a bunch more than we had before the reporting season.”

    Listen at 7:25

  6. Denny Fishon AI compute spot pricingPositive8:03

    Spot pricing for AI compute remains strong.

    “spot pricing continues to be really. strong”

    Listen at 8:03

  7. Denny Fishon Enterprise, neocloud, and sovereign AI spendingPositive8:24

    Enterprise, neocloud, and sovereign AI spending is growing rapidly.

    “And that's growing really fast.”

    Listen at 8:24

  8. Denny Fishon AI investment portfolio strategyPositive8:43

    The portfolio is seeking non-semiconductor ways to participate in AI.

    “we're trying to find ways to participate in this market that are non-semiconductor”

    Listen at 8:43

  9. Denny Fishon Microsoft CorporationPositive8:58

    Microsoft may offer substantial investment gains as it re-rates.

    “there's going to be a lot of money to be made with companies. As Microsoft starts to re-rate”

    Listen at 8:58

  10. Denny Fishon Microsoft valuationPositive9:01

    Microsoft is beginning to re-rate after its recent quarterly results.

    “As Microsoft starts to re-rate, for example”

    Listen at 9:01

  11. Denny Fishon Microsoft AI strategy and returnsPositive9:20

    Concerns about Microsoft’s AI returns and strategy are beginning to ease.

    “some of those concerns, I think, are starting to be put to bed.”

    Listen at 9:20

  12. Denny Fishon NVIDIA-based cloud infrastructure financingPositive10:38

    Third-party capital is suitable for financing NVIDIA-based cloud infrastructure.

    “it makes a lot of sense to have third-party capital to help finance that.”

    Listen at 10:38

  13. Dan Lahavon Meta trial penaltiesNegative18:56

    Meta could face penalties as high as $1.4 trillion if states prevail.

    “The $1.4 trillion is a huge number, and that's because it's at the very top end of the estimate of what penalties might be in this case if the states are successful at trial.”

    Listen at 18:56

  14. Dan Lahavon Meta trialNeutral20:15

    The Meta trial is scheduled to continue until September 23.

    “the trial is scheduled at this point to go until September 23rd.”

    Listen at 20:15

  15. Dan Lahavon ApplePositive26:29

    Apple would likely benefit if the AI trade reverses.

    “Apple would probably benefit from this kind of trend.”

    Listen at 26:29

  16. Dan Lahavon ApplePositive26:32

    Apple may experience limited downside because it lacks AI-related capital-expenditure risk.

    “It might have some muted downside just because it has no AI-related CapEx risk.”

    Listen at 26:32

  17. Dan Lahavon AI model pre-deployment testingPositive29:03

    AI models require laboratory testing before deployment for safety and security.

    “there is a necessity to use them and just test them in a lab environment to make sure that they're safe and secure.”

    Listen at 29:03

  18. Dan Lahavon AI model pre-deployment testingNegative31:32

    Skipping pre-deployment testing makes real-world failures more likely and more serious.

    “if you're not going to discover that in a pre-deployment test, you're going to discover that when it happens in the world, and that's a much greater issue.”

    Listen at 31:32

  19. Dan Lahavon AI model deployment failuresNegative32:35

    Discovering AI failures after deployment is substantially worse than controlled testing.

    “The alternative would be discovering something bad happened after deployment, which is much, much worse.”

    Listen at 32:35

  20. Dan Lahavon Traditional cybersecurity monitoring toolsNegative33:03

    Traditional cybersecurity monitoring tools have limited effectiveness for AI systems.

    “Classic cyber monitoring tools only go so far in an AI context.”

    Listen at 33:03

  21. Dan Lahavon AI cybersecurity defensePositive34:39

    AI offers substantial opportunities to improve cybersecurity defense.

    “There is a huge opportunity in what AI can do for the defender side.”

    Listen at 34:39

  22. Dan Lahavon AI code security reviewPositive34:48

    AI may review every line of code for security within a few years.

    “we can live in a world in a few years where AI is going to review every line of code for security elements.”

    Listen at 34:48

  23. Dan Lahavon Frontier AI model performancePositive35:27

    Frontier AI models may continue doubling performance for at least two years.

    “At least for the next two years.”

    Listen at 35:27

  24. Dan Lahavon AI security verificationPositive36:11

    AI systems could eventually verify sandboxes and detect security issues.

    “over the long time horizon, we can have AIs also verifying, you know, just like sandboxes, verifying that there are no issues.”

    Listen at 36:11

  25. Dan Lahavon AI security testing expertiseNeutral37:21

    Effective AI security testing requires deep expertise in both cybersecurity and AI.

    “We need people that have a real dual expertise, both a very deep cyber background and a very deep AI background”

    Listen at 37:21

  26. Denny Fishon ModularPositive42:02

    Modular enables AI models to run across different AI chips.

    “The company, Modular, is the software layer that sits on top of technology, sits on top of AI chips to make sure that any model can run on any chip.”

    Listen at 42:02

  27. Denny Fishon AI for sciencePositive46:27

    AI for science will be a major investment opportunity over the next decade.

    “Right now, we're really excited about science. I think that AI for science is going to be a fascinating place to be in the next 10 years.”

    Listen at 46:27

  28. Denny Fishon AI for sciencePositive46:30

    AI for science will attract significant interest over the next decade.

    “AI for science is going to be a fascinating place to be in the next 10 years.”

    Listen at 46:30

  29. Denny Fishon AI for sciencePositive46:36

    AI combined with leading science may produce breakthroughs in materials, physics, and mathematics.

    “using the best models with the best science in the world to find breakthroughs in things like material science, physics, math.”

    Listen at 46:36

  30. Ed Ludlowon Interlocking directorate lawNegative48:04

    A century-old law prohibits investors from serving on competing company boards.

    “you can't, that investors can't serve on the boards of competing companies.”

    Listen at 48:04

  31. Dan Lahavon Andreessen HorowitzNeutral49:03

    The DOJ will decide whether to settle or sue after investigating Andreessen Horowitz.

    “after doing an investigation, then they decide whether and how to settle it or to file a lawsuit.”

    Listen at 49:03

  32. Denny Fishon Andreessen Horowitz antitrust investigationPositive50:18

    The government could drop the Andreessen Horowitz investigation.

    “It could also just drop the investigation altogether”

    Listen at 50:18

  33. Lauren Jonason Andreessen HorowitzNegative50:25

    The DOJ could sue to force a director to leave a board.

    “the most unlikely outcome is it could sue to force one of the directors to step down”

    Listen at 50:25

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.