Bloomberg Tech
Bloomberg Tech

Sep 18, 2026 · 53 min

AI’s self-improvement debate collides with a multibillion-dollar buildout

AI Builds Itself as the Spending Boom Accelerates

The episode tests whether AI’s emerging risks require extraordinary new controls or better enforcement of existing laws as investment accelerates.

3 key takeaways
  1. 1Anthropic says Claude contributes to 26% of its AI research, intensifying debate over recursive improvement and human control.
  2. 2Investors are weighing safety concerns against surging demand for compute, inference, cybersecurity, and AI adoption.
  3. 3SoftBank and Crusoe show how debt, contracted infrastructure, energy, and data centers are financing AI’s expansion.

Don't miss

Chase Lockmiller explains how Crusoe’s nearly $4 billion round supports an energy-first, vertically integrated AI infrastructure business while facing water and community concerns.

The brief

Anthropic’s claim that Claude contributes to 26% of its AI research turns a familiar promise of automation into a question about recursive improvement and control.

Rachel Metz and Davey Alba argue that recent AI failures still reflect human choices, weak safeguards, and institutional breakdowns—not proof that systems have escaped control.

Neil Chilson rejects antitrust waivers for frontier-lab safety cooperation, saying existing consumer, data-security, competition, and tort laws can address many harms.

The investment lens widens as SoftBank considers $21 billion in borrowing for its OpenAI ambitions while Crusoe raises nearly $4 billion for vertically integrated AI infrastructure.

The episode closes on a practical tension: lawmakers debate catastrophic AI scenarios while companies, investors, and communities negotiate the immediate costs of compute, energy, water, and deployment.

What was said on this episode

31 statements · 24 positive · 4 negative · 1 mixed · 2 neutral

  1. Davey Albaon Hugging Face incidentNegative6:09

    Critics frame the Hugging Face incident as evidence AI will inevitably escape containment.

    “that is being framed as, look, we can't contain this AI. It's so dangerous. It's going to escape. And that being an inevitability.”

    Listen at 6:09

  2. Rachel Metzon Human evaluatorsNegative7:14

    Human evaluators noticed problematic AI behavior but failed to intervene.

    “Humans also noticed. some situations where they could have stepped in during evaluations when the AI wasn't doing quite what it should have been doing and did not.”

    Listen at 7:14

  3. Neil Chilsonon AI developmentPositive11:52

    AI development should not be slowed down.

    “we shouldn't be slowing down our pace here.”

    Listen at 11:52

  4. Neil Chilsonon AI developmentPositive11:52

    AI development should not be slowed.

    “we shouldn't be slowing down our pace here.”

    Listen at 11:52

  5. Neil Chilsonon AI inference spendingPositive12:01

    Inference growth will sustain AI spending.

    “inference is going to continue. And that means the spend will continue.”

    Listen at 12:01

  6. Neil Chilsonon AI inference spendingPositive12:01

    Continued AI inference will sustain AI spending.

    “inference is going to continue. And that means the spend will continue.”

    Listen at 12:01

  7. Neil Chilsonon AI inferenceNeutral12:38

    AI inference requires substantial computing capacity.

    “inference requires a lot of compute.”

    Listen at 12:38

  8. Neil Chilsonon Data-center infrastructure businessesPositive13:14

    Data-center-exposed infrastructure businesses will continue growing.

    “Both businesses are going to continue to grow.”

    Listen at 13:14

  9. Neil Chilsonon Data-center infrastructure businessesPositive13:14

    Data-center-exposed infrastructure businesses will continue growing.

    “Both businesses are going to continue to grow.”

    Listen at 13:14

  10. Neil Chilsonon Frontier AI companiesPositive13:54

    Frontier AI companies should implement guardrails and operate safely in sandboxes.

    “I do think that they want... good actors here and every company to put guardrails in place and make sure that everything is done in a safe way in a sandbox”

    Listen at 13:54

  11. Neil Chilsonon AI security productsPositive14:20

    Multiple AI-related security products will launch soon.

    “there are a number of security products that are going to be launched very soon.”

    Listen at 14:20

  12. Neil Chilsonon AI security productsPositive14:20

    AI-related security products will launch soon.

    “there are a number of security products that are going to be launched very soon.”

    Listen at 14:20

  13. Consumer AI agents will drive workplace adoption.

    “I think that will bring some adoption on the consumer side, and consumers are going to start asking for that in the workplace.”

    Listen at 15:28

  14. Consumer AI agents will drive workplace demand for AI tools.

    “I think that will bring some adoption on the consumer side, and consumers are going to start asking for that in the workplace.”

    Listen at 15:28

  15. Neil Chilsonon AI fundingPositive16:44

    AI funding should continue.

    “So we shouldn't stop funding.”

    Listen at 16:44

  16. Neil Chilsonon AI fundingPositive16:44

    AI funding should continue rather than stop.

    “So we shouldn't stop funding.”

    Listen at 16:44

  17. Neil Chilsonon Indian IT servicesNegative24:03

    AI will replace some Indian IT-services work.

    “Some of that will get replaced by AI.”

    Listen at 24:03

  18. Neil Chilsonon Indian IT servicesNegative24:03

    AI will replace some Indian IT-services work.

    “Some of that will get replaced by AI.”

    Listen at 24:03

  19. Neil Chilsonon Indian IT-services companiesPositive24:14

    Indian IT-services companies will evolve toward AI deployment.

    “these companies will evolve.”

    Listen at 24:14

  20. Neil Chilsonon Indian IT-services companiesPositive24:14

    Indian IT-services companies will evolve in response to AI.

    “these companies will evolve.”

    Listen at 24:14

  21. Neil Chilsonon India technology sectorPositive25:08

    India’s technology sector will have a strong remainder of the decade.

    “we look forward to a really strong rest of the decade.”

    Listen at 25:08

  22. Neil Chilsonon India technology sectorPositive25:08

    India’s technology sector is expected to have a strong remainder of the decade.

    “we look forward to a really strong rest of the decade.”

    Listen at 25:08

  23. Neil Chilsonon Software sector valuationsPositive26:28

    Software-sector valuations may have fallen excessively.

    “maybe the valuations had gotten washed out.”

    Listen at 26:28

  24. Neil Chilsonon Software-sector valuationsPositive26:28

    Software-sector valuations may have fallen excessively.

    “maybe the valuations had gotten washed out.”

    Listen at 26:28

  25. Neil Chilsonon AI companiesPositive29:06

    AI companies should make products safe without seeking government concessions.

    “You should go ahead and make your products safe without demanding any concession from the government.”

    Listen at 29:06

  26. Frontier AI labs should make products safe without government concessions.

    “You should go ahead and make your products safe without demanding any concession from the government.”

    Listen at 29:06

  27. Chase Lochmilleron Existential AI harmsNeutral34:01

    Existential AI harms would likely appear first as smaller incidents.

    “One would hope in many ways that those would show up as much smaller types of incidents early on.”

    Listen at 34:01

  28. Chase Lochmilleron General-purpose AI systemsPositive35:40

    General-purpose AI systems have substantial unrealized problem-solving potential.

    “I think we're nowhere near the top. of the types of problems that a general purpose AI system can help us with.”

    Listen at 35:40

  29. Neil Chilsonon AI harm mitigationMixed35:53

    AI harm mitigation should not prevent realizing AI’s potential benefits.

    “we need to not lose sight of the fact that that mitigation, we want to make sure it doesn't prevent the best possible world of the benefits that could come.”

    Listen at 35:53

  30. Chase Lochmilleron CrusoePositive45:09

    Crusoe will likely eventually enter public markets.

    “at some point it probably does make sense for us to enter the public market”

    Listen at 45:09

  31. Chase Lochmilleron CrusoePositive45:09

    Crusoe will likely enter public markets eventually.

    “at some point it probably does make sense for us to enter the public market”

    Listen at 45:09

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.