TBPN EPISODE SUMMARY

OpenAI–AMD Deal, DevDay Reactions, xAI’s Memphis Datacenter | Doug O'Laughlin, Celine Halioua

Podcast: TBPN | Hosts: John Coogan, Jordy Hayes
Date: October 6, 2025

Episode Overview

This episode dives deep into the recent flurry of AI tech news including major OpenAI partnership deals—most notably the OpenAI–AMD agreement, reactions to OpenAI’s DevDay product launches, and an analysis of xAI’s “Colossus” data center efforts in Memphis. The hosts are joined by semiconductor analyst Doug O’Laughlin to unpack the implications of these unprecedented chip deals and by Loyal CEO Celine Halioua, who shares her journey developing lifespan-extension drugs for dogs. The show is fast-paced, insightful, and peppered with humorous takes on the rapidly evolving tech landscape.

Key Discussion Points & Insights

1. OpenAI DevDay: Product Announcements and Implications

[09:34–18:37]

• Product Launches:

  • ChatGPT Apps—integrating with companies like Zillow and Figma for in-app prompts ([09:41]).
  • Agent Builder—a node-based, Zapier-like workflow tool for more intelligent automation (“Is it like Zapier but with more if-statement intelligence?” – Jordy, [10:20]).
  • Codex—now GA (general availability), no major upgrades.
  • APIs—Sora 2 API at $0.10/second (significantly cheaper than expected), with a Pro tier at $0.30/second ([12:04]).

• Market Segmentation:

  • Consumer vs. Enterprise vs. Prosumer—AI is bifurcating into three business models:
    • Free/ad-supported consumer use (knowledge retrieval/fun)
    • Enterprise SaaS (API-driven workflows)
    • “Prosumer”/pay-as-you-go for power users—potentially a key growth market ([16:36–18:37]).
  • Consumption-based pricing for heavy users could change how AI tools are adopted and paid for.

• On Sora as Platform vs. API:

  • OpenAI opening Sora API signals lack of intent to turn Sora into a social app/content destination. “If you thought it could be a new social video network, you’d keep the tool exclusive.” (John, [23:18])
  • App adoption patterns and the creation/posting dynamic are contrasted with early Instagram.

2. AI Hardware Arms Race: OpenAI Deal Frenzy, AMD, Nvidia, and Market Reactions

[26:07–54:50]

• Circular Deal Structures:

  • OpenAI is “creating gravity” by involving all major hardware players—Nvidia, AMD, Oracle, Broadcom—making AI infrastructure everyone’s problem (“It's everyone’s problem now. Everyone has a vested interest in OpenAI.” – Doug, [57:13]).
  • OpenAI’s strategy: secure warrants/partnerships at every layer (Doug calls it “too big to fail,” [57:15]).

• OpenAI–AMD Deal Details ([38:54]):

  • 5-year agreement; OpenAI to buy 6GW-equivalent lots of AMD’s MI450 chips (directly or via cloud partners like Oracle).
  • AMD CEO Lisa Su promises “tens of billions” in new revenue—OpenAI receives warrants for up to 10% of AMD (160 million shares at $0.01) if deployment milestones are hit.
  • Incentives for “mutual success” between OpenAI and AMD.
  • Hypothesis: OpenAI may inference on AMD, train on Nvidia (Jordy, [43:13]).

• Comparison with Nvidia, Broadcom, and Industry Trends:

  • Nvidia seen as preferred partner (“Nvidia remains the preferred chip supplier,” [44:27]), but everyone (including hyperscalers) is hedging with custom chips (Google TPU, Amazon’s Trainium, etc.).
  • Circular capital flows (e.g., Nvidia invests in OpenAI, OpenAI then re-invests in chips): questioned as “dotcom-bubble-style” but also part of today’s megacorp financing (Doug, [59:05]).
  • “You never really lock down Sam.” (Jordy on Altman’s partnership maneuvering, [37:51]).

3. AI Infrastructure, Power, and the Coming Energy Crunch

[47:03–72:06]

• Massive Data Center Builds:

  • OpenAI’s chip deals are expressed in gigawatts, and energy demand is at the heart of scalability (OpenAI planning for 1GW in the second half of next year, [46:34]).
  • Musk’s xAI building “Colossus” data centers in Memphis—significant local debate over power/water ([50:12]).

• Where Will the Power Come From?

  • Doug: The grid isn’t built for 100% sustained load; expect on-site (“behind the meter”) gas gen, and a scramble for supply ([69:49]).
  • “Natural gas energy consumption is just going to go through the roof.” – Doug, [69:49].
  • International expansion likely; possible role for Oklo/fusion/nuclear if openAI leans in.

• Financing the Build-out:

  • Hyperscalers have massive cash stockpiles but minimal leverage (“No one is levered at all,” [60:38]).
  • Potential for more private credit and increased leverage (Jordy: “If you use 250B FCF at 5% coupon, that's 5 trillion in notional debt,” [74:10]).
  • “When you all run out together, that’s the problem… a debt-fueled binge.” (Doug, [74:16]).
  • Mag 7 (big tech) could lever up and raise huge funds if needed.

4. xAI, Tesla, and the Value of Colossus

[50:12–89:20]

• Colossus & Asset Valuation:

  • xAI’s Memphis data center (Colossus 1 & 2) houses >200,000 Nvidia chips; new one will be even larger.
  • Musk’s ability to finance “never underestimate” (“If there’s someone you should never short, it’s Elon Musk.” – Doug, [85:04]).
  • Colossus 2 by itself could be worth “tens of billions” on breakup value ([88:48]).

• Speculation on xAI’s Future:

  • Possible merger of xAI into Tesla to solidify financing and asset base.
  • Larry Ellison cited as a continual resource for Musk.
  • Concern over whether xAI can maintain burn without product market fit—but hard to bet against Musk’s fundraising ability.

5. Industry Competition: Old Tech, New Tech, and The Threat of AI to Incumbents

[19:34–24:39; 91:32–End]

• OpenAI Encroaching Across the Stack:

  • OpenAI essentially now a “large tech company; every founder must ask ‘what if OpenAI does this?’” (John, [19:13]).
  • Previous examples of tech giants failing to crush verticals despite forays (see Facebook/Google rivalry discussion, [20:05–22:56]).

• Sora and Platform Strategy:

  • Debate on whether Sora will become a consumption/social network or remain a creative tool.
  • “If you can create a place where every Sora generation ends up in the feed…”—the missed Instagram opportunity, but also, lower creation/post ratio observed for AI video ([26:34]).

• App Store Chatter:

  • Sora and ChatGPT shooting to #1/2 in App Store downloads, above TikTok and Instagram ([29:31]).

6. Dog Longevity Drugs: Interview with Celine Halioua (Loyal)

[112:39–137:54]

• Celine’s Background & Loyal’s Mission:

  • Stemmed from working on human longevity, pivoted to dogs for feasibility.
  • Marketed as “drugs to extend dog lifespan”—huge market, high willingness to pay across SES (Celine: “I’ve had people offer me $50,000 for the drug before,” [117:35]).

• Science and Regulatory Path:

  • Targeting IGF1: “The thing that makes a Newfie grow large also governs its rate of aging” ([119:35]); their drugs attempt to shift a big dog’s biology toward that of a smaller, longer-lived breed.
  • Rigorous regulatory pathway—first ever FDA efficacy approval for a longevity drug.
  • Dog drug development is orders of magnitude faster & cheaper than in humans ([126:08]): 4–7 years and ~$20–50M for dogs vs. 10 years, $1B+ for humans.

• Potential for Human Spinout and Biotech Industry Context:

  • Dogs as a better analog/model than mice for translational aging research.
  • The commercial system for dogs (pet owners spend anything, broader impact possible) and advantages in FDA exclusivity for animal drugs ([130:37]).
  • No AI hype here: Celine is skeptical of real-world impact of AI-for-drug-discovery until much better underlying data exists ([132:07]).

• Work Culture and Leadership:

  • Fully remote team; “You get so much more benefit from hiring the best person for the job, no matter where they live” ([136:26]).

Notable Quotes & Memorable Moments

• On OpenAI’s Agent Builder:

  • “It’s almost like a Zapier… just a little more intelligent but with if-statements.” – Jordy, [10:20]
  • “I see this thing potentially creating a prosumer, consumption-priced tier in the market for OpenAI.” – Jordy, [16:36]

• On Sam Altman’s Deal-Making:

  • “There’s a world in the future where OpenAI needs a bridge round and can go to all these suppliers and say, ‘if you don’t pony up, we all go down.’” – Doug, [57:15]
  • “You never really lock down Sam.” – John, [37:51]

• On AI Bubble Talk:

  • “It’s contrarian at this point to say this is good, healthy, and natural.” – John, [32:48]
  • “It’s not tech bubble crazy—the valuation would have to be 2x from here… Last time the tech bubble was concentrated in small unprofitable companies. Today it’s the seven largest companies in the world.” – Doug, [59:05]

• On Power Constraints:

  • “Natural gas energy consumption is just going to go through the roof.” – Doug, [69:49]

• On AI’s Threat to Creators:

  • “If you’re making content that is at all substitutable with slop, you’re done. But if you’re making something uniquely human, you can’t be replaced.” – Jordy, [107:59]

• On Loyal’s Market:

  • “A certain subset of the population will spend anything on their dog. Not actually correlated to SES or how much cash people have in the bank… I’ve had people offer $50,000 for the drug.” – Celine, [117:13]

• On AI-for-Science Hype:

  • “I debated whether we should just make AI part of our story for valuation, but honestly, we’re not there. Biology is the ultimate black box.” – Celine, [132:07]

• On Remote Biotech Teams:

  • “We have to hire the best possible person for each skill set we need, and often those people are not actually, they’re almost never in the Bay Area.” – Celine, [136:26]

Timestamps for Important Segments

• [03:03] – AI hardware device speculation
• [09:34] – OpenAI DevDay: products launched
• [12:04] – Sora 2 API pricing discussion
• [16:36] – Prosumer model & AI consumption economics
• [26:07] – AMD–OpenAI deal explainer
• [38:54] – AMD deal specifics and strategic intent
• [47:03] – Energy grid and scaling AI infrastructure
• [50:12] – xAI Memphis datacenter story (Colossus)
• [56:20] – Doug O'Laughlin joins: chip market analysis
• [60:38] – Bubble analogies and capital structure
• [69:49] – Power supply & data center energy strategies
• [88:48] – The asset value of xAI’s “Colossus 2”
• [112:39] – Celine Halioua interview: Loyal and dog longevity
• [126:08] – Dog drug development: timelines and costs
• [132:07] – Discussion of AI in drug discovery
• [136:26] – Loyal’s remote team & company culture

Additional Memorable Moments

• [53:38] – GPT-generated X posts benchmarked vs. real “viral” content (“Who’s with me on turning this month into pure spooky magic?” – Grok, [54:50])
• [95:02] – Gong hit analytics and segment breakdown (“Are gongs louder as funding increases?”)
• [134:34] – Dogs on Ozempic: Celine’s take (“It takes all the joy—the dog becomes a cat.”)

Tone & Style

The episode is energetic, irreverent, and deeply knowledgeable. Hosts John and Jordy share banter, skepticism, and excitement around AI’s future, with plenty of inside jokes and meme references. Guests are given space for detailed technical explanations, blending frank industry critique with strategic speculation.

Useful For:

• Anyone tracking the business, technical, and societal impacts of generative AI, especially the coming hardware and energy crunch.
• Sector investors, founders, or technologists navigating the new “AI stack”.
• Biotech/healthtech professionals interested in the translation of longevity research into the consumer pet market.
• Fans of fast-paced, sometimes-meme-laden technology commentary unconstrained by corporate PR.

End of Summary