Better Offline — Monologue: OpenAI and NVIDIA’s Deal Is BS

Host: Ed Zitron
Release Date: September 26, 2025
Podcast Network: Cool Zone Media / iHeartPodcasts

Episode Overview

In this solo monologue episode, tech industry veteran Ed Zitron launches a pointed, expletive-laden critique at the highly publicized “$100 billion” partnership between OpenAI and NVIDIA. Through detailed financial analysis, historical context, and insider industry commentary, Zitron debunks media buzz and lays bare the contradictions, impracticalities, and hype surrounding the deal. Ed argues that the entire arrangement is built on unrealistic promises, misleading headlines, and a disregard for fiscal reality, raising bigger questions about the future of the AI boom and tech investment as a whole.

Key Discussion Points & Insights

1. The NVIDIA–OpenAI “Investment” is Largely Theoretical

(03:13–05:00)

• Headline Discrepancies:
  • Media reported a $100 billion “investment,” but in reality, Ed emphasizes that NVIDIA has “not wired OpenAI a single goddamn dollar.”
  • The only hard number is a possible $10 billion, “allegedly” to be transferred in monthly tranches, but “nothing’s actually been signed.”
• Conditional Funding:
  • Most future funds are contingent on OpenAI building unprecedented new data center capacity — a process that is slow, expensive, and largely unproven at this scale.

2. Data Center Construction — Unrealistic Assumptions

(05:05–08:20)

• Physical and Financial Constraints:
  • Ed describes OpenAI’s lack of direct experience: “OpenAI really has built nothing themselves. Everyone else is building it for them.”
  • The so-called “Stargate” locations are mostly re-packaged announcements: “One of them is...already chosen to be one of Oracle’s data centers. This was already announced. They’re re-announcing stuff.”
• Building Timeline Realities:
  • Construction of major sites (like Abilene’s 1.2 GW facility for Oracle/OpenAI) is taking “about two and a half years per gigawatt.”
  • Headlines ignore how slow and capital-intensive this all is.

3. The Ludicrous Economics of the Deal

(08:20–10:40)

• Cost Calculations:
  • The Abilene venture needed $15 billion for 1.2 GW capacity. OpenAI would need “$125 billion just to build the data centers to get the [NVIDIA] funding, which will require another $200 billion in GPUs.”
  • OpenAI’s reported deal with Oracle: “promised to spend $400 billion over the next five years.” Realistic cash requirements over five years: approximately $600 billion.
• Complex Financing Structures:
  • NVIDIA may not sell but “lease” GPUs to OpenAI, through “a special purpose entity to buy the GPUs from themselves that they would then lease to OpenAI...that’s straight Enron shit.”
• Revenue Myths:
  • Media repeats OpenAI’s projections of $100–$145 billion revenue (not profit!) by 2028–2029, but the company’s current actual revenue is much lower (“about $6.26 billion” for the year).

4. The Immense Capital Drain and Industry Bubble

(10:40–12:20)

• Market Distortion:
  • Estimated total available capital in the top 10 private equity funds: about $477 billion; US venture capital: about $164 billion; another ~$150 billion in other funds.
  • Zitron: “Are we really going to carve out half to two-thirds of available global capital for one fucking company that burns a billion fucking dollars every two fucking seconds? ... There’s no parallel in history for any company...that’s ever spent this much.”
• Downstream Effects:
  • OpenAI consumes “33% of US venture capital last year,” and likely more this year. “None of these companies are profit generating. None of these products are popular.”
  • Estimated total AI industry revenue (counting all major providers) is just $55 billion in 2025, versus capital expenditures exceeding $400 billion.

5. Industry Delusion and Potential Implosion

(12:21–13:30)

• Hype and Misaligned Incentives:
  • Ed’s biting assessment: “Is anyone having fun? Does this feel like the future? Or does this feel like the future of how we’re going to blow hundreds of billions of dollars on nothing?”
• Market Consequences:
  • When the bubble bursts, “it’s going to fuck retail investors and a ton of venture capitalists too. Ton of private equity people as well, which will be kind of funny. But the financial apocalypse will hit everyone and it’ll be grisly.”

Notable Quotes & Memorable Moments

• On the “Investment” Reality:
  • “Nvidia has not wired OpenAI a single goddamn dollar.” (03:33)
• On Recycled Announcements:
  • “As a PR professional, when you’re re-announcing shit, it’s because you run out of stuff to announce... This is real.” (04:42)
• On Data Center Economics:
  • “OpenAI will need $125 billion just to build the data centers to get the funding from Nvidia, which will require another $200 billion in GPUs.” (08:50)
  • “I’ve really hesitated to bring up any Enron comparisons...that’s straight Enron shit.” (09:50)
• On AI Revenue Fantasies:
  • “OpenAI is projected to make $13 billion this year. From my calculations, they made about $6.26 billion. It’s wank. And that’s the technical term they use.” (11:10)
• On the Bubble:
  • “Are we really going to carve out half to two-thirds of available global capital for one fucking company that burns a billion fucking dollars every two fucking seconds?” (11:50)
  • “There’s no parallel in history... I’ve never seen a company less deserving of any investment than OpenAI.” (12:10)

Episode Structure & Key Timestamps

• [03:13] — Monologue introduction; deconstructing the NVIDIA-OpenAI “investment”
• [04:20] — Data center PR hype and realities
• [05:05] — Timeline and financial needs of data center construction
• [08:20] — Detailed cost breakdown; GPU and data center cost estimates
• [09:50] — Special purpose entities: “straight Enron shit”
• [11:10] — Revenue claims vs. reality; OpenAI’s actual financials
• [11:50] — Capital availability; math on global capital allocations
• [12:21] — Insight into the scale of the AI investment bubble
• [13:20] — Closing thoughts and preview of next episodes

Tone and Takeaways

Ed Zitron’s monologue is sharp, acerbic, and laced with careful industry analysis and sarcasm. He exposes hype-driven narratives, questions the logic behind multi-billion-dollar bets on “AI,” and calls out both media complicity and industry self-delusion. Ed ultimately warns of an imminent reckoning in tech finance, positioning this NVIDIA-OpenAI deal as the latest — and perhaps most egregious — example of the sector’s bubble dynamics.

For listeners and readers:
This episode provides a brisk, biting, and data-driven takedown of one of the most-hyped tech deals of the decade, challenging both the numbers and the narratives that dominate tech reporting today.