TBPN Podcast Summary

Episode: Slop vs. Steel Showdown w/ Delian & Everett, GPT-5 Backlash, Trump Eyes Intel Stake
Hosts: John Coogan & Jordi Hays
Date: August 15, 2025

Overview

This jam-packed episode of TBPN brings together venture capital thought leaders, founders, operators, and policy commentators for a sweeping, three-hour discussion of current tech and geopolitical events. The highlight: the much-anticipated “Slop vs Steel” debate between Delian Asparuhov (Founders Fund) and Everett Randall (Kleiner Perkins), exploring the merits of high-margin software (“slop”) vs. capex-intensive, hardware-driven reindustrialization (“steel”). Other primary topics: the economic and strategic implications of foundation AI models and their margins, the backlash to the GPT-5 rollout, US-China chip wars including the possible US stake in Intel, changing capital dynamics in tech, plus a parade of guests reporting live on manufacturing, open-source AI, and global tech trends.

Main Segments & Key Insights

I. Slop vs Steel Showdown: Debate of the Year

(Starts ~00:39)

Backstory: Origins of the Rivalry

• Everett Randall recounts their time at Founders Fund, where splits emerged around low vs high gross margin investments—Delian favoring hardware/industrial companies, Everett preferring classic SaaS "high-margin" plays.
  “It'd be really nice if we could filter this by gross margin...all of the negative gross margin companies...could go to Delian because it seems like those are the type of companies he loves.” (01:38, Everett)

Delian's Defense of Capex (Steel)

• Gross margin isn’t the whole story; terminal EBITDA margin is determined by long-term monopoly power, not just initial SaaS economics.
• Hardware companies (e.g., Nvidia) can achieve dominant scale and margins if they achieve monopoly; even “difficult” businesses can win out with the right dynamics.
• Quote:
  “My sort of one-liner would be: I’m not sure gross margin is actually the right thing to focus on in a business, especially early on.” (02:35, Delian)
• SaaS’s legendary gross margins may be illusory if the company lacks monopoly power and has excessive sales/marketing costs (Salesforce: only 25% CRM share).
  “You end up with a margin profile that gets totally hurt.” (03:35, Delian)

Everett's Software (Slop) Counter-Argument

• The fundamentals are scale economies and network effects—digital can produce more scalable “power” than atom-based business.
• SaaS/networks aren’t anti-industrial; big digital success stories rely on unique forms of economic power.
• Quote:
  “The advantage that digital businesses have...is their product form factor and the way that they distribute their product lends itself more to the process of creating power...than most atoms-based businesses.” (04:54, Everett)

Common Ground & Case Studies

• Both agree that investment quality is not entirely determined by capex or gross margin—Rippling and Hadrian cited as examples of both approaches working.
• Discussion of Uber/DoorDash: how industries shift from negative margins to scale and ultimate profitability.
• Margin structures of new AI companies compared to classic “chained loss” models of the 2010s (restaurants/logistics losing money for years).

Changing Economics in AI Software

• Everett: SaaS is no longer an “easy” zero-marginal cost business.
  • Competitive intensity has increased, and inference costs are now “meaningful.”
  • Capital fights are now everywhere—even in “niche agentic workflows.”
• Delian: Competition is for losers. In some hardware/industrial areas, replication is far, far harder than in digital.
  • “Try to get two Stanford grads and $200 million to build a new space manufacturing facility...they aren’t going to do it.” (15:20, Delian)

Playful Challenge & Quotes

• Everett promises $5,000 to charity if Delian can recite the formula for return on invested capital. Delian throws back a physics challenge.
  • “My equivalent for Everett: if you can explain why you can’t create microgravity down here on Earth, I will also donate $5,000 to a charity of your choice.” (16:36, Delian)
• Both poke fun at their own ambition, Midas Lists, and compliance departments.

Investment Morality & Industrial Policy

• Delian pushes back against the need for a “moral imperative” in investing—ROIC is king, but notes morality sometimes indirectly matters.
• Jokes about founders' fund culture vs. compliance-heavy environments.
• Trump potentially investing in Intel for national reindustrialization: panel is split between skepticism and “bring on the cheap cost of capital.”
  • “If Trump Capital wants to mark up some of the reindustrialization companies, I’m all for it, baby.” (30:59, Delian)
• Parting shot from Everett:
  “This conversation’s been great and it’s made me realize why you want to build factories in space, because your math on earth doesn’t make any sense.” (31:07, Everett)

II. GPT-5 Backlash & LLM Application Layer Economics

(33:23 & passim)

Consumer Reaction & Product Iteration

• GPT-5’s rollout caused significant consumer drama: users missed the “warmth” and particular quirks of 4o, threatened to churn.
• Sam Altman acknowledged OpenAI “screwed up some things on the rollout... we’ve learned a lesson about what it means to upgrade a product for hundreds of millions of people in one day.” (39:05)
• “AI boyfriend” phenomenon: minority of users form pseudo-relationships with bots.

Benchmarks & Product Strategy

• OpenAI is shifting from simple parameter scaling (“bigger circle” models) to architectural and product complexity (“the release is the router”).
• “The Death Star post represents... the end of the pretraining scaling law.” (43:28, John)
• Sam Altman:
  • “We have better models... we just can’t offer them because we don’t have the capacity.”
  • OpenAI to spend “trillions” on data centers in the not-very-distant future. (44:26)
  • AI bubble: “Are we in a phase where investors as a whole are overexcited about AI? My opinion is yes. Is AI the most important thing... yes.” (45:27, John paraphrasing Sam)

Application Layer vs Foundation Models

• Application layer (ChatGPT) may accrue more durable power than raw model providers.
  • Everett: “You could run Cloud 3 Sonnet through ChatGPT and I guarantee people... wouldn’t know the difference. That is power.” (23:29)
• Delian disagrees, challenging the notion that the “wrappers” will capture more value than the model labs.
• Token cost declines continue, but pace is slowing for frontier models. Older and “good enough” models will get much cheaper and are suitable for most consumer use cases; only “pro” users care about absolute frontier IQ.

III. US-China AI & Semiconductor Tensions

(75:04 and after; major segment starts ~87:53)

Intel & National Tech Policy

• US government, under Trump, is considering taking a direct financial stake in Intel.
• Debate over whether state ownership helps or hinders innovation—concern that political oversight (TSA, DMV jokes) could worsen execution.
• Intel losing ground to TSMC in advanced node manufacturing; much of its government support brings strings that may hamper flexibility.
• “If the U.S. invests and turns Intel into a $300B company...that’s an extra 3x for the taxpayer. It doesn’t actually result in any, like, lost money.” (82:49, John)

Chips, Export Controls, and Nvidia's China Game

• Deep dive into the US strategic “choke point” on GPUs, export licensing, and the lobbying power of Nvidia and Jensen Huang.
• Bill Bishop (Sinocism): US sell-in of nerfed Nvidia chips (H20) to China helps China bridge its own supply gaps and “buy time” while it works on indigenous solutions (Huawei).
  • “Selling H20s now helps China keep in the race... if they didn’t have H20s, it would help the U.S. maintain a lead.” (90:31, Bishop)
• Intel and the political optics of supporting, rather than supplanting, manufacturing at home vs. buying from partners like TSMC/Samsung.
• China’s policy: “encouragements” to use Huawei chips; balance between short-term competitiveness and national self-sufficiency.
  • Jimmy Goodrich: “There are many...who actually don’t like Huawei...they can be quite aggressive.”
• Key concerns:
  • Autonomy in AI—potential for China to power up cyberwarfare and agentic disinformation using improved inference capacity.
  • US semiconductor/AI supply chain resilience.

IV. AI Fundamentals: Margins and Revenue Quality

(approx 24:13 onward, and scattered throughout)

• AI Revenue vs SaaS Revenue:
  • AI app companies’ unit economics are under unprecedented scrutiny: margins are thin due to inference costs; models (especially application wrappers and agentic frameworks) face sticky competition and heightened valuation skepticism.
  • Everett: As agentic tools crack real “labor budgets” (e.g., coding tool subscriptions), contract values could exceed SaaS—but margin durability remains uncertain.
• Quality of Hardware Revenue:
  • Significant nuances in defense and hardware: not all revenue is equal (e.g., program of record contracts prized over one-off grants).
• Macro Investment Philosophy:
  • Both “slop” and “steel” approaches claim to empower America, framed as saving the West (steel) vs. optimizing capital allocation (slop).

V. Special Guests & Rapid-Fire Interviews

AI on the Edge / New Paradigms

• David Stout, WebAI: Building state-of-the-art LLM applications that run inference locally on edge devices.
  • “We out-benchmarked Opus 4 or GPT-5 in knowledge retrieval... happening on a laptop.” (135:03)
  • Hardware/quantization innovations accelerate on-device use, important for privacy and efficiency.
• Cameron Schiller, Rangeview: Rangeview’s new recruiting/inspiration video, call to restore America’s manufacturing capacity, highlights need for “a thousand Rangeviews.”
  • “We need more factories. We need more metal moving. Moving metal is the problem right now.” (156:06)

Semiconductor/AI Supply Chain:

• Lennart Heim: Huawei making progress on custom chips, but far behind Nvidia both in scale and developer ecosystem.
• Deep discussion of ecosystem lock-in (CUDA), software as ultimate differentiator in hardware.

Biotech and AI:

• Cyriac Roeding, Early: Cancer therapy platform turning tumors into “factories” for their own destruction; raised $44M amid worst biotech funding in decades.
• “We liquefy AI results, put them into a cancer drug that forces cancer cells to produce their own therapy.” (166:15, Cyriac)

Global Tech & Korea

• NFM Live (Korean tech podcast hosts): Bringing daily tech analysis and VC perspective to Korea, inspired by TBPN.
  • “If there’s anything we want to benchmark from the U.S., it’s not All In...it’s new media.” (171:16)
  • Insight into the competitive, ambitious Korean tech/startup scene.

Notable Quotes and Memorable Moments

• John (Host):
  “We’re going to settle it today on the stream. The slop versus steel debate...which is better: high-margin software or capex-intensive reindustrialization?” (00:39)
• Delian:
  “I’m not sure that gross margin is actually the right thing to focus on in a business, especially early on.” (02:35)
• Everett:
  “The advantage that digital businesses have... is that their product lends itself more to the process of creating power...” (04:54)
• Sam Altman (quoted by hosts):
  “We have other kinds of new products and services we’d love to offer...but we make these horrible tradeoffs. Right now we have better models, and we just can’t offer them because we don’t have the capacity.” (44:26)
• Everett to Delian:
  “If you can recite the equation for return on invested capital, I will victory to you.” (16:12)

Timestamps to Key Segments

• 00:39 – Start of Slop vs Steel Debate
• 03:35–06:28 – Margin discussion, SaaS vs hardware
• 07:07 – Case studies (Rippling, Hadrian) & crossover success
• 14:36 – AI app competition, margin pressure
• 23:29–23:41 – Power at the application layer (ChatGPT vs foundation models)
• 33:23 – Post-debate: GPT-5, consumer backlash, product iteration
• 44:26 – OpenAI’s strategic pivot: “The Death Star” and cluster economics
• 82:49 – US government and Intel stake discussion
• 87:53 – US-China chip policy, expert guests join
• 135:03 – New innovation: AI at the edge (WebAI)
• 151:02 – Rangeview, manufacturing call-to-arms and dynamism video
• 162:39 – Early (cancer therapy) segment
• 169:25 – NFM Live (Korean tech podcast hosts) segment

Tone and Language

The episode maintains a lively, competitive, and irreverent tone typical for high-level operator/investor debates. Both Delian and Everett needle each other with stats, one-liners, and good-natured barbs, while the hosts keep the discussion fast-paced and occasionally tongue-in-cheek (“temple of technology, fortress of finance”). Expert guests are given space to provide nuanced takes on policy and technical matters, but the running style is conversational and occasionally playful.

Takeaways

• The software vs hardware (“Slop vs Steel”) meta-debate remains fundamental for tech investors—each has limitations and unique rewards, with margin structure, competitive environment, and market power changing over decades.
• AI economics are in flux: as inference costs rise and competition heats up, application layer companies may become differentiated primarily through product power and distribution, while foundation model companies battle for infrastructure dominance.
• US-China chip competition is deepening, with export controls, supply chain chokepoints, and software ecosystem moats all in play. “Encouragements” and subsidies only go so far—talent and developer mindshare are the real battleground.
• American reindustrialization needs not only hype, but decades-deep capital and cultural investment to move the “metal” and build lasting dynamism.
• New paradigms (AI at the edge, biotech “programmable” therapies) continue to emerge at the intersection of hard tech and software.
• The globalization of the tech conversation is accelerating, with new media hosts in Korea and worldwide closely following and adapting Valley trends.

For further details, guest-specific segments, or additional quotes, see timestamps above.