Cheeky Pint: Ben Thompson from Stratechery on AI Ads, The End of SaaS, and The Future of Media

Host: John Collison (Stripe)
Guest: Ben Thompson (Stratechery)
Date: February 12, 2026

Overview

In this episode, Stripe cofounder John Collison sits down with Ben Thompson, the founder of Stratechery, for a wide-ranging discussion touching on the shifting landscape of media, the evolving economic models in tech, how AI is reshaping advertising and commerce, the future of SaaS, bundle economics, the threat and opportunity of agentic AI, the ongoing importance of semiconductor manufacturing, and more. The tone is thoughtful, occasionally cheeky, with Ben’s signature big-picture perspective and John’s hands-on industry curiosity.

Key Discussion Points & Insights

Living in Taiwan and Global Perspectives

• Taiwan as a Place to Live and Innovate

  • Ben discusses why Taiwan is convenient and rich in culture, while John notes it's underrated by the SF crowd.
  • The rise of delivery (Uber Eats) has changed local dining dynamics, sometimes closing beloved 'hole-in-the-wall' restaurants.
  • Advice to visitors: Don’t over plan; follow your gut at night markets.
  • [02:25] John: "Don’t try to over plan. Go to a night market and follow your belly."

• City Design and Heritage

  • Taiwan’s blend of commercial and residential, legacy of Japanese colonial planning, and its “ugly outside, palatial inside” vibe.

Aggregation Theory and the AI Era

• Aggregation Dynamics

  • Ben revisits aggregation theory: in the Internet era, power flows to demand aggregators rather than suppliers (ex: Booking.com versus hotel chains).
  • [05:25] Ben: "Booking.com aggregates all the hotels but are also aggregated by Google. So they're Google’s biggest customer even as they’re also on the other side."

• AI’s Disruption of Theoretical Models

  • The impact of AI on aggregation is “TBD”, with potential for new aggregators (like ChatGPT) to develop platform power, but much still unsettled.

The Debate: Ads in AI Products

• Silicon Valley’s Ad Skepticism

  • Ben notes a Valley reflex against advertising.
  • [07:46] Ben: “With Stripe, you’re on the skepticism side. I think ads are amazing... Stratechery has gotten tremendous traction just by not hating ads—even though I’m not an ad model.”

• 効iciency and Democratization

  • Ads as a democratizing monetization method, granting access worldwide (including for those who can’t pay direct fees).
  • Instagram and Meta’s ad targeting—sometimes better than organic search—delivers surprising, useful products.

• AI Advertising Strategy

  • Critique of OpenAI’s bland, context-triggered banner ads: it shrinks their addressable market and can erode user trust in answer integrity.
  • [12:34] Ben: “The ads should be more like Facebook ads than Google Ads. Not on-the-spot prompt-based, but based on a deep understanding of the user.”

• Broader Ad Model Critique

  • Search ads can cannibalize organic value; Facebook/Meta’s model shines by surfacing “things I didn’t know I wanted.”
  • The ideal for AI: build a profile of user interests and anticipate needs—avoiding direct answer-ad entanglement.

• Trust and Privacy Backlash

  • Ben dismisses paranoia over “targeting that’s too good” as overblown and largely a tech-elite concern not shared by most users.

Platforms, Feeds, & Media: The Meta–TikTok Paradigm

• Meta’s Identity Crisis

  • Facebook’s growth plateaued when it thought of itself as a social network, failing to pivot fast enough.
  • TikTok succeeded by switching to pure algorithmic entertainment: “It’s not a social network… it’s personalized TV.”
  • [23:09] Ben: “What actually matters is absolute numbers. Better to have 0.1% of a billion pieces of content be good than 10% of 100.”

• TikTok, China, and Regulation

  • Ben highlights concerns over Chinese influence via algorithms, not data itself. Despite forced U.S. sale, China retained algorithmic control.
  • [26:18] Ben: “It seems insane to have a primary information source controlled by your chief geopolitical adversary.”

• Writer Bundling Dilemmas

  • Discussion of why “bundles” (à la cable TV or Spotify) are optimal but hard to enforce—writers prefer sovereignty; users resist friction.

AI’s Impact on Commerce: “Agentic Commerce” and Beyond

• Agentic Commerce Breakdown

  • Three levels:
    1. Just-in-time UI/automation (autofilling forms, simplifying actions)
    2. Better discovery/search (natural language, parameterized queries for products)
    3. Persistent, AI-powered profiling for recommendations and anticipatory targeting
  • [41:41] Ben: “The largest and most successful agent in the world today is Facebook advertising... You set a price, they find you the user. That’s already agentic.”

• The Promise and Limits

  • Replacing manual search and discovery; possibility of AI anticipating needs (e.g., surfacing a winter coat ad before you search for one).
  • Caution: moving toward “perfect competition” can commoditize everything, eroding margins and some categories’ soul/"humanity".

• Amazon Basics Problem

  • Utilitarian but soulless goods proliferate (“no soul in an AmazonBasics power adapter and that's fine”), but the rising tide lifts all boats: more access, lower prices.

The End of SaaS?

• Market Malaise and Structural Change

  • SaaS multiples are down: reasons include end of headcount-driven growth and the threat of commoditized, in-house cloud code.
  • Still, core “systems of record” are robust (e.g., Workday) and unlikely to be displaced by custom code soon.
  • Future success: small, focused businesses (the “pond” model) rather than one-size-fits-all mega-SaaS.
  • [47:52] Ben: "Focus on your strengths, pay others for the rest—US business excels at this."

• AI’s Role

  • AI may unlock bespoke, tiny “pond” businesses (enabled by low-cost tools/platforms) and challenge the seat-based SaaS revenue model.

The Future of the Premium Newsletter Model

• From Unbundling to Mini-Bundles

  • Ben pioneered the direct-to-reader paid model, learning from earlier (sometimes failed) attempts (ex: Andrew Sullivan’s leaky paywall and burnout).
  • Key: keep costs low and offer “pay for more” rather than “pay or lose.”
  • Big challenge for bundles: successful writers reluctant to join; bundling works best with few strong players (cf. music industry/Spotify).

• AI in Writing

  • Ben mostly uses AI for rapid, deep research and as an advanced Google. He remains highly selective, not using it to generate final content.

Hardware Constraints: The Fab Bottleneck

• TSMC as The Strategic Pinch Point

  • AI progress is chip-bound; nearly all AI chips are fabbed at TSMC, which is conservative about expanding capacity due to high CapEx risk and the perils of overbuilding.
  • [67:01] Ben: "The risks for fabs are larger than for anyone else… If you have too much capacity, your costs are locked in."

• Solutions?

  • Hyperscalers may need to prepay for capacity or foster competition (Intel, Samsung). Market structure (monopoly with TSMC) combines with underinvestment: “We’re heading for a massive shortage in 2029."

Societal Change in an Abundant AI World

• On Education

  • Students should integrate AI into homework; in-person evaluation will rise in importance.
  • Shared experiences (live events, group readings) will be more valuable as content becomes individualized.
  • [74:21] Ben: “Whoever can use AI most effectively in their jobs is going to have a big advantage.”

• On Community and Content

  • Community remains largely unsolved for content—comments and message boards attract poor dynamics, but sharing within private groups can foster better discussion.
  • Audio content (e.g., the Stratechery podcast) now dominates consumption, improving retention but reducing virality and sharing.

• On Crypto’s Relevance

  • Even more valuable as a marker of digital authenticity in a world of infinite AI-generated content.

Big Tech: Execution and Strengths

• Apple

  • Manufacturing champion, but platforms stewardship lags; products lead, software rougher, and corporate culture aging.
  • “Apple gets platforms because they make great products, but are terrible stewards of the platforms.”

• Google

  • Suboptimal, yet resilient; core business is so profitable it can afford inefficiency, tending toward amorphous ‘slime’ that eventually swallows new fields.

• Microsoft

  • Thrives by offering 'good enough', bundled tools, especially for non-expert users—“Best of breed isn’t what most people want.”
  • Remains powerful due to entrenched distribution; SaaS disruption slower than predicted.

• Meta/Facebook

  • Technically strongest operator, especially in ads and scaling; sustained by engagement and cash flow. Potentially more threatened by AI/agents than Google.

• Amazon

  • Pioneered cost optimization strategies (AWS, Graviton, Trainium), but faces question whether the same playbook works when hardware leaps are rapid and competition for chips is intense.

Notable Quotes & Memorable Moments

On Meta’s Core Strengths

"[Meta’s] ad model is underrated. The trick with them is keeping engagement—that’s what makes the whole thing go.”
— Ben Thompson [87:51]

On Ad Models in AI

“The best Instagram ads don’t have anything to do with the stuff I’m surfing. It’s from Meta’s understanding of me broadly.”
— Ben Thompson [12:34]

On TikTok and Information Control

“It seems fairly insane to have a primary information source controlled by your chief geopolitical adversary.”
— Ben Thompson [26:18]

On Bundling and Incentives

“Bundles are good for everyone involved, and no one wants to be a part of it.”
— Ben Thompson [57:10]

On Perfect Competition and Commoditization

“In a world of AI-mediated everything, how many things that can’t get measured, fall by the wayside?”
— Ben Thompson [33:22]

Segment Timestamps

• 00:20–04:48 | Taiwan as a home base and living in Asia
• 05:14–06:03 | Aggregation Theory explained
• 06:03–11:14 | AI, OpenAI, and the politics of digital advertising
• 11:29–17:27 | Critique of ad models in AI; Google, Meta, and user trust
• 19:35–23:53 | Big tech success, feed era, absolute vs. relative content
• 24:01–28:29 | TikTok, China policy, and the limits of U.S. regulation
• 29:57–35:15 | AI as aggregator; agentic commerce scenarios
• 37:34–41:16 | Agentic commerce: automation, better search, persistent profiles
• 45:19–52:18 | SaaS’ future; seat-based pricing vs. structural change
• 53:12–63:09 | The newsletter model, bundling, and Substack
• 63:41–65:59 | AI research in writing and information reliability
• 66:06–73:21 | TSMC bottleneck, chips, and solutions
• 73:52–83:45 | Societal shifts: school, content sharing, crypto, audio
• 83:48–90:03 | Big tech execution—deep dive on Apple, Google, Microsoft, Meta, Amazon

For More

• Stratechery: stratechery.com
• Cheeky Pint Podcast (Selected Episodes)

Summary by PodcastGPT
For those who haven’t listened, this episode is a masterclass in connecting strategy, technology, business models, and societal outcomes—punctuated with sharp, contrarian takes and real-world anecdotes.