At Long Last… (This Week in Stratechery)
1. Key Themes
Theme 1: Frontier AI lab economics are finally visible, and the picture is mixed
Anthropic's S-1 ends four years of speculation
The S-1 gives the first hard numbers on frontier lab economics, though this newsletter only previews the analysis.
"Monday morning brought us Anthropic's S-1, and not a moment too soon. After four years of arguments about the economics at frontier labs, we now have some real numbers to sink our teeth into."
Profitable core business, but competition and dilution are the threats
The headline takeaway from Ben's analysis is that profitability today doesn't guarantee durable returns for investors.
"Anthropic's IPO is finally here; the core business is very profitable, but competition may eat all of the profits, even as investors are set up for endless dilution."
Compute cost is the central question
The episode and article titles point to unresolved questions about who ultimately funds frontier compute.
"Anthropic S-1, The Cost of Compute, Who Is Paying?"
"Anthropic Opens the Kimono, Bubble Questions and Frontier Costs"
Theme 2: Meta's compute spending is a bet on Muse
The investment case for Meta is now tied to Muse
Ben frames Meta's capex as logically inseparable from belief in its frontier model.
"Meta's earnings come in the context of Zuckerberg delivering on his promise; if investors believe in Muse they must support Meta's spending on compute."
The SEC filing context is also notable: the discussion covers "why the SEC forced Anthropic to help us understand Meta's Muse bet," suggesting Anthropic's disclosures offer a read-through on what frontier compute costs.
Theme 3: AI pressure is breaking the global Internet, and localization may be an improvement
A two-part prediction: rupture, then something better
"First, the pressure applied by AI will eventually (or imminently?) rupture the global internet that we've come to know for the past 25 years. But second, as society moves to more localized solutions and products, that may in fact be a welcome development for all of us who have come to accept degraded experiences throughout our technological lives."
The old open Internet was already dying
"The old Internet, where anyone could host a server is over, but it's just the final blow in an ending that has been coming for years. The future, though, is brighter than it appears."
The trigger was "a handful of ominous cybersecurity incidents and... continued upheaval."
Theme 4: China's political risk is rising
Anti-corruption campaign is producing a climate of fear
"an extended look at the climate of fear across the Chinese bureaucracy as Xi Jinping continues an anti-corruption crackdown that was absolutely necessary but may be going further than Xi's earlier proponents ever expected."
Capital markets signal: the STAR 50 AI bubble
The Sharp China episode title includes "The STAR 50 AI Bubble Has Burst," a signal on Chinese AI equity sentiment (no detail provided here).
2. Contrarian Perspectives
1. The death of the global Internet is a feature, not a bug. Most commentary on cybersecurity incidents and AI disruption is apocalyptic. This piece argues the opposite for end users.
"that may in fact be a welcome development for all of us who have come to accept degraded experiences throughout our technological lives."
Supporting evidence: Andrew Sharp's personal behavior shift, "an avid group chatter who's more or less abandoned public social media," suggests demand already favors localized, private products.
2. A highly profitable AI lab can still be a poor investment. Conventional wisdom treats profitability as validation of the frontier lab thesis. Ben's framing separates product profitability from shareholder returns.
"the core business is very profitable, but competition may eat all of the profits, even as investors are set up for endless dilution."
3. "Chinese century" talk is overblown. Andrew argues that a bizarre debate over Mao's governing history, sparked by a soccer star, reveals structural dysfunction.
"connipitions like this one, anchored in a bizarre debate over Mao's governing history, are evidence that rumors of a Chinese century have been greatly exaggerated."
"another reminder that the CCP cannot get out its own way."
3. Companies Identified
Anthropic
- Description: Frontier AI lab, now public via S-1 filing
- Why mentioned: Its S-1 provides the first real economics of a frontier lab
- Quotes: "Anthropic's IPO is finally here; the core business is very profitable, but competition may eat all of the profits, even as investors are set up for endless dilution."
Meta
- Description: Social media/AI company; developer of the Muse model
- Why mentioned: Earnings analysis tied to Muse and compute spend
- Quotes: "if investors believe in Muse they must support Meta's spending on compute."
- Description: Data/AI software company
- Why mentioned: CTO Shyam Sankar interviewed on AI diffusion, safety, and forward-deployed engineers
- Quotes: "An interview with Palantir CTO Shyam Sankar about AI diffusion and safety, forward-deployed engineer, and what the battle in tech is actually about."
- Description: Payments infrastructure (sponsor of the issue)
- Why mentioned: Presented as foundational infrastructure for Internet-native businesses
- Quotes: "Stripe provides the foundation for Internet-native companies — including Stratechery." Also: "No business succeeds because of Stripe... But it does make more things practical to try."
- Description: Newsletter publishing platform
- Why mentioned: Example of integrating a fragmented stack, with Stripe underneath
- Quotes: "building Stratechery was complicated and involved gluing together disparate pieces; Substack made it easy by delivering an integrated offering."
- Description: Consumer hardware/software company
- Why mentioned: Subject of the weekly video, "Do You Trust Apple?"; Dithering also covered "Touchscreen MacBook Pros"
- Quotes: "This week's Stratechery video is on Do You Trust Apple?"
Starlink
- Description: SpaceX satellite internet service
- Why mentioned: Q&A topic on Sharp Tech
- Quotes: "Q&A on Meta, Starlink, Brewers"
4. People Identified
- Description: Stratechery author and analyst
- Why mentioned: Authored the Anthropic S-1 and Internet articles
- Quotes: "Ben did just that with his Update on Tuesday."
Andrew Sharp
- Description: Stratechery writer and podcast host
- Why mentioned: Wrote this overview and the China analysis
- Quotes: "I make the point that connipitions like this one... are evidence that rumors of a Chinese century have been greatly exaggerated."
Shyam Sankar
- Description: CTO of Palantir
- Why mentioned: Interviewed on the "AI religious war"
- Quotes: "what the battle in tech is actually about."
- Description: Chinese leader
- Why mentioned: Leading an anti-corruption crackdown creating bureaucratic fear
- Quotes: "an anti-corruption crackdown that was absolutely necessary but may be going further than Xi's earlier proponents ever expected."
- Description: Sinocism author, Sharp China co-host
- Why mentioned: Co-hosted the China episode on bureaucratic fear and Chinese AI equities
- Quotes: "Sharp China with Andrew Sharp and Sinocism's Bill Bishop"
Kylian Mbappé
- Description: International soccer star
- Why mentioned: Center of a China political controversy over Mao
- Quotes: "a glorious mess surrounding Mao Zedong and international soccer star Kylian Mbappé."
John Gruber
- Description: Daring Fireball author, Dithering co-host
- Why mentioned: Discussed "The End of the Internet" with Ben
- Quotes: "Dithering with Ben Thompson and Daring Fireball's John Gruber"
Jon Yu
- Description: Asianometry creator
- Why mentioned: Covers semiconductor/computing history
- Quotes: "How Linux (and Windows) Ate UNIX"
5. Operating Insights
1. Usage-based billing matters most when usage drives your costs. AI companies in particular should price on consumption rather than flat rates.
"AI companies can charge for what customers actually use, which is particularly important when usage also drives costs, without having to build the systems to track and bill for it themselves."
2. Buy the commodity layer; build the differentiation. Infrastructure should lower the cost of experimentation, not substitute for product judgment.
"Stripe doesn't answer the hard questions about what to build or whether anyone will actually use it. But it does make more things practical to try."
3. Integrated offerings beat assembled stacks for creators and small operators.
"building Stratechery was complicated and involved gluing together disparate pieces; Substack made it easy by delivering an integrated offering."
6. Overlooked Insights
1. Forward-deployed engineers as an AI diffusion mechanism. The Palantir interview flags this model as relevant to how AI actually spreads into enterprises, an angle largely absent from mainstream AI discourse.
"An interview with Palantir CTO Shyam Sankar about AI diffusion and safety, forward-deployed engineer, and what the battle in tech is actually about."
2. Fintech can now build entire business models without owning the underlying rails. Card-issuing and processing infrastructure is abstracted away, lowering barriers to launching financial products.
"Fintechs can build entire business models around card programs without ever having to build the issuing and processing infrastructure underneath it."