Nvidia's Historic Quarter, SaaS Comeback, Bessent vs Druck, America's Debt Crisis, Cancer Vaccine
Chamath Palihapitiya, David Sacks, David Friedberg, Jason Calacanis
29 Aug 20266 min read1h 30m
TL;DR
Nvidia posted the most profitable quarter of any public company in history — $60B net profit on $96B revenue, guiding 70% growth vs. Wall Street's 45% expectation — proving the AI capex boom has real legs. Salesforce surged 20% as the 'SaaS apocalypse' narrative collapsed: systems of record like CRM are not being replaced by AI but are becoming the canonical data layer that agents depend on. The hosts argue Benioff's willingness to let Claude be the front-end to Salesforce data is the smart play — trading user interface control for deeper entrenchment as AI's source of truth.
Key Moments
David Sacks
“He is integrating Salesforce into Claude. He's willing to allow Claude to be the front end to have the sort of the primary user relationship.”
Sacks explaining Benioff's strategic bet — risking disintermediation by letting Anthropic's Claude own the UX in order to stay relevant as the data layer.
“The simple extrapolation here was AI agents can code therefore all software is going to zero. That was the simplistic extrapolation.”
Sacks arguing that the SaaS apocalypse narrative — like the job-loss narrative — was a flawed linear extrapolation that Salesforce's results now shred.
All-In is a weekly podcast hosted by four close friends and tech investors: Chamath Palihapitiya, David Sacks, David Friedberg, and Jason Calacanis. The show covers tech, finance, politics, and science with candid takes from each host's unique vantage point. Known as the 'besties,' the four bring together perspectives from venture capital, operating, and policy.
Takeaways
1
Letting Claude own your UX is Benioff's masterstroke Salesforce is integrating its full data layer and workflows into Anthropic's Claude, accepting that the AI assistant — not Salesforce's own UI — will be the primary interface for users. Sacks argues this willingness to risk disintermediation is precisely what makes Salesforce more entrenched, not less, because agents need a canonical source of truth to write back to.
2
Nvidia's guidance gap dwarfs the beat itself Nvidia guided to 70% revenue growth next year when Wall Street expected 45% — and management said that number was supply-constrained, not demand-constrained. That delta between guidance and consensus is rarer and more significant than even a blowout quarter, signaling sustained AI capex well into the future.
3
Build for agents, not users — SaaS founders take note Sacks argues the strategic imperative for any SaaS company is now to invest in great APIs and CLI (command-line interfaces) that agents can call, not just slick human UIs. Founders who insist only their own agents can interact with their system will drive customers to more open competitors.
4
Agents will unlock Salesforce's 'trap value' Benioff told Sacks that most Salesforce customers never discover the platform's full feature set — but AI agents will, surfacing capabilities proactively and acting on them once users grant permission. This turns the agent into a power-user consultant, potentially eliminating the expensive implementation and admin layer that has historically slowed adoption.
5
Horizontal SaaS survives; vertical SaaS is the real risk Chamath and Sacks converge on a distinction the market missed: monolithic systems of record like Salesforce, Workday, and SAP hold irreplaceable data gravity, while vertical SaaS products are essentially workflow automations that AI can replicate. The SaaS apocalypse narrative hit Salesforce hardest when it should have been focused on single-vertical niche tools.
6
AI phases: models → harnesses → domain context Chamath maps AI development as three phases: foundation models (the brain), agent harnesses (arms/eyes/memory), and now the emerging third phase of giving agents deep domain context to function as true professionals. Companies sitting on rich contextual data — like CRMs — are the gatekeepers of phase three.
7
Science stagnation stems from funding monoculture Friedberg and Sacks argue that grant funding, tenure, and lecture circuits all punish heterodox thinking, forcing researchers into incrementalism rather than breakthrough science. The parallel to COVID lab-leak suppression and string theory orthodoxy illustrates how the apparatus has structurally lowered the ceiling on discovery.