Chamath Palihapitiya is building AI factories to dismantle the SaaS empire Salesforce perfected. He argues that only $1 trillion of the $5 trillion software market pays for licenses - the rest fuels a ‘middleman economy’ of consultants forced to patch rigid platforms. Now, AI collapses the cost of custom software, letting 8090 turn raw business inputs into production-ready code. One partner claims $5 billion in legacy licenses already unbundled.
"Only $1 trillion goes to licensing. $4 trillion is spent on the middleman economy of consultants."
- Chamath Palihapitiya, This Week in Startups
The model is catching fire. 8090 expects $100 million in bookings this year, targeting $500 million next year. But the real twist? Mark Benioff and Salesforce Ventures are leading the $100 million Series A. Benioff is effectively funding the torpedo aimed at his own fleet. The round also pulled in Bill Ackman, David Sacks, and Adam D’Angelo - proof that even the architects of enterprise software see the shift.
Palihapitiya isn’t just rewriting software - he’s redesigning the company. He treats departments as hardware chips with fixed inputs and outputs. Marketing takes money and content, outputs leads. Sales takes leads, outputs revenue. AI agents monitor the interconnects. If the signal degrades, the system shows exactly where it broke. No tribal knowledge. No politics. Just math.
"We’re building a system on a chip. Every function is a chip. Inputs and outputs are fixed."
- Chamath Palihapitiya, All-In with Chamath, Jason, Sacks & Friedberg
While 8090 scales, China closes the AI gap. Jason Calacanis notes GLM 5.2 beats GPT-4 in coding and runs at 85% lower cost. Chinese labs harvest reasoning traces from U.S. APIs, distilling breakthroughs at a fraction of the price. David Sacks adds they’re training on Huawei’s Ascend 910B chips - bypassing U.S. sanctions. The moat is gone.
Meanwhile, Micron’s memory surge hits $42 billion in revenue, up from $9 billion. DRAM and HBM now consume 30-40% of hyperscaler capex. The AI race isn’t just about models - it’s about who controls the memory, the power, and the physical build. Palihapitiya bets that relaxed design constraints - shipping-container data centers, air cooling - will let new players bypass years of infrastructure planning. The factory model scales only if the hardware can keep up.


