Nobody made money owning the internet, everybody made money building on top of it (Save this). That's @friedberg argument because Netscape built the first browser and server software, charged for both, went public, then got wiped out once open source alternatives showed up. Mozilla built Firefox for free, Google built Chrome for free, Apache gave away server software so anyone could get online without paying a toll. The value didn't vanish, it just moved off the gatekeepers and onto everyone building on top, Google, eBay, Etsy, Amazon, and millions of small businesses that never could have existed otherwise. Friedberg's claim is that AI is following the same script. If open models keep closing the gap with proprietary ones, value stops concentrating in two or three labs and spreads across every company building on top of it and the data backs this up. Stanford's 2026 AI Index found the gap between the best closed model and the best open model has shrunk to about 3.3%, down from double digits two years ago. Chinese open models now account for 41% of downloads on Hugging Face over the past year and more than 30% of Fortune 500 companies have verified accounts pulling models straight from that ecosystem. Enterprise sentiment backs it too, 41% of organizations say they'd drop proprietary models entirely once open ones hit parity, and another 41% are expanding open source usage regardless. The market already got a preview of this snapping into place. DeepSeek's open weight model release in January 2025 wiped close to $600 billion off Nvidia's market cap in a single day. That's what Chamath means by value capture evaporating in months, not decades. A sector with what looked like permanent pricing power turned out to have a shelf life measured in product cycles. Where this gets uncomfortable is capex and if enterprises can get most of what they need from a free open model on their own hardware, the case for endlessly compounding training spend weakens. Analysts have flagged this as a risk that isn't priced into current capex forecasts, precisely because open models keep closing the gap faster than commercial labs are widening it.
