Featuring
The besties opened 2023 by putting reputations on the prediction market, from the Republican primary to consumer credit and the first ChatGPT boom. Two exchanges survived the debate filter: Chamath and Sacks handicapping Haley against DeSantis, then Friedberg and Jason fighting over whether OpenAI was a bargain or a copyright bonfire at $29 billion. Hindsight is feeling cheeky: Chamath won the political spread, Friedberg's OpenAI ten-bagger call was conservative, and Jason's legal warning became real without becoming fatal.
Spice rack
Could OpenAI justify a $29 billion valuation despite its copyright and monetization risks?
Original point: OpenAI at roughly $29 billion could still be a ten-bagger because its tools might become infrastructure for a large application ecosystem.
What everyone argued
Chamath Palihapitiya
Chamath argued that general models trained on similar public data could converge, making proprietary data, hand tuning, and distribution the real moats. He also agreed that derivative-work questions created a meaningful legal threshold.
Jason Calacanis
Jason argued that the valuation had outrun fundamentals and that commercializing models trained on unlicensed material would invite crushing copyright litigation. He treated rights clearance and substitution for creators as existential constraints.
David Friedberg
Friedberg argued that OpenAI could become an AWS-like layer for AI applications, earn from the ecosystem built on its tools, and plausibly grow from $29 billion to roughly $300 billion.
Winner circle
Friedberg wins the central valuation question. He correctly identified the application-platform path and even his tenfold upside estimate proved conservative by later private-market marks. Jason deserves credit for spotting a genuine copyright overhang, while Chamath best explained why model leadership alone might not be a moat, but neither rebutted the core claim that OpenAI could create value far beyond $29 billion.
Commentary
Chamath Palihapitiya
Assumptions and fact checks
Models trained with comparable compute on the same data will converge enough that unique data and tuning determine defensibility.
Why it mattersThe framing usefully identifies data and product differentiation, but architectures, training methods, talent, inference systems, and distribution can also create persistent differences.
Copyright treatment of training and derivative outputs would require new legal clarification.
Why it mattersThe Copyright Office later devoted a full report to generative-AI training, and major cases remain contested.
Jason Calacanis
Jason found the strongest missing risk in the bull case, then converted a serious unresolved liability into a certainty of commercial failure. The warning aged better than the conclusion.
Assumptions and fact checks
Copyright exposure would make OpenAI's platform model commercially unworkable.
Why it mattersThe exposure became real litigation, but OpenAI continued selling products, signing partnerships, raising capital, and expanding its platform.
Training on an entire copyrighted work without permission necessarily defeats fair use.
Why it mattersAmount used matters, but no single factor automatically decides fair use. The Copyright Office describes licensing as appropriate in some circumstances while recognizing that outcomes depend on the use, source material, and market effects.
Commercial purpose, amount used, and harm to the market for the original are relevant to fair use.
Check17 U.S.C. 107 requires courts to weigh purpose and character, nature of the work, amount used, and market effect.
David Friedberg
Friedberg saw the platform before the revenue model was obvious, and his ten-bagger remark proved conservative on the private-market mark. He would have made the case sturdier by acknowledging compute, governance, and licensing costs.
Assumptions and fact checks
Developer tools and applications on top of OpenAI would create platform-scale value.
Why it mattersOpenAI's later funding announcement explicitly described APIs, enterprise deployment, consumer distribution, and Codex as a reinforcing platform flywheel.
A $29 billion entry price offered tenfold upside.
Why it mattersThe reported valuation later rose far beyond $290 billion, although private marks do not by themselves prove durable profitability or investor returns.
OpenAI would receive a billion-dollar-plus investment in 2023.
CheckOn January 23, 2023, OpenAI announced a multi-year, multi-billion-dollar Microsoft investment.
OpenAI could become a roughly $300 billion company.
CheckOpenAI announced a $300 billion post-money valuation in March 2025 and later an $852 billion valuation in March 2026.
Was Nikki Haley a better 2024 Republican primary bet than Ron DeSantis?
Original point: Go long Nikki Haley and short Ron DeSantis because DeSantis had peaked too early and Haley could emerge later as a more normal, broadly acceptable alternative.
What everyone argued
Chamath Palihapitiya
Chamath framed the prediction as a relative trade. DeSantis would absorb attacks as the early front-runner, while Haley could emerge from behind as a southern governor with establishment credibility and a more moderate profile.
David Sacks
Sacks argued that DeSantis could unite the establishment and populist wings of the GOP, while Haley lacked meaningful populist support and therefore could not consolidate the party.
Winner circle
Chamath wins the narrowly framed trade. He correctly saw more downside in the early DeSantis boom and identified Haley as the candidate likelier to survive after the field consolidated. Sacks diagnosed Haley's ceiling, but he did not establish that DeSantis could inherit Trump's populist voters, and the primary showed he could not.
Commentary
Chamath Palihapitiya
Chamath won by defining a relative bet instead of pretending Haley would win the nomination. His mechanism was only half right, but the trade paid exactly because DeSantis faded first.
Assumptions and fact checks
DeSantis's early front-runner status created more downside than advantage.
Why it mattersHis support eroded under sustained scrutiny, and he left after Iowa despite entering as Trump's strongest apparent rival.
Republican voters were moving toward a moderate alternative.
Why it mattersHaley did become the last alternative standing, but Trump overwhelmingly controlled the electorate. The relative trade worked without the broader moderation thesis working.
David Sacks
Sacks supplied the strongest objection to Haley but answered the broader nomination question instead of the actual spread trade. That claim mismatch cost him the round.
Assumptions and fact checks
DeSantis was broadly accepted by both establishment and MAGA voters.
Why it mattersThe primary showed that nominal approval was not transferable support. Trump retained the populist electorate, while DeSantis could not build a durable anti-Trump coalition.
Haley's weak populist support made her less viable than DeSantis.
Why it mattersIt prevented Haley from winning, but she still outlasted DeSantis. The assumption answered who could beat Trump, not Chamath's narrower relative trade.
Chamath gave the round its best strategic distinction: a valuable product is not automatically a durable moat. He improved the debate without taking as clean a valuation position as Friedberg or Jason.