Biggest LBO Ever, SPAC 2.0, Open Source AI Models, State AI Regulation Frenzy
Speakers
Chamath Palihapitiya
— CEO, Social Capital
David Friedberg
— CEO, The Production Board
Jason Calacanis
— Angel Investor / Founder, LAUNCH
David Sacks
— General Partner, Craft Ventures
Summary
The besties discuss the $55B take-private of Electronic Arts by Saudi Arabia's PIF, Silver Lake and Affinity Partners - the largest LBO ever - and what it says about gaming as the anchor of digital entertainment and about a private-equity industry that Chamath believes is headed for zero returns, with private credit as the next bubble. Chamath explains his institution-only SPAC 2.0 structure and warns retail to avoid SPACs, while Friedberg sees a public-market opportunity in mature companies that can execute an AI transformation. After Sacks joins, the group covers OpenAI's Sora and Meta's Vibes video apps, China's lead in open-source models (DeepSeek 3.2, Kimi K2, Qwen), the electricity-cost problem created by AI data centers, and the frenzy of state-level AI regulation versus a single federal standard.
- EA is being taken private for $55B at $210 a share by PIF, Silver Lake and Affinity Partners; Chamath calls it a killer deal and a multi-hundred-billion-dollar asset if EA can build distribution outside Xbox and PlayStation.
- Friedberg argues AI will accrue most to video games and that entertainment demand grows as AI frees up people's time; PIF is concentrating its bets in gaming.
- Chamath says private equity is 'totally screwed' as capital floods in and distributions dry up, and that private credit is the next bubble building.
- Chamath's SPAC 2.0 (American Exceptionalism) is built for institutions with performance-gated sponsor pay and no founder warrants; both Chamath and Jason tell retail investors to avoid SPACs.
- Friedberg sees a chance to beat the market by picking mature public companies that can execute AI transformation; Chamath says PE portfolio companies lack the people to do it and only owner-operators or motivated public CEOs will.
- DeepSeek 3.2-Exp cuts API costs about 50%; Chamath's 8090 moved workloads to Groq-hosted Kimi K2 because open-source models are far cheaper than OpenAI, Anthropic and Amazon Bedrock; Sacks says open source is the one AI area where the US trails China.
- Chamath warns electricity rates could double within five years from data-center demand; Sacks sees natural gas as the near-term power answer, nuclear longer term, and grid peak-shaving as the bridge; Jason is buying TAO for distributed AI compute.
- Sacks and Friedberg criticize the state AI regulation frenzy (California SB53, Colorado SB24-205) and call for federal preemption; Jason is torn on states' rights.
Ideas
EA take-private is a killer deal
Chamath really likes the $55B take-private of EA at $210 a share by Saudi Arabia's PIF, Silver Lake and Affinity Partners. Bull case: video games are the anchor pillar of internet usage (about 3 billion daily players, per Unity's CEO), EA is the 800-lb gorilla of gaming IP, and going private lets it take its time to clean up the opex model, use next-gen AI tool chains and build distribution outside the Xbox/PlayStation gatekeepers, whose grip is eroding (Xbox's 50% subscription price hike caused so many cancellations that the site went down). Done right, EA is a multi-hundred-billion-dollar asset; the public market was getting it totally wrong, and JPMorgan underwrote the $20B of debt the same day. The bear case - AI tool chains increasing the number of games by orders of magnitude and social platforms distributing them - is low probability, so Jared and Egon did a killer deal.
AI IP erosion: gaming wins, studios lose
Extending his recurring view that the value of patents, IP and copyright will erode as AI tool chains make content cheap to create, Chamath sees a spectrum in which some content-IP holders lose and others win: gaming is on the winning side, while traditional content studios - the Disneys, the Hulus, the Netflixes - are on the losing side.
AI IP erosion: gaming wins, studios lose
Extending his recurring view that the value of patents, IP and copyright will erode as AI tool chains make content cheap to create, Chamath sees a spectrum in which some content-IP holders lose and others win: gaming is on the winning side, while traditional content studios - the Disneys, the Hulus, the Netflixes - are on the losing side.
AI accrues most to video games
AI will accrue far more to video-game entertainment than to social media or traditional media, because games can deliver dynamic, back-and-forth experiences: AI-driven players already lift engagement and retention (Fortnite tunes AI opponents so new players stop churning). The broader macro bet everyone should consider: if AI raises productivity and is deflationary, people in the industrialized world will have more free time, so the entertainment market grows - gaming is the future of entertainment and the future of gaming is AI. PIF's aggressive gaming bets (Savvy Games, Scopely, Niantic, stakes in Nintendo and Take-Two, now majority ownership of EA) reflect the same 10-year thesis.
Private equity returns are going to zero
Private equity in general is 'totally screwed' (this deal, Silver Lake and Affinity excepted). Zero rates gave PE near-infinite cheap borrowing to manufacture returns faster than venture or hedge funds, capital poured in from 60/40 portfolios, and now a flood of laggards overpays for and under-manages assets; as with venture and hedge funds, when an asset class hockey-sticks in size the returns go to zero. Distributions (DPI) have been few and far between for four or five years, and PE portfolios are mostly B and C companies run by C and D people who cannot execute an AI transformation (8090 gets almost no revenue from PE firms despite trying hard to sell into them). Money will come out of PE and concentrate in the few firms that know what they are doing.
Private credit is the next bubble
The money leaving private equity has already leaked into private credit - lending to businesses - which is growing on the same hockey-stick chart as PE and is the next big bubble that is building.
Don't buy SPACs beyond 1% allocation
SPACs such as Desktop Metal, Opendoor, Virgin Galactic, Joby, SoFi and MP Materials are venture-stage investments: 80% of venture goes to zero and 20% pays for the rest, but retail investors treated them like Netflix or Nvidia without understanding the stage of the companies or the venture-style portfolio math. Don't buy SPACs unless they are less than 1% of your portfolio.
Retail should avoid SPACs, including his
Retail investors should avoid SPACs - 'maybe not all SPACs, but definitely my SPAC, just avoid it... do not invest in these things.' His SPAC 2.0 vehicle (American Exceptionalism, 'Raptor 2') was deliberately built for institutions: 98.7% of the capital came from blue-chip institutional investors, there are no founder warrants, and the sponsor earns nothing unless the stock is up 50% and then 75%. It fits an institutional portfolio construction with a very different risk model; retail should not be that far out on the risk curve because you cannot predict where these things go.
Watch Apple's on-device AI push
Apple, though furthest behind in AI, has a really interesting open-source efficient language model (OpenELM) worth keeping an eye on, and its plan is to run LLMs locally on personal devices (M4 Mac minis, phones) with a distributed, SETI-at-home-style compute layer. People will not necessarily want their AI jobs to go to the cloud, and Apple's silicon focus is aimed at that on-device future.
AI power: gas now, nuclear later
AI will create a huge need for power over the next 5-10 years. On a 5-10 year horizon the answer is probably nuclear, or at least nuclear is a big part of it, but nuclear takes at least five years; within the next five years the answer is natural gas, although gas turbines have a two-to-three-year backlog. The bridge is squeezing an extra ~80 GW out of the existing grid by shedding roughly 40 peak hours a year to backup generators and diesel, then adding a lot more gas and eventually nuclear.
Buying TAO for distributed AI compute
Jason is a partner in the emerging crypto fund Still Core Cap, which is buying TAO and studying Bittensor's subnets built for distributed AI compute - a SETI-at-home-style network with an incentive layer. He believes decentralized and local compute will be a big part of AI because people will not necessarily want their AI jobs to run in the cloud.
Up Next
This All-In Podcast video, published October 03, 2025, features Chamath Palihapitiya, David Friedberg, Jason Calacanis, David Sacks discussing EA, ESPO, DIS, NFLX, PSP, BIZD, SPACS, AEXA, AAPL, UNG, URA, TAO. 11 trade ideas extracted by AI with direction and confidence scoring.
Speakers: Chamath Palihapitiya, David Friedberg, Jason Calacanis, David Sacks · Tickers: EA, ESPO, DIS, NFLX, PSP, BIZD, SPACS, AEXA, AAPL, UNG, URA, TAO