E131: 2024 Fantasy President picks, debt ceiling agreement, Dollar dominance & more

David Sacks said his political influence was being overstated: he was funding both Ron DeSantis and RFK Jr., while he and Chamath planned an RFK event, but he denied being DeSantis’s hidden “architect.” He said one press story rested on a call that never happened with someone he did not know, describing himself instead as a “mid-level donor.”

Chamath had hosted Joe Manchin for dinner at his home the previous week, giving the group direct personal access to a senator they were discussing as a possible presidential entrant. Chamath praised Manchin but refused to speculate on whether he would run if Biden left the race, which was more informative than the hosts’ guesswork.

Friedberg’s debt warning was that the ceiling deal changed the slope only marginally: roughly $2.1 trillion in six-year cuts against about $42 trillion of spending. He argued rising interest costs were already competing with defense and would increasingly squeeze Social Security and Medicare unless both parties addressed the structural budget gap.

The de-dollarization debate produced a more nuanced picture than either side’s rhetoric: foreign holdings of marketable Treasuries had fallen sharply as a share, but the yuan still represented only about 2.2% of Swift payments. China-Brazil agreements pointed toward diversification, while dollar-serviced Belt and Road loans showed how far the dollar remained from displacement.

Nvidia’s Grace Hopper strategy was treated as a potential lock-in play, not merely a faster chip: integrated CPU, GPU, memory and CUDA could make switching hardware painfully expensive. Chamath said the next two quarters were crucial; without credible alternatives from AMD and hyperscalers, Nvidia could effectively run away with the AI hardware layer.

AI model economics were simultaneously collapsing: participants said work costing $2–3 million a year earlier could now cost about $200,000 through software improvements. Narrower domain models and cheaper inference could reduce compute per model, although Friedberg expected falling costs to unlock much more total demand.

Chamath described a ruthless attention-management rule: he estimates he has only four or five useful thinking hours a day and refuses to spend them on remote disaster scenarios. He focuses on probable outcomes and long-tail risks only when they can make money, explaining how he filters macro fear from investable opportunity.