Trump vs Powell, Solving the Debt Crisis, The $10T AGI Prize, GENIUS Act Becomes Law
Speakers
GavinSBaker
— Portfolio Manager, Atreides Management
David Friedberg
— CEO, The Production Board
Jason Calacanis
— Angel Investor / Founder, LAUNCH
David Sacks
— General Partner, Craft Ventures
Bo Hines
— Executive Director, President's Council of Advisers on Digital Assets
Bill Hagerty
— Senator
Summary
All-In hosts Jason Calacanis and David Friedberg are joined by investor Gavin Baker to work through tariffs, the trial balloon about firing Fed Chair Jerome Powell, and why a 5 percent 30-year Treasury yield makes the US deficit finally matter. The conversation then turns to the roughly $10 trillion AGI prize, Grok 4's benchmark leap, hyperscaler capex and Baker's argument that low-cost token production and distribution will decide the AI race. David Sacks and Bo Hines join from the White House to explain the GENIUS Act becoming law and the CLARITY market-structure bill, and Sacks separately interviews Senator Bill Hagerty about the stablecoin legislation. The episode closes on the Pittsburgh energy summit, about $90 billion of announced AI infrastructure investment, and the lifting of Nvidia H20 export restrictions to China.
- Gavin Baker sees tariffs settling into reciprocity outside China and says AI now overwhelms every other market driver.
- Baker warns stocks would fall far more than the 1 percent trial-balloon move if Powell were actually fired.
- David Friedberg frames the real crisis as the 5 percent 30-year yield and a potential $2 trillion annual interest bill on $36 trillion of debt.
- Baker adds that the 30-year rose even as the Fed cut, so the deficit finally matters and needs spending cuts, revenue and growth.
- Jason Calacanis sizes a $10 trillion AGI prize from roughly $100 per user per month of AI spending.
- Baker argues distribution beats the best product and sees industrial logic in an xAI-Apple partnership that would also help Apple and Google on antitrust.
- David Sacks and Bo Hines detail the GENIUS Act becoming law, the CLARITY Act moving to the Senate, and dollar stablecoins creating new Treasury demand.
- The Pittsburgh summit announced about $90 billion for AI and energy, and Baker calls the H20 export reversal good for Nvidia and AMD against Huawei.
Ideas
Trump put caps tariff-driven equity volatility
Baker worries that the market fell only about 1 percent on the reports that Trump had drafted Jerome Powell's termination letter. He believes stocks would fall considerably more if Powell were actually fired, that there are very sound reasons for Federal Reserve independence, and that Trump should not read the shallow 1 percent reaction from that trial balloon as evidence the move would be cheap. He calls it a mistake he hopes does not happen, which makes the Fed-chair headline risk a live downside setup for US equities.
AI overwhelms every other market driver
Baker says the single most important thing in the market by far is AI and that it overwhelms everything else, including tariffs and the usual macro debates. He expects the hyperscaler capex cycle to continue for as long as the economic returns are there, noting Google alone is spending about $70 billion this year, and he accepts that roughly $1 trillion a year of AI revenue is a reasonable framing. But given the market capitalizations of the companies involved, he says the prize needs to be at least $10 trillion for current spending and valuations to be justified, which makes the size of the eventual AI payoff the key variable for the whole market.
Long-end yields, not Powell, are the crisis
Friedberg argues that firing Powell does nothing for the actual US fiscal problem, which sits on the long end of the Treasury curve. The 30-year is at exactly 5 percent, the highest borrowing cost since 2007, while the average rate across $36 trillion of outstanding debt is only 3.3 percent, producing about $1.2 trillion of annual interest expense. Refinancing toward 5 percent takes interest expense toward $2 trillion a year, and the balance keeps growing because the government still runs a deficit. The Fed can only cut the short end to stimulate activity; the long end is the market telling the US that spending and taxation have to be addressed first.
Deficit finally matters because rates stay high
Baker backs Friedberg with the empirical point that the 30-year yield has gone up since the Fed started cutting rates. The deficit has been a talking point since Ross Perot in 1992 but never mattered while rates fell, because interest expense stayed a small share of the budget. Now that rates are higher and do not look like they are coming down soon, interest expense is on a path within a few years to exceed Medicare and Medicaid, Social Security or the military as the largest line item. He sees a way out through slower government spending, extra revenue (tariffs acting as the consumption tax America never had, since income-tax receipts have never gone much above 18-19 percent of GDP) and faster growth through deregulation, which would start a virtuous cycle of a smaller deficit and lower rates.
AI is a ten trillion dollar prize
Calacanis sizes the AGI prize at roughly $10 trillion without needing superintelligence. He argues a knowledge worker in the developed world should consider about $75 to $100 a month, or roughly $1,200 a year, a no-brainer spend on AI; with about a billion people in the developed world that is about $1 trillion of annual revenue, which at typical multiples is roughly $10 trillion of market capitalization. He anchors the number to the PC era, where a $3,000 to $4,000 machine lasting three years also cost about $1,000 a year per person, and treats this productivity case as the silver medal that already justifies today's data-center spending.
Apple and xAI are natural AI partners
Baker argues the best product does not always win in technology: on the internet distribution wins championships, and Grok 4 faces competitors with far stronger distribution in Google, Microsoft and Meta. The industrial logic therefore points to an xAI-Apple partnership. OpenAI's purchase of Jony Ive's hardware startup puts it in direct competition with Apple, the rumored Perplexity or Mistral acquisitions are band-aids that do not get Apple what it needs, and Apple already has a decade-long, multi-billion-dollar precedent in its Google search deal. A credible independent AI provider would also help both Apple and Google in the DOJ antitrust remedy phase, and would give Grok enterprise credibility, so he sees solid logic for the deal for both sides.
Stablecoins entrench dollar and Treasury demand
Sacks argues dollar-backed stablecoins entrench the dollar rather than threaten it. Under the GENIUS Act every issued token must be fully reserved, so each dollar token circulating on a wallet anywhere in the world corresponds to a physical dollar in a US bank account invested in a Treasury, and studies cited suggest the bill could create trillions of dollars of new demand for US debt. Stablecoin market share is already about 98 percent dollar versus 2 percent euro because there is a flight to quality and no reason to hold multiple fiat stablecoins. In countries where the local fiat is not trusted, people and merchants can now transact in dollars through phone wallets, dollarizing those economies from the bottom up and extending reserve-currency status into the online realm against BRICS-style challengers.
AI data centers follow gas and nuclear
Sacks argues that energy dominance produces AI dominance. The Pennsylvania energy and innovation summit announced roughly $90 billion of new investment because the state is the second-largest energy producer in the US, with abundant natural gas, room to expand fracking and existing nuclear capability through Westinghouse, and because it makes sense to site large AI data centers next to the power source. He adds that the beneficiaries of the AI boom are much broader than big tech: hardware and robotics firms, gas and nuclear companies, trade associations, construction, electricians and carpenters all get growth from it, so this is not just a Silicon Valley trade.
AI needs all US electricity generation
Baker says electrical production is fundamental to AI and that the US is already at a disadvantage to China on electricity generation, a gap he wants closed as quickly as possible. He is deliberately agnostic on the source: natural gas is great, nuclear is great, solar is great, batteries are great, and he says we need it all. The implication is that the AI buildout is a demand driver for the entire domestic power-generation complex rather than for one favored technology.
H20 sales protect Nvidia and AMD
Baker says the H20 is a deprecated version of the last-generation Hopper chip and many years behind the American state of the art, but it is roughly two years ahead of what Huawei can build, which makes allowing its export devilishly clever. China has more electricity than the US and can afford power-inefficient designs such as Huawei's CloudMatrix 384, which links chips with fiber optics instead of copper and does not need Blackwell-level power efficiency. Selling H20s keeps Chinese demand on American silicon and slows the emergence of a domestic Huawei alternative that could eventually challenge Nvidia, AMD and other US AI-accelerator champions globally. He thought the ban was a mistake and is happy the restriction is being reversed.
Up Next
This All-In Podcast video, published July 19, 2025, features GavinSBaker, David Friedberg, Jason Calacanis, David Sacks discussing SPY, AI-SECTOR, TLT, GOOGL, AAPL, UUP, SHV, UNG, URA, TAN, BATTERIES, NVDA, AMD. 10 trade ideas extracted by AI with direction and confidence scoring.
Speakers: GavinSBaker, David Friedberg, Jason Calacanis, David Sacks · Tickers: SPY, AI-SECTOR, TLT, GOOGL, AAPL, UUP, SHV, UNG, URA, TAN, BATTERIES, NVDA, AMD