Trump vs Powell, Solving the Debt Crisis, The $10T AGI Prize, GENIUS Act Becomes Law
Summary
Tariff risk has narrowed into a mostly reciprocal regime, but China and AI hardware remain the live exceptions. Gavin Baker said semiconductor tariffs would make it “hard to win in AI,” while the apparent China deal—rare earths for NVIDIA H20s and AMD MI308Xs—and Vietnam’s special carveout for goods transshipped from China show where enforcement remains concentrated. Trump’s renewed celebration of market highs should provide a “dampening mechanism on tariff volatility.”
Replacing Jerome Powell would not solve the fiscal problem and could damage confidence in the Fed’s independence. The initial reports drove stocks down 1% and long yields up roughly 10 basis points; Baker warned that an actual firing would cause a much larger reaction. Meanwhile, the 30-year Treasury reached 5%, versus a 3.3% average rate on $36 trillion of federal debt—math that could lift annual interest expense from $1.2 trillion toward $2 trillion if debt were refinanced at those rates.
The deficit matters now because falling interest rates no longer conceal it. Baker noted that the 30-year yield has risen since Powell began cutting, illustrating why easier overnight money cannot dictate long-term funding costs. Without slower spending, additional revenue and faster growth through deregulation, interest could soon exceed Medicare and Medicaid, Social Security or defense: “The deficit finally does matter.”
Grok 4 delivered what Baker called the largest model leapfrog in some time, despite being trained on Hopper rather than Blackwell. He cited roughly twice the performance of state-of-the-art Google, OpenAI and Anthropic models on the semi-private ARC-AGI-2 benchmark, plus scores in the 40s on Humanity’s Last Exam versus roughly 5% for exceptional humans. Grok 5, “‘05” from OpenAI and the next Gemini, trained on Blackwell, could bring another step-function gain.
The real AI prize is leverage over problems that currently consume decades or millennia, not merely automating office work. Baker separated economically useful AGI from superintelligence whose returns are “definitionally unknowable”; Friedberg reframed the latter as “leverage on time,” using fusion’s 40- or 50-year research cycle as the example. The upside ranges from routine productivity to longevity, abundance and scientific breakthroughs, though Baker preserved the hedge that AI could “go sideways.”
A plausible consumer-productivity market alone underlies Jason Calacanis’s $1 trillion revenue and $10 trillion value thought experiment. At roughly $75-$100 monthly across one billion developed-world users, annual revenue approaches $1.2 trillion; Baker called the arithmetic reasonable but stressed that Google alone was spending about $70 billion that year. Because producing more tokens can create more intelligence through test-time compute and reinforcement learning, “being the low-cost producer is really going to matter in AI.”
Model quality will not decide the AI platform war without distribution. Baker’s maxim was that “on the internet distribution wins championships,” making Apple and xAI natural potential partners as OpenAI moves into hardware and Google, Meta and Microsoft bring entrenched channels. Safari, Chrome, Perplexity’s Comet browser, ChatGPT’s virtual desktop and Grok companions all point toward the browser or agent as the controlling interface.
The GENIUS Act turns regulated dollar stablecoins into payment infrastructure and a prospective source of Treasury demand, while CLARITY remains the next legislative catalyst. GENIUS became law after passing both chambers; Sacks said the leading offshore stablecoin provider would have to come onshore within three years, while Baker said Tether would need to be 100% in Treasuries to enter the U.S. market. Sacks argued that global adoption could produce “dollarizing from the bottom up”; Calacanis kept pressing on Tether, and Bo Hines added that it would be the fourth-largest Treasury buyer that year.
Deep dive
1. Tariff volatility is receding, except where China and AI intersect
Baker put tariffs below AI as the market’s dominant variable, but left sectoral policy unresolved. His clearest warning was on semiconductors: “It’s going to be hard to win in AI if we put tariffs on semiconductors,” especially given the downsides of tariffs for constructing U.S. data centers.
The broad endgame looked like “reciprocity and reasonable reciprocity” for most countries, with China treated differently. Baker described an apparent exchange of rare earths for NVIDIA H20s and AMD MI308Xs, while Vietnam’s finalized agreement included a special carveout for goods transshipped from China.
Trump’s sensitivity to equities was Baker’s proposed stabilizer. After roughly three months of saying he did not care about stocks, Trump was again citing all-time highs; that market feedback should dampen tariff volatility, though Baker said they would see where things landed in August.
2. Firing Powell would trade institutional credibility for little control
Reports that Trump had drafted Jerome Powell’s termination letter briefly sent equities down 1%, raised bond yields about 10 basis points and pushed Polymarket’s probability of Powell leaving in 2025 to 30% before it fell toward 20%. Powell’s term was due to expire in May of the following year.
Baker did not treat CPI’s move from 2.4% to 2.7% as decisive. Base effects and core or “supercore” measures still suggested contained inflation, while foreign exporters appeared to be absorbing more tariff cost than expected; nevertheless, a weak dollar plus tariffs should theoretically exert upward pressure.
His categorical judgment was that dismissing Powell “would be a mistake.” The world would continue functioning, but the market’s 1% trial-balloon response understated the likely damage, and the Federal Reserve’s independence mattered.
3. The 5% long bond makes the debt burden a current crisis
Friedberg’s load-bearing calculation began with $36 trillion of debt carrying an average 3.3% rate, producing roughly $1.2 trillion of annual interest expense. Refinancing near the 30-year Treasury’s 5% yield—the highest cited since 2007—would push the run rate toward $2 trillion, before adding future deficits.
Lower overnight rates can encourage borrowing, construction, car purchases and business formation, but Friedberg distinguished that stimulus from the price investors demand for lending to the government for 30 years. “Getting rid of a prudent individual may be more challenging than it is beneficial.”
Baker supplied the empirical counterpoint to political pressure: the 30-year yield had risen after Powell began cutting. Deficits once seemed painless because rates kept falling even as debt grew; at the current course and speed, interest could become the government’s largest line item in the not-too-distant future.
There was no silver bullet. Baker’s path combined slower spending, extra revenue, tariffs and faster growth through deregulation. He described consumption taxes as regressive and tariffs as a form of consumption tax that can incentivize domestic manufacturing, while acknowledging Ricardo’s comparative-advantage case for free trade. Closing the deficit should lower rates, creating a mutually reinforcing cycle.
4. A centrist party has room, but politics may misuse Musk’s edge
Baker saw space for a fiscally conservative, reasonably socially liberal party supporting all forms of American energy, including solar. He said targeting a few House and Senate races could have a large impact without requiring a national victory.
His reservation concerned whether an America Party would be the highest and best use of Elon Musk’s talents. An America Party might be a good idea, but Baker argued that AI and Mars could be a better use of Musk’s particular abilities. Sacks pushed back that Musk views politics as existential; Baker answered that Neuralink and the Boring Company are “games of perfection,” while politics is “a game of compromise.”
5. Grok 4 reset the frontier before the Blackwell generation arrived
Baker called Grok 4 the biggest leapfrog in some time. On semi-private ARC-AGI-2—important because some questions were held back and had not been seen by the model—it scored roughly twice as well as state-of-the-art Google, OpenAI and Anthropic models.
On Humanity’s Last Exam, he cited exceptional humans at roughly 5% and Grok 4 in the 40s. His conclusion was that this was “incredible progress,” not a marginal reshuffling of a leaderboard.
The hardware context made the result more consequential. Grok 4 was trained on Hopper, which Baker regarded as approximately the limit of NVIDIA’s prior generation; Grok 5, “‘05” from OpenAI and the next Gemini would be trained on Blackwell and might deliver another substantial step function.
His AGI definition was deliberately practical: an AI taking economically useful actions across many domains and outperforming the average human in most or all of them—drafting a contract, diagnosing a patient or booking travel. None of those tasks requires superintelligence.
6. Superintelligence is better understood as leverage over time
Baker defined superintelligence as smarter than any human and able to access all human knowledge. Its returns are “definitionally unknowable”: if humans have already pushed the limits of physics, biology and chemistry, returns might disappoint; if it cures cancer or invents warp drives, they could be extraordinary.
Friedberg rejected a sharp binary between general and superintelligence in favor of a spectrum of “leverage toward complexity.” His example was fusion: humans have spent roughly 40 or 50 years across science, discovery and engineering, while stronger digital intelligence could provide leverage on problems that otherwise require centuries or millennia.
In Friedberg’s framing, humans are limited less by ability than by time. The resulting leverage could support longer lives, abundant food and resources, easier travel and reclaimed personal time; he did not frame AI as an independently motivated species or race.
Calacanis proposed productivity improving 10% monthly, but Baker called that aggressive. Technology diffuses more slowly than frontier demonstrations imply, although software is already the clearest exception: coding gains are “undeniable,” with startups also reporting meaningful results in customer support and sales.
7. The $10 trillion prize rewards cheap intelligence, not just smart models
Calacanis priced a personal AI subscription at roughly $75-$100 monthly. One billion developed-world users paying about $1,200 annually yields roughly $1 trillion in revenue, supporting his back-of-the-envelope $10 trillion valuation for the “silver medal” of broad economic usefulness alone.
Baker found the revenue logic reasonable but corrected the capex scale: Google alone was spending roughly $70 billion that year, not $10 billion across the industry. Whether that spending repeats for five years depends on realized returns, but companies will keep deploying capital if token economics justify it.
For most of Baker’s 25-year technology-investing career, low-cost production was secondary; AI reverses that. More tokens can translate into more intelligence through test-time compute and post-training reinforcement learning, so generating more tokens from a given $30 billion, $50 billion or $75 billion of capex is a “profound advantage.”
8. Distribution could turn Apple and xAI into natural allies
“The best product doesn’t always win in technology,” Baker cautioned. Grok 4 still faced Google, Meta and Microsoft, and his internet-era rule was simple: “Distribution wins championships.”
Baker proposed Apple and xAI as natural partners. OpenAI’s purchase of Jony Ive’s hardware startup puts it on a path toward competing with Apple; Google has Android and Pixel, Anthropic is closely tied to Amazon, and Meta is in what Baker called a “death match” with Apple.
An Apple-Grok arrangement could give xAI consumer and enterprise distribution while offering Apple a credible alternative amid scrutiny of its lucrative Google search deal. Reports of Apple considering Perplexity or Mistral looked to Baker like “a Band-Aid,” not what Apple needed.
The browser is another distribution layer: Safari and Chrome begin with structural advantages, while Perplexity’s Comet and ChatGPT’s virtual desktop turn browsing into delegated chores. Baker also viewed Grok’s companions as a more engaging route toward an agent interface.
9. GENIUS moved stablecoins from regulatory limbo into law
Bo Hines called GENIUS the foundation for the remaining crypto framework because it updates “archaic” payment rails: “We’re fixing the plumbing of our financial system.” He linked that plumbing to dollar dominance, tokenized public securities and eventual 24/7 markets.
Sacks carefully separated the two bills. GENIUS had passed the Senate and House and was subsequently signed into law; CLARITY, covering market structure for other tokens, had passed the House but still needed Senate hearings and markup, with Tim Scott targeting completion by the end of September and a possible October signing.
Sacks said 12 House members whose votes were needed to advance the bill to that day’s vote were brought into the Oval Office. He said Trump listened to genuine objections, rebutted others and personally negotiated the votes; Hagerty similarly described months of compromise and a 60-vote Senate hurdle that required Democratic support despite some Republican losses.
Hagerty’s first enacted Senate bill broke what he described as a decade-long drought for the Senate Banking Committee. He credited Kirsten Gillibrand’s market expertise for helping bring Democrats along, while Sacks and Hines said completing GENIUS and CLARITY would deliver roughly 90% of the industry’s requested legal framework.
10. Stablecoins can export dollars while absorbing Treasury supply
Sacks said roughly 98% of stablecoins were dollar-backed and 2% euro-backed because users exhibit a “flight to quality.” Once residents of countries with untrusted currencies can transact through phone wallets, he expects portions of those economies to begin “dollarizing from the bottom up.”
Sacks said the leading offshore stablecoin provider would have to come onshore within three years. Baker said Tether would have to be 100% in Treasuries to enter the U.S. market. Sacks separately argued that every dollar token traded globally would correspond to a physical dollar in a U.S. bank account invested in a U.S. Treasury.
Sacks argued that this reserve structure would create demand for U.S. debt, citing studies suggesting trillions of dollars of potential Treasury purchases. Hines added that Tether would rank as the fourth-largest purchaser of U.S. Treasuries that year.
Calacanis called Tether’s past “sordid” and emphasized that he was speaking for himself. Sacks said it was unfair to say Tether did not have its act together and argued that prior U.S. uncertainty had driven innovation offshore. Jason’s broader point was that a legal framework would be preferable to leaving the industry without rules.
11. The bank compromise removes an explicit interest feature but leaves room for rewards
Community banks feared that interest-bearing stablecoins could drain deposits, a concern Sacks considered “wildly overblown.” The legislative concession nevertheless removed an explicit interest feature.
The restriction did not eliminate all possible incentives: marketing, promotions, rebates and other rewards remained available. Sacks floated mileage-style points or gift cards as possible mechanisms, subject to interpretation of the final language.
Hagerty framed the legislation against the central-bank digital currency favored, in his account, by Elizabeth Warren and others. He argued that such a system could centralize visibility and control over transactions, citing reporting thresholds and “Choke Point 3.0,” and said GENIUS put the “final nail in the coffin” of that approach.
12. AI infrastructure broadens the trade beyond Silicon Valley
Sacks described roughly $90 billion of announced Pennsylvania investment around an energy-and-innovation summit organized by Senator Dave McCormick and Dina Powell McCormick. Pennsylvania’s natural gas, Westinghouse nuclear presence and proximity between generation and data centers made it a logical AI hub.
His striking takeaway was the breadth of beneficiaries: Google and other big technology companies, smaller hardware and robotics firms, gas and nuclear producers, construction companies, electricians and carpenters. Calacanis pointed to Google investing in hydro and upgrading dams as one example of AI demand financing existing infrastructure.
Friedberg’s prescription was an all-of-the-above buildout—natural gas, nuclear, solar and batteries—because electrical production is fundamental to AI. He said the United States was already at a disadvantage to China in electricity generation, making energy expansion inseparable from AI competitiveness.
13. Selling H20s can preserve U.S. standards without exporting the frontier
Baker supported reversing the H20 restriction because the chip was neither current Blackwell nor full Hopper, but a less powerful, deprecated variant “many, many years behind” the American frontier. Yet he estimated it remained about two years ahead of Huawei, making the policy “devilishly clever.”
China’s electricity advantage changes hardware design constraints. Huawei’s CloudMatrix 384 links chips with fiber optics rather than copper and is far less power-efficient than NVIDIA’s Blackwell NVL72, but China can spend power where U.S. systems must optimize it more carefully.
Baker’s strategic conclusion was that permitting the H20 gives America an AI advantage while reducing China’s incentive to develop a domestic NVIDIA alternative that could later challenge NVIDIA, AMD and other American accelerator companies globally. The argument was for selling a deprecated, less powerful tier rather than exporting Blackwell or the latest capability.