Bernie Sanders Says Stop All AI, China's Breakthrough, Inflation Down, Golden Age in 2026?
Summary
Sacks argued that Bernie Sanders’s proposed data-center moratorium would stop US development without stopping China’s AI progress, creating “the biggest own goal ever.” Sanders’s case combined job displacement, billionaire power, child welfare, and a demand that “democracy can catch up”; the group acknowledged risks around Character.AI, “spicy chats,” and parasocial relationships while distinguishing them from useful research tools. Sacks’s investor-level conclusion was categorical: AI carries economic and national-security consequences, and surrendering leadership would make America poorer.
The industry’s immediate political constraint is social license, even as capital is available and value capture remains unresolved. Friedberg noted that enormous infrastructure investment has so far produced obvious value concentration mainly in Nvidia, while Chamath said circular financing, rising stocks, energy fears, and job-loss headlines leave non-shareholders expecting a “sword of Damocles.” His prescription was Gilded Age-style public benefit—Carnegie’s 2,500 libraries, Rockefeller’s institutions, Ford’s wages, and Bell Labs—funded from corporate balance sheets that “Wall Street values at zero.”
Sacks argued that current labor data contradict economy-wide AI job destruction, but the hosts did not deny that specific occupations may disappear. He cited Vanguard figures showing AI-exposed occupations with 1.7% job growth versus 0.8% elsewhere and 3.8% wage growth versus 0.7%, plus Yale Budget Lab’s finding of “no discernible disruption” during 33 months after ChatGPT. Jason countered with Uber reducing driver recruitment where Waymo operates; Sacks conceded driver losses while arguing productivity and newly created work may still produce net gains.
Sacks sees disinflation, private hiring, tax cuts, and AI CapEx producing a “gangbusters 2026,” while Jason sees a politically dangerous gap between rate-of-change data and household price levels. CPI was 2.7% against 3.1% expected, core CPI 2.6%, and three-month core inflation 1.6%; private employment rose 121,000 from September to November while government employment fell 162,000. Yet unemployment reached 4.6%, prices remained elevated, and Jason argued that the bottom half—often without equities—does not experience claims of a “golden age.”
China’s reported EUV prototype matters less as a single copied machine than as evidence that the chip-control clock is accelerating. Reuters said the system had not produced working chips, with China targeting 2028 and some sources saying 2030; Friedberg instead emphasized a roughly $48 billion Phase Three semiconductor fund and Tsinghua research using AI to compensate for imperfect optics. Sacks’s warning was that China is “not just in a catch-up race, they’re in a primacy race.”
Even before indigenous EUV, Huawei’s DUV-based chips and networking have weakened the intended force of export controls. Sacks said China pushed DUV from an expected 14-nanometer limit to 7 and then 5 nanometers, while CloudMatrix 384 links 384 Ascend chips to approach an Nvidia rack’s performance at much higher power. Chamath added that inference silicon may become memory-centric and SRAM-heavy, shifting advantage away from the smallest nodes toward compilers and software, where he believes the US remains “light-years ahead.”
California’s proposed billionaire tax is already affecting relocation decisions even if it never becomes law, according to the group. They estimated that the threat could redirect $100 billion to $200 billion over five to ten years, shrinking the revenue base and intensifying demands for further taxation; Jason separately claimed the state pension system is underfunded by roughly $1 trillion. The tradeable implication is jurisdictional: founders, investors, and taxable balance sheets are treating property-rights uncertainty as sufficient reason to relocate.
Deep dive
1. Sanders’s moratorium would stop America, not China
Bernie Sanders’s argument, as Jason summarized it, was that billionaires seek greater wealth and power, leading technologists predict mass unemployment, AI reduces children’s social interaction, and Congress should halt new data centers so “democracy can catch up.” Jason called the remedy absurd while acknowledging that the underlying employment, energy, and child-welfare concerns were not invented.
The group separated educational chatbots from social-media addiction. Sacks said his children described AI as useful—“It’s like Google. I couldn’t do school without it”—while Jason located the sharper danger in Character.AI, “spicy chats,” and parasocial relationships rather than research and tutoring.
Sacks’s core logic: America can restrict domestic progress but cannot stop China from advancing these technologies. A moratorium would therefore be “the biggest own goal ever,” converting US leadership into Chinese leadership while making Americans poorer and less secure.
His political read was equally blunt: states’ rights and affordability are often “a red herring” for a constituency that wants technological progress stopped altogether. Europe, he argued, now has half its global GDP share from 30 years ago because of hostility to innovation—“socialism always results in the equal distribution of poverty and misery.”
2. AI needs a visible social dividend to retain its license
Friedberg cautioned that the economics remain unresolved despite the rhetoric around “tech barons.” Vast capital is being risked on infrastructure, yet Peter Thiel’s formulation still holds: “There’s really only one company that’s making any money, and that’s Nvidia,” whose roughly $3 trillion of market-cap creation does not establish where application-layer value will ultimately accrue.
His internet analogy was instructive: investors initially expected fiber-optic switch companies to capture the gains, but Google, Amazon, Uber, and other applications became the major beneficiaries. “The jury’s still out. We don’t even know what AI is,” making claims of permanently concentrated value premature.
Chamath nevertheless diagnosed a severe perception problem: circular deals push a handful of stocks higher while households read about electricity bills, lost jobs, weakened education, and a “sword of Damocles” above their children. His message to industry leaders was personal: “Enough of the stupid haircuts, dumb watches, ugly clothes, ostentatious displays of wealth. We’ve all done it. I’ve been guilty of it.”
The historical playbook was Carnegie’s 2,500 libraries, Rockefeller’s institutions, Ford’s wage policy, and AT&T’s Bell Labs—measurable public artifacts that made industrial progress legible. Chamath said corporate cash is valued at zero by Wall Street and should finance housing, education, and health-care benefits that tens of millions can feel.
3. Sacks said today’s labor data rebut mass displacement, not future risk
Sacks cited Vanguard’s comparison of AI-exposed occupations with all others: job growth was 1.7% versus 0.8%, while wage growth was 3.8% versus 0.7%. His causal interpretation was that productivity raises the value of labor, producing higher pay and greater demand rather than immediate redundancy.
That followed Yale Budget Lab’s assessment of 33 months after ChatGPT, which found “no discernible disruption to the job market.” Sacks also attributed roughly two percentage points of current GDP growth to the AI CapEx boom, with no aggregate job loss yet visible in the cited data.
Jason’s concrete counterexample came from Uber: where Waymo is operating, Uber has reduced efforts to recruit drivers in Los Angeles and San Francisco to avoid onboarding people into a deteriorating experience. Uber is also offering work such as data labeling—evidence, in Jason’s telling, that companies are already planning around displacement.
Sacks conceded the category-level point: autonomous driving could eliminate driver jobs, and “some jobs are going to change, some jobs will be eliminated.” His dispute concerns the net—whether new jobs and productivity gains more than offset losses—and the “motte and bailey” move between claims that mass disruption is already here and unfalsifiable predictions about the future.
4. A funded anti-AI ecosystem has shaped the US narrative
Sacks pointed to reporting that the Future of Life Institute funded journalism fellowships associated with NBC News, Bloomberg, Time, The Verge, and the Los Angeles Times. He said Vitalik Buterin’s donated Dogecoins unexpectedly became a roughly $600 million war chest supporting journalism, academic research, and organizations opposing data centers.
Citing an analysis of the “doomer industrial complex,” Sacks named Dustin Moskovitz, Jaan Tallinn, and Buterin as major funders behind hundreds of organizations, with more than $1 billion donated to the broader cause. Chamath’s framing was “whack-a-mole”: once water-use claims are rebutted, critics can move to cooling chemicals, air cooling, or the next environmental concern.
The claimed narrative effect is visible in polling: 83% of respondents in China believed AI’s benefits would outweigh its harms, versus 39% in the United States. Sacks said AI companies had played into the job-replacement narrative, suggesting it was either, as Chamath had said, tied to their next fundraising round or simply laziness; his practical answer was private power generation, factual rebuttal, and concrete affordability benefits.
5. Disinflation and private hiring underpin the 2026 bull case
CPI arrived at 2.7% versus 3.1% expected, while core CPI reached 2.6%, its lowest level since March 2021. Sacks added Kevin Hassett’s figure that core inflation had run at 1.6% over the preceding three months, reading the trajectory as evidence that inflation is “just about a solved problem.”
Employment from September through November fell by 41,000 overall, but Sacks split that into 121,000 additional private jobs and 162,000 fewer government jobs. He attributed the public-sector drop to employees accepting DOGE-related buyouts effective October 1, rather than to broad private-economy contraction.
His supporting scoreboard included a 10.7% reduction in federal workers—from 2.4 million to 2.15 million—a $600 billion year-over-year deficit reduction, national gasoline below $3, and real-wage gains averaging more than $1,000. He specified $1,300 for factory workers and $1,800 for construction workers.
Vanguard’s outlook projected 3% growth in 2026 and its strongest labor-market improvement in the second half. Sacks layered on lower rates and tax provisions covering tips, overtime, Social Security, and the standard deduction: “Sit back and enjoy this. I think we’re headed for a gangbusters 2026.”
6. Better flow data have not repaired the price-level problem
Friedberg saw fewer federal employees as a productivity story beyond payroll savings. His battery company faced a 700-page Department of Energy filing, millions of dollars in costs, and teams of 50 to 100; shifting responsibility toward existing state and local regulation could lower the burden on domestic manufacturing intended to reduce dependence on China.
Jason assessed the administration’s “3-3-3” framework—3% GDP growth, a 3% deficit-to-GDP ratio, and 3% inflation—as succeeding on growth and inflation but not yet on the deficit. Friedberg separately pointed to overtime and tip provisions, car-loan interest deductibility, and accelerated depreciation as stimulative 2026 tax measures.
Jason’s pushback was about levels and distribution, not whether 2.7% beat a forecast. Trump promised lower prices, yet grocery prices continued rising, unemployment moved from 4.0% in January to 4.6%, and people without stocks missed the equity rally: “It’s not the golden age for the bottom half of Americans.”
Sacks’s most important concession was that workers remain roughly $2,000 underwater versus five years ago after losing about $3,000 in real wages and recovering $1,000. His answer was time: the economy is “a giant supertanker,” as with Reagan’s low popularity in 1983 before “morning in America” and a landslide in 1984.
7. China’s EUV prototype compresses the export-control timeline
Reuters reported that China had assembled a prototype resembling ASML’s extreme-ultraviolet lithography system, allegedly with help from former ASML engineers. The machine had not produced working chips; China was targeting 2028, while some sources considered 2030 more realistic.
The bottleneck is unusually concentrated: ASML, valued in the discussion at roughly $400 billion, is the sole supplier; each machine costs about $250 million and takes six months to build. US pressure beginning in 2018 led the Netherlands to restrict exports, making EUV what Sacks called perhaps the West’s “single biggest advantage” in AI.
Sacks never supported selling EUV to China, though he considered attempted replication inevitable: “No advantage is forever.” If the report is accurate, it shortens the estimated chip-gap timeline from being adjacent to being a matter of years, even if one prototype remains only a data point.
8. Beijing is redesigning lithography, not merely copying ASML
Friedberg placed the prototype inside a decade-long funding structure. Phase One of China’s National Integrated Circuit Industry Investment Fund began in 2014 around manufacturing and fabs such as SMIC; Phase Two began in 2019 around design and materials; the roughly $48 billion Phase Three, established in 2024, explicitly targets manufacturing “choke points.”
A July review from a Tsinghua-linked group in Light: Science & Applications described AI-based inverse lithography. Friedberg said neural networks can compensate for suboptimal optics and predictable diffusion or shadow effects, changing the mask so the printed result approaches three-nanometer precision without requiring the same Zeiss optical performance.
Sacks’s warning was conceptual: “China is not just in a catch-up race, they’re in a primacy race.” Alternative production methods could reach parity with ASML or move ahead of it, yielding manufacturing leadership, AI leadership, and “economic leverage over the planet”—a broader threat than stolen machine schematics.
9. DUV and networking already make Huawei’s chips serviceable
Sacks noted that deep-ultraviolet lithography was expected to top out around 14 nanometers, yet China pushed it to 7 and then 5. Huawei’s Ascend chips remain inferior to US chips in his assessment, but they are “serviceable” despite the controls.
Networking closes part of the remaining gap. Huawei’s CloudMatrix 384 links 384 Ascend chips into a rack that Sacks said can perform comparably to an Nvidia rack, though at substantially greater power consumption—an expensive workaround, but still a functioning one.
The export regime therefore bought time rather than permanent exclusion. Friedberg predicted that lithography technology he believes exists in China but is not publicly discussed would be deployed in Huawei and mainland fabs in 2026; Jason summarized the likely sequence as announcements in 2026 and manufacturing impact in 2027. The Reuters report did not, in Friedberg’s view, capture the full scope of progress.
10. Memory-centric silicon could devalue the smallest-node race
Chamath distinguished today’s compute-centric chips from a possible future built for “infinite inference.” That workload may favor memory-centric architecture with large amounts of SRAM, allowing commercially valuable silicon to be manufactured at 14 nanometers and other less-advanced nodes rather than requiring every gain to come from EUV.
Such a shift transfers difficulty into compilers and software. Chamath believes the United States remains “light-years ahead” there, so successful Chinese EUV would be “bad” without necessarily becoming “the end of the world.”
He also stressed the brittleness of leading-edge production: sub-two-nanometer systems push physical limits and can suffer low yields, while compute-heavy designs depend on high-bandwidth memory supplied by SK hynix and Samsung. The strategic question is not simply who owns the smallest node, but who delivers the cheapest, fastest, highest-yield system.
11. Onshoring may reduce Taiwan’s semiconductor leverage
Friedberg suggested that simultaneous fab construction in the United States and mainland China could reduce Taiwan’s role as a strategic flash point. If the US has domestic manufacturing, it may feel less compelled to defend Taiwan for chip access; if China has mainland fabs, its economic incentive to seize Taiwanese capacity may also decline.
Sacks’s immediate prescription was more leading-edge capacity in America. TSMC’s Arizona fab and planned expansion matter, but he said permitting restrictions must be removed, contrasting that with the moratorium Sanders advocated. Jason had described Ro Khanna’s support as limited and said Khanna’s clarified position was that data centers should use renewable energy.
Friedberg explained why Chinese researchers still publish: scientists seek recognition, while local governments and competing teams need visible breakthroughs to win the next funding round. Publication can stop when a national champion emerges or research transfers into industrialization; he resisted assuming that every disappearance reflects a direct CCP order.
12. California’s tax threat is already repricing residency
Discussing moves toward Austin, the group argued that a proposed billionaire tax could alter California’s trajectory even if it never passes. Their estimate was $100 billion to $200 billion of economic impact over five to ten years because the mere threat of “property seizures” changes where wealthy residents live and where capital and the tax base remain.
Sacks called the proposal “an enormous self-own,” while Jason said people had been putting up with high taxes until property-rights threats emerged; he also linked broader relocation decisions to personal-security concerns. The feared feedback loop is fiscal: mobile taxpayers leave, revenue falls, and politicians respond with further taxation or attempts to recover the missing base.
Jason separately claimed that California’s pension system is underfunded by roughly $1 trillion while beneficiaries expect payments for another 30 to 50 years. Sacks warned that the resulting fiscal pressure could drive a “socialist spiral,” making additional state taxation harder to defend.
13. Cannabis rescheduling shifts attention toward potency labels
The hosts closed by noting that President Trump had signed an executive order moving marijuana from Schedule I to Schedule III. They supported distinguishing it from heroin, while stopping short of calling modern cannabis harmless or treating public smoking as desirable.
Friedberg said contemporary resins and similar products can apparently be about 200 times as potent as the joints of their youth and may induce psychotic states. Jason and Friedberg prioritized toxicity or potency labels, with Jason arguing for FDA regulation more like tobacco and stronger safeguards against children accessing the products.