Summary

San Francisco’s National Guard argument turned on whether a recovering city still needs a federal backstop. Jason Calacanis cited crime down 30% citywide and 40% downtown, homicides at a 70-year low, car break-ins at a 25-year low, and convention and hotel bookings up 50% and 60%; David Sacks countered that Market Street remains an open-air drug market that authorities proved they could clear within 48 hours for President Xi’s visit. Jason proposed a three-month, tightly bounded deployment around BART and the Tenderloin, preserving the episode’s core disagreement: “If the city’s doing their job and the city’s improving, why do you need that?”
The sharper San Francisco call was to stop financing addiction and institutionalized homelessness. Friedberg argued that the city provides something like $2 billion to nonprofits rewarded for expanding their “flock,” while Jason estimated a $700 million-$800 million homeless budget against 8,000-20,000 people—roughly $52,000 per person at the midpoint. Sacks’ emblematic case was the $5 million-a-year managed alcohol program dispensing beer to homeless alcoholics; Jason summarized the absurdity as “the treatment is beer.”
China’s rare-earth escalation turned an anti-price-control panel into a case for strategic industrial policy. China announced controls on 12 of 17 critical minerals effective December 1, after decades of subsidies that could destroy Western competitors by flooding spot markets; Chamath Palihapitiya argued that the US therefore needs a “last resort buyer of record” and strategic reserves analogous to petroleum. Friedberg preferred deregulation, tax incentives, new mines and cleaner processing, but agreed America could unlock deposits and capacity rather than accept Chinese dependence.
The near-term trade setup favored de-escalation even as both countries reduced strategic dependence. Scott Bessent said the Trump-Xi meeting remained on track, the threatened additional 100% tariff for November 1 “does not have to happen,” and the sides had “substantially deescalated”; prediction markets assigned only 15% odds to the tariff and 73% odds to an agreement by November 10. Sacks’ call was that a top-level bargain can contain bureaucratic escalation without ending long-term competition.
The panel framed China as a reascending power whose state-directed capital system is structurally formidable, not invulnerable. Chamath said China held the world’s largest GDP during 70% of the years since 1500 and described national priorities cascading into roughly 300 provincial “VCs,” producing companies such as BYD and Xiaomi. Sacks countered with 20% youth unemployment, a demographic time bomb, a moribund real estate market and very little to no foreign direct investment.
Local resistance is becoming a material constraint on the AI capex supercycle. Google withdrew a proposed $1 billion Indianapolis data center, while Microsoft in Wisconsin and Amazon near Tucson also pulled or mothballed projects amid electricity, water and noise concerns. Chamath’s warning was that “tokens are not jobs”: hyperscalers may need to subsidize local power, solar and storage with their own free cash flow instead of expecting communities to absorb the costs.
The AI labor debate remained unresolved because strong aggregate growth and shrinking technology payrolls can both be true. Sacks attributed 40% of a 3.8% growth quarter to AI—implying 2.3% without it—and argued that humans remain “end to end” while AI is “middle to middle.” Jason replied that GDP is not wages and raised the exposure of drivers, developers and other workers; Sacks then cited Alphabet at 187,000 employees versus a 190,000 peak, Meta at 75,000 versus 86,000, Uber at 31,000 versus 33,000, and Amazon at 1.55 million versus 1.6 million. His hedge mattered: the new jobs might not arrive as quickly as automation removes or suppresses the old ones.

Deep dive

1. San Francisco’s rebound weakens the case for an indiscriminate intervention

Jason said Mark Benioff’s National Guard controversy began more innocuously than headlines implied: while discussing several hundred off-duty officers hired for Dreamforce, Benioff was asked whether he would use Guard members and answered, “Sure, if they can be cops.” Sacks nevertheless endorsed federal help because Market Street still resembles a “zombie city” organized around open-air drug markets.
Jason’s empirical case was that city data showed crime down 30% overall and 40% downtown, homicides at a 70-year low, tents at record lows, car break-ins at a 25-year low, and the first net increase in police officers in seven years. Convention bookings were up 50% and associated hotel bookings 60%, matching his own observation that the city is “on the upswing.”
Jason also cited District Attorney Brooke Jenkins’ figures: felony conviction rates reached 88% for narcotics, 85% for robbery, 72% for burglary and 70% for gun possession. Friedberg then asked the load-bearing question: “If the city’s doing their job and the city’s improving, why do you need that?”

2. A targeted federal operation became the compromise position

Sacks praised Mayor Daniel Lurie and Jenkins while arguing that left-wing judges and the city’s legal culture still constrain them. His proof of possibility was Gavin Newsom’s cleanup before President Xi’s visit: the drug dealers and addicts “magically” disappeared within 48 hours, showing that the remaining blight is a policy choice rather than an immovable condition.
Jason’s proposed split-the-difference was three months of Guard presence at every BART station and four targeted Tenderloin corners, requested by Lurie and presented as a limited test. He expected perhaps 10%-20% of San Franciscans to protest or riot, while the rest might accept a clearly bounded deterrence operation.
Sacks pointed to Washington Mayor Muriel Bowser as the political template: she publicly disliked Trump’s Guard deployment but cooperated, and residents benefited. He rejected protest threats as a veto over enforcement—“We should not be held hostage”—and favored a precise operation against the network supplying downtown fentanyl, followed by deportations.

3. San Francisco’s homelessness economy rewards continued failure

Friedberg’s diagnosis moved beyond policing: the city provides “something like $2 billion a year” to nonprofits that manage homeless and addicted populations, rewarding them for adding people to their “flock.” Because benefits attract people from outside San Francisco, he argued, spending functions as a regional magnet while sustaining the addiction it nominally treats. Sacks agreed with the point.
Sacks advocated forced transitions into treatment and used the managed alcohol program as the defining absurdity: $5 million annually funds hotel-based beer distribution to homeless alcoholics. He also described a closed-off area where people can use needles and remain in a permissive environment. Jason summarized the program’s logic as “the treatment is beer.”
Jason separately placed the homeless budget between $700 million and $800 million, against estimates of 8,000-20,000 people. Splitting those ranges produced roughly $52,000 per homeless person annually, while Sacks cited Thomas Sowell’s aphorism: “You get as much homelessness as you’re willing to pay for.”

4. Rare-earth price floors are security policy disguised as market intervention

China’s new controls covered 12 of 17 critical rare-earth minerals beginning December 1; Trump answered with a threatened additional 100% tariff on Chinese imports from November 1 and accused Beijing of trying to “hold the world captive.” Markets sold off, then recovered as both sides signaled de-escalation.
Friedberg explained the conventional objection to price floors: mandating a minimum price inflates inputs and distorts markets. His preferred remedy was regulatory relief, tax incentives and other economic incentives, because the US regulatory environment helped move mining and processing offshore in the first place.
Sacks’ exception was that rare earths have not operated as a free market for roughly 30 years. China used massive subsidies—and WTO rules that allowed so-called developing nations to subsidize industries—to eliminate competitors, then retained the ability to slash prices whenever Western investors attempted to rebuild capacity. Price certainty is meant to neutralize coercive leverage, not bureaucratically price an ordinary consumer good.
Mountain Pass supplied the cautionary history: Molycorp failed, its asset became MP Materials, and the US government later entered a major deal to revive output. Neodymium moved from roughly $50,000 per ton in 2020 to nearly $250,000 during COVID before falling again, volatility that makes financing multiyear capex exceptionally difficult.

5. A strategic reserve can break China’s spot-market weapon

Chamath illustrated the financing mismatch through take-or-pay contracts. MP Materials and his LFP cathode business secure commitments from buyers such as General Motors, then take those contracts to Wall Street to finance factories; Chinese competitors can instead receive provincial subsidies, loan guarantees and direct state support before producing anything.
Once a US factory begins operating, China can dump enormous volume into the spot market, crater prices and leave contracted buyers paying more than competitors. Chamath therefore favored the federal government as “a last resort buyer of record,” absorbing shocks through strategic reserves and swapping materials to Tesla, GM or robotics companies when spot supply becomes unavailable.
Friedberg said the most important future market may be robotics rather than consumer electronics. He emphasized that permanent magnets enable the actuators moving robotic joints, while Sacks noted a long tail of perhaps 20,000 magnet types embedded throughout electric motors and industrial manufacturing; dependence extends well beyond simply mining ore.

6. Domestic processing offers an escape from permanent subsidy

Friedberg separated mining from refinement: Mountain Pass produces bastnäsite ore, but strong-acid leaching, extraction and separation create hazardous waste and worker exposure. America stopped meaningfully developing this chemical and metallurgical stack about 40 years ago, leaving dirty legacy processes as a major reason production concentrated in China.
His optimistic case was that automation and newer techniques could make those well-understood steps safer and scalable. Radar, microwave imaging, interferometry and the US Geological Survey’s “Earth MRI” could also reveal deposits across Texas, Wyoming, Colorado and Missouri; he suggested there could be well over 1,000 times the proven reserves currently catalogued.
The disagreement was therefore about mechanism, not the goal. Chamath wanted methodical federal purchasing rather than trying to time commodity markets; Friedberg believed deregulation, discovery and modern processing could eventually let private supply outpace China. Both rejected continued dependence as the default.

7. A Trump-Xi bargain could contain escalation without ending rivalry

Bessent said the leaders’ South Korea meeting remained on track, the 100% tariff need not occur, and the conflict had “substantially deescalated.” Prediction markets agreed: 15% odds of the tariff taking effect by November 1 and 73% odds of a US-China trade agreement by November 10.
Sacks argued that this relationship works best through agreement “at the top,” because ministries otherwise produce misunderstandings and bureaucratic food fights. China’s mineral controls followed a September 29 US rule extending export restrictions to affiliates more than 50% owned by listed entities, illustrating how defensible actions can still trigger disproportionate retaliation.
His expected equilibrium was continued trade with mutual de-risking. Neither side wants the other to convert supply-chain leverage into geopolitical coercion, so a grand bargain can stabilize competition without recreating the old assumption that economic interdependence eliminates security conflict.

8. China is reascending through a state-directed venture network

Chamath rejected “rising power” for “reascending power”: since 1500, China had the world’s largest GDP during roughly 70% of those years. The national motivation, in his telling, is not discovering potential dominance but reclaiming a position China believes it historically held.
He described capital allocation cascading from Xi Jinping and a close group of seven or eight people, through the Politburo, provinces and prefectures. Once EVs became a priority in 2006-2007, roughly 300 local capital allocators behaved like venture investors competing against a national mandate; that system ultimately produced companies such as BYD and Xiaomi.
Sacks placed the same project institutionally: China has built the Shanghai Cooperation Organization, promoted BRICS and expanded Belt and Road as alternatives to the US-led NATO, G7, IMF and World Bank architecture. Sacks countered Jason’s view that China was simply advantaged by citing 20% youth unemployment, a demographic time bomb, a moribund real estate market and very little to no foreign direct investment.

9. America’s China error combined ideology, greed and fragile optimization

Sacks called the post-Cold War mistake hubris: leaders treated the 1991-2017 “unipolar moment” as permanent, embraced the “end of history,” and assumed wealth would turn China into a liberal democracy. Instead, modernization made China richer and more capable without making its government more Western.
Jason’s dissent was simpler—“It wasn’t hubris, it was greed.” CEOs moved production to maximize margins; Chamath added that compensation increasingly rewarded gameable measures such as EPS, encouraging executives to optimize global profit flows while ignoring national resilience and long-duration supply risk.
Friedberg added the mistaken belief that America could monopolize knowledge work while outsourcing labor. The internet diffused expertise globally. Jason separately pointed to industrial concentration and the vulnerability created by roughly 90% of chips coming from Taiwan. Huntington’s competing thesis survived better: other civilizations sought modernization, not Westernization.

10. Local opposition is now gating the AI capex cycle

Chamath identified three withdrawals in one week: Google abandoned a planned $1 billion Indianapolis data center before an expected adverse council vote, Microsoft pulled a Wisconsin proposal, and an Amazon project near Tucson was mothballed. Residents repeatedly cited higher electricity prices, uncertain water neutrality and noise pollution.
His distinction was crucial: these communities were not primarily voting from immediate fear of losing their own jobs. They were processing a story in which AI might eventually take employment while data centers raise bills today—“tokens are not jobs”—and then using municipal zoning to cancel infrastructure essential to the sector.
The political risk extends beyond three sites. Sacks counted roughly 1,000 AI bills moving through state legislatures, often driven by pressure to “do something,” while Chamath argued that 50 different regulatory regimes would enrich lawyers, slow deployment and hand China an avoidable advantage.

11. Hyperscalers may need to purchase their social license

Chamath urged hyperscalers to use free cash flow as a community cushion: pay higher electricity tariffs, subsidize residents’ bills, or finance local solar and storage. He argued that public markets do not value their idle cash, noting that Facebook’s stock rose when it burned through cash, so these concessions could become part of the capex supercycle rather than mere philanthropy.
Jason’s Base Power example made the mechanism concrete: he said he thought households pay roughly $500-$1,000 to install batteries that charge when electricity is cheap and discharge when it is expensive. Equipping an entire 3,000-home development could create a distributed network serving residents and data-center demand without requiring rooftop solar.
Chamath also called for credible spokespeople able to translate technical progress into plain product value. Between Anthropic AI doomerism and parental alarm over OpenAI’s push toward erotica, he saw an industry allowing its most inflammatory narratives to define public opinion: “We need to have better, more eloquent, reasonable spokesmen.”

12. AI’s growth dividend does not settle the labor-displacement debate

Sacks’ macro case began with 3.8% quarterly growth, 40% of which he attributed to AI; without the boom, his math yielded 2.3%. He saw no evidence yet of widespread job loss and asked whether policymakers wanted an economy growing around 2% or one approaching 4%.
His labor model was “humans are end to end; AI is middle to middle.” People choose objectives, prompt, validate hallucinations and iterate, while models perform intermediate work. As agriculture fell from 50% of US employment in 1900 to 2% in 2000, displaced workers did not remain permanently idle; they moved into factories and services.
Friedberg’s stronger claim was that recruitment precedes obsolescence: workers would be recruited into new, potentially higher-paying jobs before buggy driving disappeared. Sacks and Friedberg extended the analogy to highways, tires, motels, drive-throughs, taxis and car culture—downstream industries impossible to enumerate when the enabling technology first appeared.
Jason argued that GDP is not wages, noted that only roughly the top 50%-55% of the country owns equities, and raised the exposure of drivers and young developers. Sacks then cited Alphabet falling from 190,000 employees in 2022 to 187,000, Meta from 86,000 to 75,000, Uber from 33,000 to 31,000, and Amazon from 1.6 million to 1.55 million even as earnings expanded. With developer unemployment rising, Sacks’ concern was timing: new jobs “might not show up as quickly as the old jobs.”

Chinese AI – They're Just Like Us? With Beijing-Based Concordia AI CEO Brian Tse Episode 159 - October 17, 2025