Key Moments
Socialists Sweep NYC, China Catches Up in Coding, AI Memory Crunch, Micron's Blowout Quarter
All-In PodcastWant to know something specific about what's covered?
We've already dissected every moment. Ask and we will deliver (with timestamps).
Key Moments
Socialists are surprisingly sweeping NYC primaries, fueled by a new generation of voters and a growing disillusionment with traditional politics, while China's open-source AI models rapidly close the gap with frontier US AI.
Key Insights
Three socialist-endorsed candidates, including one who wants to 'end Western civilization,' swept New York City congressional primaries, with one candidate's win attributed to a 26% chance before election day.
The Democratic Party is facing a populist takeover from the DSA, which uses the party as a 'ballot access vehicle' to push a radical agenda including abolishing the Senate and the carceral state.
China Z.AI's GLM 5.2, an open-source AI model with 744 billion parameters and a 1 million token context window, scored 51 points on the artificial intelligence index, besting GPT-5.5 on a coding benchmark and trailing Claude Opus by less than 1%.
Micron's revenue surged 4x year-over-year to $42 billion, with its stock up 10x, driven by demand for High Bandwidth Memory (HBM) for AI, leading to significant price increases for consumer electronics like Apple's Mac Studio (up 25%).
The critical bottleneck in AI infrastructure is DRAM, with memory capacity and bandwidth being foundational to AI model performance. Experts estimate DRAM will constitute 30-40% of hyperscaler capex next year.
Orbital compute could become significantly cheaper than terrestrial data centers, with a 1-gigawatt data center costing $60 billion terrestrially versus an estimated $40 billion in space, largely due to the deflationary impact of reusable Starship launches on launch costs.
Socialist surge in NYC Congressional primaries signals a shift in the Democratic Party
A significant political upset occurred in New York City's congressional primaries as three socialist-endorsed candidates, championed by Mayor Eric Adams, swept their races. Notably, a 32-year-old Democrat socialist, who has made controversial statements including wanting to 'end Western civilization' and celebrating the slaughter of Israeli civilians, defeated a five-term incumbent in the 13th district. Another candidate unseated a two-term incumbent in a wealthy district. Prior to the election, the chances of this 'trifecta' win were estimated at only 26% by partners like Poly Market. These victorious candidates resonated with younger, college-educated, and higher-income demographics, who are described as 'people who can afford to be socialists.' This trend is seen as a populist takeover within the Democratic Party, with the Democratic Socialists of America (DSA) effectively using the party as a 'ballot access vehicle' rather than an ideological ally. The DSA's agenda includes radical proposals such as abolishing the Senate, defunding the police, abolishing prisons, and ending Western civilization, which alarm establishment Democrats who feel they are losing control of the party.
China's AI capabilities are rapidly advancing, challenging US dominance in open-source models
China is making significant strides in AI development, particularly with its open-source models. Z.AI's GLM 5.2, a 744 billion parameter model with a 1 million token context window, has demonstrated impressive capabilities, scoring 51 points on the artificial intelligence index and outperforming GPT-5.5 on a software engineering benchmark. It trails only slightly behind Claude Opus and is offered under an MIT license, making it highly accessible. This rapid progress is partly attributed to 'distillation,' a process where models leverage extensive data and reasoning traces harvested from API usage of frontier models, essentially creating a 'cheat sheet' for catch-up. Experts believe this raises questions about the efficacy of current legal restrictions and self-imposed limitations on AI development in the US, as frontier capabilities may become widely available sooner than anticipated. The implication is that the economic value may shift from frontier labs to infrastructure providers, while open-source models greatly benefit AI infrastructure providers like NVIDIA.
The critical bottleneck in AI is memory, driving up costs and impacting hardware
The ongoing AI boom has exposed a critical bottleneck: DRAM, particularly High Bandwidth Memory (HBM). Micron, one of the few companies producing HBM, has seen its revenue skyrocket by 4x to $42 billion, with its stock increasing 10x. This surge is driven by the insatiable demand for HBM, which sits atop NVIDIA GPUs and is essential for AI model performance. Micron's entire 2026 supply is sold out. This scarcity is leading to significant price increases across consumer electronics; for instance, Apple's Mac Studio saw a 25% price hike. Experts highlight that memory capacity and bandwidth are foundational to AI, and this bottleneck is expected to persist. It's estimated that DRAM will account for 30-40% of hyperscaler capital expenditure next year. While other companies are attempting to enter the market, producing specialized DRAM for AI servers is extremely difficult, with only three companies currently capable. This situation underscores the strategic importance of memory in the AI supply chain.
Social media's impact on youth and the debate over potential bans
The conversation touched upon the potentially detrimental effects of social media on young people. Some suggest that banning social media for individuals under 16, as implemented in Canada and explored in Florida, could be a crucial step to mitigate radicalization and promote healthier information consumption. This approach is viewed as a way to prevent young people from becoming addicted to social media 'drugs' too early, allowing them more time to develop critical thinking skills. However, a counterargument posits that such bans could be a precursor to broader censorship regimes. By forcing young users to age into social media access, the goal might be to deanonymize users and enable authorities to control content, particularly dissenting views. This highlights a tension between protecting youth and preserving free speech in the digital age.
The economic case for orbital compute and distributed AI infrastructure
The discussion explored the burgeoning economics of orbital compute and distributed AI infrastructure as potential solutions to escalating terrestrial data center costs. The high cost of building a 1-gigawatt data center, estimated at $35 billion for semiconductors and $25 billion for power and cooling, coupled with inflationary pressures and regulatory hurdles, makes space-based compute increasingly attractive. Reusable rockets like SpaceX's Starship could dramatically reduce launch costs, potentially bringing the cost of putting a gigawatt of compute into orbit down to around $40 billion. This cost advantage, combined with the potential for deflationary launch economics, could make orbital compute more cost-effective than ground-based solutions in the coming years. Furthermore, the concept of distributed inference, where computing power is spread across numerous smaller, accessible nodes, could lower the cost of AI services and enable greater participation in the AI ecosystem.
AI's role as an economic leveler and the importance of responsible communication
AI is presented as the ultimate economic leveler, capable of democratizing access to expertise and intelligence. By transforming vast amounts of information into actionable insights, AI can empower individuals with capabilities previously only available to a select few. The analogy is made to having a 'superfounder' or a brilliant co-founder available to every person, regardless of their background. However, the speakers lamented that the potential of AI is being undermined by poor communication and the spilling of internal Silicon Valley conflicts into public discourse. This has led to a loss of credibility and allowed alternative, often more radical, narratives to gain traction, filling the vacuum of trust. To harness AI's potential, a concerted effort is needed to present it as a tool for widespread benefit rather than a source of fear and job loss.
The challenge of pricing IPOs in a booming AI market
The segment covered the complexities of Initial Public Offerings (IPOs) in the current market, particularly for AI companies like Cerebras. The conversation highlighted the tension between aggressive IPO pricing to capitalize on market enthusiasm and the risk of stocks falling below their offering price, which can trigger automatic selling by portfolio managers. Companies are advised to price IPOs thoughtfully to avoid breaking deal price within the first nine months. The sheer volume of potential offerings, including Anthropic's projected $3 trillion valuation, raises questions about absorption by global capital markets. However, it's argued that the market has already absorbed much of this value through private investments, and the shift to public markets involves moving from larger private markets to even larger public ones. The discussion also touched on the historical context of IPO pricing, recalling when $1.5 billion or $14 billion were considered groundbreaking figures, underscoring the unprecedented scale of current valuations.
The struggle to assimilate migrants and rising anti-American sentiment
The conversation also delved into the challenges of migrant assimilation in the US. It was suggested that the breakdown of the traditional 'melting pot' approach due to multiculturalism and 'wokeism' has led to a situation where newly arrived migrants and, in some cases, specific political factions, are not integrating into American society. This lack of assimilation, coupled with the rise of socialist ideologies, has resulted in candidates openly declaring anti-American sentiments and advocating for radical societal transformations that mimic the systems of their countries of origin. This trend is seen as alarming, particularly when combined with policy choices that may have inadvertently empowered these movements and alienated segments of the traditional Democratic base, such as working-class and minority voters.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Books
●Concepts
●People Referenced
Common Questions
The socialist sweep in NYC elections indicates a shift in the Democratic Party, where candidates endorsed by figures like Zohran Mamdani, often with radical platforms, are gaining traction. This is seen by some as a populist takeover similar to Trump's influence on the Republican Party.
Topics
Mentioned in this video
Travis Kalanick's new venture.
Partnering with Elon Musk on the TerraFab project.
Company that makes racks for data centers.
Company co-founded by Elon Musk, recently had an IPO that Gavin Baker discussed.
Platform whose data showed a big upset for the Mandami sweep chances.
Used as an example of a past economic unlock that harvested the world's knowledge, but lacked transformation into expertise.
Mentioned in the context of anti-AI groups receiving funding and its models being rolled back due to jailbreak issues.
Mentioned funding pro-AI groups and its models also experiencing issues with approval hoops.
The Chinese company that released the GLM 5.2 open-source AI model.
Considered the American open-source champion, makes GPUs and also potentially at odds with OpenAI regarding chip development.
Reportedly building OpenAI's Jalapeno chip.
Chinese national champion in chip development, used as a training platform for GLM5 family.
An AI model claimed to be trained on Huawei chips.
One of three companies globally making High Bandwidth Memory (HBM), which smashed earnings and revenue, with 2026 supply sold out.
Another company that makes High Bandwidth Memory (HBM).
Another company that makes High Bandwidth Memory (HBM).
Discussed for increasing prices on its products due to rising DRAM costs.
Company that was once priced at 14 billion for a private round, indicating a shift in capital market expectations.
Company that makes modules for data centers.
Key semiconductor manufacturer, mentioned for the 4-month lead time to produce wafers after an order.
Chinese company going public, expected to flood the market with cheap consumer-grade DRAM.
Mobile network operator, rumored to be a potential acquisition target for Elon Musk.
Mentioned in the context of a potential merger with SpaceX and the Megapod trademark.
A project for distributed compute where individuals contribute computing power.
Company in the AI chip space, had a tough IPO with its stock breaking deal price.
DSA platform proposal to abolish ICE.
A project related to distributed compute.
Organization gaining momentum in New York City primaries, with radical proposals.
DSA platform proposal to defund the Department of War.
Mentioned as a model for a desired AI regulatory agency.
Co-host of the podcast, known for his past role at Uber, now involved with Adams.
Guest from Artemis Management, discussing markets and economic perspectives.
Mentioned in a quote about treating every game as 0-0, regardless of score.
New York City politician whose endorsed candidates won their primaries, seen as a "spiritual leader" of the DSA movement.
10th district leader who defeated incumbent Dan Goldman in a New York congressional primary.
Two-term incumbent in New York's 10th district, defeated by Brad Lander.
House Speaker whose backed incumbent was defeated by Jiovalier, facing a headache from new DSA members.
Won an open seat in New York's 7th district, which is a DSA stronghold.
Mentioned in comparison to the populist takeover of the left and his playbook being adapted by Mandami.
Discussed for his contributions to decarbonization, his focus on memory for the terrafab, and potential for a Tesla/SpaceX merger.
Author of 'The Fountainhead', which was Travis Kalanick's Twitter avatar.
Mentioned as leaving the Republican party over the Israel issue.
Journalist mentioned as 'going mental' over the Israel issue and where Chamath discussed the separation of Jews, Israelis, and Netanyahu.
Former US President, mentioned for his charisma in comparison to Mamdani.
Prime Minister of Israel, whose actions are a source of concern and whose policies are conflated with the broader Jewish and Israeli identity by some.
Former Tesla AI director, mentioned as someone who coined the term 'council of LLMs'.
Friend mentioned in the context of the 'social contract'.
Person whose company, Crusoe, is working on modularly assembling data centers.
Location of recent socialist political sweeps and primaries.
A key issue in the DSA platform, advocating for a 'free Palestine'.
Mentioned as a country about which the Chinese AI model GLM had a censored response.
Historical site about which the Chinese AI model GLM had a censored response.
City for which the Chinese AI model GLM provided exceptional tourism recommendations.
A political and economic theory discussed as a growing force in US politics, with specific criticisms of its outcomes in Canada, UK, and Australia.
Discussed as an inherent human tendency toward laziness, and as the underlying reality of the DSA's 'democratic socialism'.
Elon Musk's initiative focusing on memory production, seen as a solution to the DRAM bottleneck.
A type of DRAM mentioned as hard to make and increasingly needed in AI data centers.
A Frontier-class open-source AI model released by China Z.AI with 744 billion parameters, strong coding performance, and low API cost.
OpenAI's AI model that GLM 5.2 beat on the Frontier SWE coding benchmark.
Anthropic's AI model that GLM 5.2 nearly matched in performance.
An AI model, recommended for business owners due to its dedication to truth.
Company that demonstrated fixing censorship in Chinese open-source models.
A platform where users can rent H200 GPUs by the hour.
AI model developed by Anthropic.
A project related to distributed compute.
Huawei chips used to train the GLM5 family of AI models.
A recently filed trademark for modular data center hardware for AI computing, speculated to be battery packs with GPUs at supercharger stations.
Apple product mentioned as being purchased in higher capacities due to AI requirements, now facing price increases.
Apple product mentioned as being purchased in higher capacities due to AI requirements, now facing price increases.
Apple product, possibly misheard as 'MacBook Neo', a $699 laptop whose price increased.
Mentioned as a competitor to Apple's lower-priced laptops.
Microsoft's console, whose price was raised, indicating price increases in consumer electronics due to AI demand for components.
Tesla's home battery product, speculated to potentially integrate GPUs for distributed compute.
Gaming console whose price might be affected by rising component costs.
Gaming console whose price might be affected by rising component costs.
SpaceX's satellite internet service, potentially used in a distributed compute system with Powerwalls.
A type of GPU available for rent on platforms like Targon.
Older NVIDIA GPU model that can be made competitive for decoding with Grock or Cerebras chips.
Older NVIDIA GPU model that can be made competitive for decoding with Grock or Cerebras chips.
More from All-In Podcast
View all 426 summaries
64 minGameStop CEO Ryan Cohen’s $56B Plan to Take Over eBay
85 minWorld's First Trillionaire, Anthropic Fable Banned, The New Oligarchs, Iran Peace Deal
102 minAnthropic's Fable Backlash, Nationalizing AI, Inflation Heats Up & California’s Broken Elections
68 minAll-In's Best Ideas Pitch Competition: 4 Investors Present Their Top Trades Live
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free