Key Moments
SpaceX’s $2T Case, Nvidia’s Shock Selloff, America Turns on AI, Trump Pulls AI Order, Bond Crisis?
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
SpaceX's $2T IPO hinges on its 'Elon Web Services' boasting a $15B/year deal with Anthropic, a valuation that dwarfs all others, but the deal has a 90-day cancellation clause.
Key Insights
Andrej Karpathy has joined Anthropic to lead a new pre-training team focused on recursive self-improvement, potentially accelerating AI model development by orders of magnitude annually, akin to a new form of Moore's Law.
Anthropic achieved EBIT positive status in the most recent quarter, contributing to a narrative of strong ROI for AI investment, with projected ARR for private LLM companies potentially reaching $200-400 billion by year-end.
SpaceX's S-1 filing reveals Starlink as the current revenue driver with $11.4 billion in revenue and $4.4 billion in operating income, while its AI business (XAI) and space business reported significant operating losses.
SpaceX's valuation is heavily supported by its 'Elon Web Services' (EWS), which includes a $1.25 billion/month, $45 billion over three years, deal with Anthropic for compute rental, a revenue stream comparable to Starlink.
NVIDIA reported staggering Q1 earnings with $81.6 billion in revenue (up 85% YoY) and $58 billion in net income, while also announcing $80 billion in additional share buybacks and a significant dividend increase.
The market is signaling potential economic distress with elevated oil prices driving inflation, projected CPI hitting 6%, rising bond yields globally, and record borrowing by retail investors in Korea for AI chip stocks.
Andrej Karpathy bolsters Anthropic's AI advancement
The discussion kicks off with Andrej Karpathy joining Anthropic to spearhead a new pre-training team focused on recursive self-improvement. Karpathy, a luminary in AI with previous roles at OpenAI and Tesla, is expected to drive breakthroughs in AI models that can improve themselves. This move is seen as a significant development, potentially enabling AI models to achieve order-of-magnitude improvements annually, creating a new form of Moore's Law for AI. Anthropic's recent financial success, including achieving EBIT positive status, strengthens the narrative of AI's profitability. With projected ARR for private LLM companies potentially reaching $200-400 billion by year-end, the economic viability of large language models is becoming increasingly apparent. Karpathy's work on recursive self-improvement and continual learning are highlighted as potential "final frontiers" for AI, capable of accelerating future advancements significantly.
The evolving landscape of AI perception and regulation
A significant portion of the conversation explores the growing public apprehension towards AI, labeling it as 'anti-human.' This sentiment is partly attributed to the narrative surrounding AI-driven layoffs and the perceived power imbalances it creates. Speakers discuss how CEOs sometimes frame AI in ways that stoke these fears, intentionally or unintentionally. There's a call to focus on AI's positive end-user applications, such as solving complex mathematical problems or identifying drug candidates, to counter the 'boogeyman' narrative. The inclusion of AI models like Gemini Nano in operating systems without explicit user notification, though not deemed 'covert,' raises privacy concerns and highlights the need for clearer communication and user awareness.
SpaceX's monumental IPO and the rise of 'Elon Web Services'
SpaceX's S-1 filing for its upcoming IPO, aiming for a $2 trillion valuation, is a central topic. The filing details three core business units: Starlink, the current revenue generator ($11.4 billion revenue, 50% growth, 10+ million subscribers); the Space business (4 billion revenue, 17% growth, operating losses); and AI (XAI) ($3.2 billion revenue, doubling YoY, but significant operating losses). The most compelling aspect is 'Elon Web Services' (EWS), driven by a staggering $1.25 billion per month or $45 billion over three years deal with Anthropic for compute rental. This single deal nearly matches Starlink's revenue and significantly bolsters SpaceX's valuation case. The company’s rapid data center build-out, with decreasing construction times (122 days for the first, down to 66 for the third), is also a key advantage, enabling them to deploy compute infrastructure much faster and at a lower cost than competitors. This compute capacity is crucial for AI companies like Anthropic, positioning SpaceX as a critical infrastructure provider.
NVIDIA's dominance and market signals
NVIDIA's earnings report continues to impress, with Q1 revenue reaching an astounding $81.6 billion (up 85% YoY) and $58 billion in net income. Despite this stellar performance and a $5.3 trillion market cap, the stock's performance is noted as relatively subdued, with an 16% year-to-date gain. The company announced an additional $80 billion in share buybacks and a significant dividend increase, highlighting its immense profitability and shareholder return strategy. NVIDIA's new reporting structure separating data center and AI segments, including hyperscalers and AI clouds, allows for a better comparison with competitors like Broadcom. Despite narratives of share loss, NVIDIA's growth rate in key segments appears to outpace rivals, fueled by its unique co-design approach with AI model developers.
Macroeconomic concerns and differing market outlooks
The broader market presents mixed signals. Elevated oil prices are a significant driver of inflation, with projections suggesting CPI could reach 6% in Q2. Bond yields are rising globally, and retail investors in Korea are reportedly borrowing record amounts for AI chip stocks, reminiscent of past crypto bubbles. While some analysts express concern about a potential credit crisis and economic downturn (labeled 'Dr. Doom' predictions), others, like Gavin Baker, suggest that while rates are rising, the strength of AI fundamentals and America's self-sufficiency in energy and food position it favorably. The comparison is made to the tech bubble, where Cisco was a darling; now, NVIDIA is seen as potentially undervalued relative to its growth, with its CPU business also showing remarkable expansion.
The debate on AI regulation and global competition
The discussion touches upon the potential need for regulation, particularly around 'frontier models,' with a debate on whether the US should regulate unilaterally or in coordination with other nations, like China. While some advocate for government oversight, others express concerns about the government's increasing power and the difficulty of unwinding regulations once implemented. On the geopolitical front, the recent US-China tech summit yielded few definitive policy breakthroughs, though some business development occurred, such as agricultural and chip sales. The participants note that China's growing technological capabilities necessitate a careful balance of competition and cooperation to avoid a global imbalance of power.
Space-based data centers and the future of infrastructure
A forward-looking perspective explores the feasibility of space-based data centers. While technically challenging, the rapid progress of SpaceX with Starlink and Starship makes this a plausible future. The idea is to create a communication and compute infrastructure not controlled by terrestrial governments, serving as a potential backup for civilization and progress. This concept aligns with SpaceX's original vision of backing up the biosphere. The development of rapid reusability in rockets like Starship is critical for making such ambitions economically viable. While orbital compute is still a longer-term play, the presence of working GPUs in space today indicates that such advancements are progressing.
The human element in AI adoption and job displacement
The conversation highlights the human impact of AI, particularly regarding job displacement. The 'measurable' layoffs at Cloudflare, justified by AI efficiency, are presented as a particularly poor communication strategy that alienates employees. Conversely, the positive impacts of AI, such as transforming a child's life through genetic mutation treatment or allowing nurses more patient time, are emphasized as crucial stories to tell. The discussion also touches on the potential for phased rollouts of AI technologies like self-driving cars and the need to consider the desires of affected workers, such as truck drivers, rather than making assumptions about their preferences.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Books
●Concepts
●People Referenced
Common Questions
Andrej Karpathy, a legendary figure from OpenAI and Tesla FSD, joined Anthropic to lead a new pre-training team. Their focus is recursive self-improvement, meaning AI models like Claude will be designed to improve themselves, potentially leading to unprecedented yearly advancements akin to a new Moore's Law.
Topics
Mentioned in this video
Founder of Treaties Management, joining the podcast as a guest for a market and tech discussion.
Legendary tech figure, founding member of OpenAI and former FSD lead at Tesla, who recently joined Anthropic to lead a new pre-training team focused on recursive self-improvement in AI models.
A prominent Google Fellow noted for his consistent presence at the forefront of AI waves within Google.
CEO of SpaceX and XAI, who has publicly stated a vision of incredible abundance and optional work due to AI, funded by universal basic income.
CEO of NVIDIA, praised for his effective advocacy for AI's positive impacts and his efforts to ensure NVIDIA GPUs are actively used.
Former Google CEO, whose commencement speech was booed by young people due to anti-AI sentiment.
A venture capitalist whose wife, Felicia Horowitz, has done incredible work with the Las Vegas Police Department to dramatically reduce crime using technology.
CEO of Cloudflare, whose memo about laying off 'measurers' due to AI-driven efficiency gains was criticized for its poor messaging and dehumanizing language.
CEO of Meta, who conducted dystopian layoffs while simultaneously announcing plans to record employee computer activity to train AI models, fueling public fear about job displacement.
CTO of Palantir, who advocates for focusing on AI's positive end-user impacts in factories and hospitals rather than just listening to model makers.
The President of China, who held high-level meetings with US tech CEOs and the US President to discuss de-escalation of tensions and potential partnerships.
Mentioned as the host of a talk where Gavin Baker made an interesting point about the discrepancy in revenue multiples between chip and data center manufacturing companies.
A famous investor known for predicting the 2008 financial crisis, who made a bear case that GPU/CPU amortization schedules were too long, overstating profits for hyperscalers.
A brilliant Rhodes Scholar whose fund has grown exponentially, noted for his portfolio's significant puts on the chip sector, though the context suggests these might have been short-term hedges.
Former US President, whose policy goals are seen as relatively benefiting from closed straits (like Hormuz) due to America's energy self-sufficiency, and his perceived ability to bond with dictators.
President of Russia, mentioned in the context of performing a 'relationship bonding moment' with Xi Jinping following a US visit to China, signaling ongoing geopolitical dynamics.
A leading AI research and deployment company, co-founded by Andrej Karpathy, known for models like Codex and being a major player in the AI industry.
An AI safety and research company that recently hired Andrej Karpathy and is experiencing extraordinary success and profitability, making significant advancements in recursive self-improvement.
A leading graphics processing unit (GPU) manufacturer whose Q1 earnings were mind-boggling, showing massive growth and profitability, making it the most valuable company in the world.
A platform where Andrej Karpathy's open-source tool, 'auto research,' gained over 82,000 stars, allowing AI models to improve themselves through five-minute experiments.
A tech giant whose culture previously singled out singular technical talents as Google Fellows. They also covertly included the Gemini Nano model in their Chrome browser.
Elon Musk's electric vehicle and clean energy company, where Andrej Karpathy led the Full Self-Driving (FSD) team, pioneering brute force computation.
Elon Musk's AI company, which is an important player in the AI landscape, currently offering Grok 4.3 on the Pareto frontier of frontier models.
A company building small, verticalized language models for corporations, which are considered the future of AI.
A company making humanoid robots, with a robot observed sorting packages for a week, prompting discussion about the impact of such robots on jobs.
A self-driving taxi service, discussed in the context of cities potentially banning or limiting self-driving technology to protect human jobs.
Google's self-driving car company, with its service mentioned in the context of cities potentially banning or limiting self-driving technology.
A ride-sharing service, used as an analogy for how inconvenient cities without self-driving cars might feel, similar to early days without Uber.
An AI tool using cameras to monitor crime, discussed in terms of privacy concerns and local government control, with features like rolling databases and audit trails.
A public company whose CEO's memo about AI-driven layoffs of 'measurers' (managers/data measurers) sparked controversy due to its insensitive messaging.
A software company where Sham Sankar serves as CTO, advocating for a focus on AI's end-user benefits.
A 'neo cloud' provider that was 'saved' by NVIDIA's strong earnings, enabling it to secure asset-backed financing for GPUs at low rates and sign 6-year contracts.
The Saudi Arabian oil company, whose $29 billion IPO was the largest ever, set to be more than doubled by SpaceX's anticipated $1.75 trillion IPO.
Elon Musk's aerospace manufacturer and space transportation services company, which filed its S1 for a $1.75 trillion IPO, featuring Starlink, a growing space business, and 'Elon Web Services' (EWS) for AI compute.
A networking hardware company, referred to as the 'NVIDIA of the tech bubble' in comparison to its stock valuation during the dot-com era.
A semiconductor company whose AI semiconductor revenue growth (143% YOY) is compared to NVIDIA's, in the context of a narrative about NVIDIA losing market share to TPUs.
A company invested in by Treaties Management, whose chips (not designed for space) are used in Starlink satellites, highlighting SpaceX's expertise in engineering payloads for commercial semiconductors.
Anthropic's AI model, which Andrej Karpathy will focus on improving through recursive self-improvement, aiming for orders of magnitude improvement yearly.
Google's AI model, with the Nano version covertly included in the Chrome browser, performing tasks like proofreading and autocomplete.
Google's web browser, which quietly included the Gemini Nano model without prior announcement, raising privacy concerns.
A predictions market platform mentioned for its 71% chance prediction that SpaceX would close its first day of trading with a market cap above two trillion.
OpenAI's coding AI model, mentioned as an example of a 'harness' that Grok previously lacked.
An AI model development company that was 'dead in the water' for compute access, but after Elon Musk granted them access to Colossus 2, their Composer 2.5 model achieved Pareto dominance.
XAI's AI model, with version 4.3 on the Pareto frontier of frontier models, and the recent release of 'Grok Build' providing an essential harness/runtime environment.
A university whose papers indicate significant room for re-architecting AI models and deployments.
The former main security agency of the Soviet Union, mentioned in the context of state actors intervening in media activities to slow scientific progress during the Cold War.
Mentioned for its impressive implementation of technology like gunshot detection and drones to significantly reduce crime, serving as an example of effective policing.
A venture capital firm that published an essay on Flock Safety, advocating for its role in solving crime.
A humanoid robot mentioned in the context of potentially working in Amazon factories and the debate around taxing such robots to retrain human workers.
SpaceX's large AI compute clusters, rented out to Anthropic for $1.25 billion per month as part of a $45 billion, three-year deal, forming 'Elon Web Services' (EWS).
SpaceX's heavy-lift launch vehicle, described as their 'workhorse' and used as a comparison for the larger Starship.
SpaceX's satellite internet constellation, described as a 'money printer' due to its $11.4 billion revenue and potential for hundreds of millions of subscribers, and its role as a revenue driver for SpaceX.
NVIDIA's high-performance GPU, with a working version currently in space, used for training and inference, and expected to have a space-designed version with different heat sink and reinforcement.
An AI chip not currently being submitted for benchmarks, suggesting it might underperform against competitors like NVIDIA's GPUs.
SpaceX's fully reusable super heavy-lift launch vehicle and spacecraft, capable of fitting hundreds of people, designed for rapid reusability and essential for ambitious goals like Mars colonization and orbital compute.
Google's Tensor Processing Units, mentioned in the narrative that NVIDIA might be losing market share to them, though NVIDIA's growth rates suggest otherwise.
A country where retail investors are borrowing record amounts to trade in AI chip stocks, drawing parallels to its peak crypto betting, raising concerns about market overheating.
A country whose nuclear secrets were leaked after the Manhattan Project, and whose relationship with China is framed as a strategic alliance after US-China talks.
A country whose 30-year bond yield hit an all-time high of 5.1%, indicating global bond market stress.
A strategically important waterway for oil transport, whose potential closure is seen as terrible globally but relatively beneficial for the US due to its energy self-sufficiency, forcing re-industrialization.
A geopolitical flashpoint, mentioned in the context of potential US-China agreements or 'handoff deals' under a Trump administration.
A country whose bond yields are at their highest since 2011, signaling broader economic challenges.
A geopolitical rival to the US, discussed regarding its AI advancement, potential for a CCP-funded campaign against US AI, and recent diplomatic meetings with US tech CEOs and President Xi Jinping.
An oil-producing country mentioned as a source of oil for China, which could be 'off the chessboard' if China acts against US interests, particularly regarding Taiwan.
More from All-In Podcast
View all 410 summaries
77 minTrump-Xi Summit, Benioff: "Not My First SaaSpocalypse," OpenAI vs Apple, Multi-Sensory AI, El Niño
96 minHow We Grew Koch Inc. to $150 Billion Without Going Public: Charles & Chase Koch
72 minSpencer Pratt on Fixing LA: Wildfires, Homelessness, Corruption & the Fight to Take It Back
83 minElon’s Anthropic Deal, The Next AI Monopoly?, “FDA for AI” Panic, Trading the AI Boom
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