SpaceX’s $2T Case, Nvidia’s Shock Selloff, America Turns on AI, Trump Pulls AI Order, Bond Crisis?

Andrej Karpathy and Anthropics Strategic Shift

• Andrej Karpathy joined Anthropic as the leader of a new pre-training team focused on recursive self-improvement, aiming to enable the model to improve itself during training.
• Recursive Self-Improvement is viewed as a critical "holy grail" frontier that could unlock an order-of-magnitude annual improvement in model quality, effectively creating a new form of Moore's Law.
• Continual Learning remains a distinct challenge where models must learn from experiences in real-time like humans, which has not yet been fully unlocked.
• Anthropic's Financials reached an estimated $100 billion ARR with ~80% gross margins on inference, signaling that AI companies have achieved tangible positive returns.
• Future ARR Projections suggest the private AI sector could reach $200-$400 billion in ARR by year-end when accounting for competitors like Gemini, Cursor, and xAI.
• Architectural Innovation is shifting toward networks of smaller, specialized models working together to reduce energy costs per token compared to single monolithic models.
• Google's Stealth Deployment of the "Gemini Nano" model (4GB) directly into Chrome OS for local tasks sparked privacy concerns, though discussed as a utility-focused move rather than a security threat.
• End-Use Case Focus is deemed more valuable than model-centric narratives; examples include a hedge fund manager using AI to identify a safe drug to treat his daughter's rare genetic mutation, increasing her neuron firing from 30% to 90%.

AI Narrative, PR, and Regulatory Landscape

• Public Backlash against AI is driven by fears of power concentration, the asymmetry of technology diffusion, and the perception that benefits accrue to a small elite while the broader population faces job displacement.
• Geopolitical Influence includes accusations of coordinated anti-AI sentiment campaigns by foreign actors, though the primary driver is also deep-seated psychological shifts regarding human centrality in the universe.
• Dario Amodei's Strategy of warning about job loss is interpreted by Gavin Baker as a rational business tactic to establish a "regulatory moat" to protect Anthropic's market position.
• Leadership Communication Failure is exemplified by Cloudflare CEO Matthew Prince's internal memo labeling laid-off employees as "measurers" (managers who measure data), which was widely criticized as tone-deaf.
• Mark Zuckerberg's AI Strategy involves laying off hundreds of contractors and employees to harvest their work for training data, explicitly stating that training models to replace human tasks is a primary goal.
• Regulatory Skepticism exists regarding government oversight of frontier models; the preferred approach involves self-regulation, liability courts, and international KYC (Know Your Customer) verification rather than US-centric federal mandates.
• Self-Driving Adoption faces resistance in specific municipalities (e.g., Cambridge, Boston, NYC) where voters prioritize human employment over safety and convenience, though the long-term trend favors autonomous vehicles due to the 50,000+ annual traffic deaths in the US.
• Crime Prevention AI (e.g., Flock Safety) is presented as a municipal choice where communities can opt-in to gunshot detection to reduce crime, with Las Vegas cited as a highly successful case study of low-cost, high-impact deployment.
• Tech CEO PR Deficits are identified as a major vulnerability, with leaders praised for technical execution but criticized for poor communication regarding the societal impacts of AI and layoffs.

SpaceX IPO and "Elon Web Services"

• SpaceX S1 Filing targets a $75 billion raise at a $1.75 trillion valuation, positioning it as the largest IPO in history, with a listing expected in mid-June under the ticker "SPCX."
• Elon Web Services (EWS) revenue is the dominant growth driver, with a $1.25 billion/month ($15 billion/year) contract with Anthropic to rent "Colossus" data centers, potentially quadrupling SpaceX's AI division.
• Data Center Construction Speed has improved dramatically, with build times dropping from 122 days to 66 days, enabling rapid scaling of gigawatt-scale facilities.
• Cursor AI Breakthrough saw its Composer 2.5 model achieve "Pareto dominance" (superior performance at lower cost) in just three weeks using reinforcement learning on Colossus II, leveraging Cursor's proprietary coding token dataset.
• Grok Build launches as a new AI "harness" for the xAI ecosystem, providing an environment for state management and memory that integrates with tools like Slack and Gmail, crucial for agentic AI workflows.
• Orbital Compute is validated as technically feasible by the deployment of working H100 GPUs in space, with space-optimized versions of hardware expected in the 2028-2030 timeframe.
• Starship Reusability remains the critical engineering hurdle; achieving "rapid reusability" (multiple flights per day) is necessary for mass driver and colony economics but is not yet fully operational.
• Valuation Thesis for SpaceX includes buying into the most significant internet infrastructure project since the internet itself, AI compute dominance, and the potential for space-based backup infrastructure independent of terrestrial government control.

Nvidia Earnings and Market Dynamics

• Nvidia Q1 Results showed $81.6 billion in revenue (up 85% YoY) and $58 billion in net income, maintaining a $5.3 trillion market cap and 75% gross margins.
• Shareholder Returns include an $80 billion buyback program and a 25x dividend increase to 25 cents per share, with a commitment to return 50% of free cash flow to shareholders.
• Sector Valuation Divergence exists between Nvidia (reasonable multiples) and memory/power/cooling stocks (discounted multiples), suggesting the market is pricing in different risks and growth trajectories across the AI supply chain.
• Nvidia CPU Business is projected to reach $20 billion in revenue for the year, driven by "co-design" partnerships with every major lab to architect domain-specific accelerators.
• Asset Lifespan Reality contradicts bearish narratives; financing contracts for data centers (e.g., CoreWeave) span 6+ years, proving GPUs and AI hardware have useful lives of 10-15 years when mixed with domain-specific accelerators.
• Jensen Huang's Frustration stems from competitors releasing ASICs that refuse to undergo independent benchmarking, making it difficult to verify claims of Nvidia losing market share.

Macro Economics and Geopolitics

• Inflation and Rates are rising due to geopolitical tensions in the Strait of Hormuz, with 10-year US Treasury yields hitting 4.6% and global yields reaching multi-year highs.
• Federal Reserve Outlook has shifted from rate cuts to potential rate increases as inflation forecasts (CPI) are revised upward to 4-6% for Q2.
• Global Debt Risks are flagged at 310% of GDP, with concerns that currency devaluation and money printing could trigger a credit crisis as debt servicing becomes unsustainable.
• US Strategic Advantage is highlighted by energy self-sufficiency and a re-industrialization boom; while global straits closures hurt allies, they relatively benefit the US as the world's largest oil exporter.
• China-US Trade Talks resulted in limited concrete deals (soybeans, aircraft, deprecated GPUs) but no "grand bargain," with relations likely remaining competitive and tense.
• Geopolitical Chessboard analysis suggests the US may leverage oil supply control (cutting off Venezuelan, Iranian, and potentially Russian oil access) to deter conflict in Taiwan and the Middle East.
• Market Strategy for the current volatility involves focusing on concentrated holdings of high-conviction companies for long-term growth rather than speculative trading in the face of macro instability.