newsfilter
← Latest News
Earnings Call Transcripts·

All-In Podcast - All-In Podcast

ALLN

  • Explicit Guidance:

    • Anthropic Revenue & Margins: Q3 2024 net revenue of $16.8B (approx. $5.6B/month); Adjusted operating margin of 11%; Adjusted gross margin of 85%; Net revenue retention of 900%.
    • Anthropic Revenue Projections: ARR (Annual Recurring Revenue) estimates range from $70B to >$100B, though actuals undisclosed; Q3 growth rate noted as slowed compared to Q2.
    • Anthropic IPO Financing Needs: Chamath Palihapitiya recommends raising $200B minimum (ideally $250B) to survive potential corrections, suggesting a lower valuation (e.g., $1.5T) to clear the market rather than a higher valuation (e.g., $2T+) for less capital.
    • CoreWeave Debt Constraints: Current average interest rate on financing is ~8.5%; bond yields spiked to 15%, implying future financing costs could rise into the 10–15% range.
    • CoreWeave Break-even Risk: Management/Analysts estimate CoreWeave is currently within +/- $5B of a covenant break-even point if financing rates rise and spot prices do not cover obligations.
    • AI Sector Growth Rates: AI startups are reported to be doubling revenue every quarter; Anthropic's revenue growth is described as "violent" with 16.8B in a single quarter for a three-year-old entity.

  • Business-Segment Outlook:

    • GPU Economics: GPU lifespan (duration) expected to extend from 3–5 years to 6–8 years for inference workloads; older GPUs (H100/A100) continue to rent at premium prices despite depreciation schedules.
    • Spot vs. Contract Pricing: Spot GPU pricing remains elevated and expected to stay high; CoreWeave's fixed-price contracts (2.5-year payback vs. 5-year term) create a widening arbitrage gap with spot markets (potential $20B–$30B upside if they switch to spot).
    • Data Center Construction: Severe shortage of power, cooling, and physical labor; data center construction timelines extending beyond 2028; supply chain constraints (packaging, HBM) dominated by NVIDIA.
    • Compute Duopoly: Anthropic and OpenAI projected to capture ~60% of all compute expected to be deployed globally over the next 12 months, establishing a compute duopoly.
    • Small Language Models (SLMs): Growing enterprise adoption of small, specialized, open-source models for specific workloads (e.g., fraud detection) due to efficiency and lower compute costs compared to frontier models.
    • CoreWeave Financing Structure: Business model relies on "boxes" (independent asset-backed vehicles) with 75% of revenue from three investment-grade customers; risk of contagion is mitigated by isolated structures but reliant on tenant creditworthiness.

  • Market and Competitive Outlook:

    • AI Market Dynamics: Demand for compute is "off the charts" and supply-constrained; no glut of supply expected in the near term due to physical construction limits and regulatory hurdles.
    • Software Economics: Software prices expected to decline due to commodification and open-source alternatives (Jevons Paradox); however, total software volume and bespoke creation will surge, driving productivity.
    • Competitive Landscape: Anthropic and OpenAI viewed as threats to legacy SaaS, healthcare, and workflow businesses; Meta (Muse) and Google models approaching but not yet leading in frontier capability.
    • IPO Market: Strong demand anticipated for Anthropic and OpenAI IPOs; no "lack of money" in the system to finance these valuations (trillions).
    • Open Source vs. Closed: Tension expected to increase as open-source models improve; Anthropic accused of fear-mongering to stifle competition and push for regulation against open weights.

  • Macro / Industry Outlook:

    • Interest Rates: Risk of 10-year Treasury yields rising to 5.5%–6.0% driven by deficit spending and debt-to-GDP; this would severely impact businesses relying on financing spreads.
    • Energy Supply: US faces a massive energy deficit (~1.4 Terawatt-hours by 2050, equivalent to six Californias); energy availability is identified as the primary bottleneck ("moat") for AI infrastructure.
    • Geopolitics: Conflict in the Middle East cited as a headwind for global data center development due to security risks and inability to secure sites.
    • Regulatory Environment: Concerns over potential government intervention to restrict open-source models; prediction that regulatory capture attempts will fail as companies must manage technical risks themselves (KYC/technical defenses).
    • Deflationary Pressure: AI expected to be broadly deflationary for software and productivity, potentially removing 90% of value from certain tasks, though total economic output is expected to rise.

  • Strategic Initiatives:

    • CoreWeave Restructuring: Potential recommendation to cancel long-term fixed contracts in favor of spot pricing to capture higher arbitrage and improve liquidity, despite covenant risks.
    • Anthropic IPO Preparation: Planning for a public listing around Thanksgiving 2024; strategic need to raise significant capital ($200B+) to secure compute capacity (5–10 gigawatts) and avoid down rounds during market corrections.
    • Open Source Deployment: Trend toward enterprises building bespoke software and in-house models rather than purchasing SaaS; "everyone is now a developer" with 95% of code assisted by AI.
    • Security & Moats: Shift in competitive advantage from software code to physical assets (land, power, infrastructure) and data moats (feedback loops from vertical-specific usage).
    • Governance Changes: Call for OpenAI CEO Sam Altman to receive a normalized equity grant (2–4% or $20–$40B) to fix governance conflicts and align with founder incentives.

  • Risks and Headwinds to Outlook:

    • Financing Spreads: CoreWeave and similar neoclouds face existential risk if risk-free rates rise, eroding margins between fixed revenue and variable debt costs; potential 15% chance of default in any given year per credit markets.
    • Spot Price Volatility: If spot GPU prices drop or demand subsides, payback periods for data centers will lengthen, magnifying debt risks.
    • Regulatory & Political Headwinds: Potential for local communities to block data center entitlements; political pressure to regulate AI safety and open weights.
    • Operational Complexity: Data center cooling and power density increasing complexity; risk of losing investment-grade wrappers from hyperscalers (Microsoft, Google) if service level agreements are missed.
    • Open Source Competition: Rapid commodification of software value; Anthropic accused of attempting to use "distillation" and "worm" narratives to stifle open-source competitors.
    • Security Risks: Hypothical risks of AI-powered worms on IoT devices; actual risks of existing non-AI worms (e.g., Mirai, Kim Wolf) exploiting unpatched hardware.
    • Alignment Risks: Concerns that Anthropic's training methodology (Claude Constitution iterations) may induce misalignment and model self-agency, creating a "Frankenstein" risk.
    • Bill Gates Narrative: Perception that doomerism promoted by Bill Gates is a strategic play for influence and power rather than a genuine risk assessment.

  • Management Tone on the Future:

    • Confidence Level: High confidence in long-term AI demand and the viability of Anthropic/OpenAI, despite near-term volatility.
    • Market Volatility: Views CoreWeave stock/bond selloff as "volatility" rather than a fundamental bubble pop; compares current AI dynamics to the 1999 internet bubble but notes lack of demand is not the issue (contrast: AI has excess demand, 1999 had excess supply).
    • Resilience: Belief that as long as payback periods remain short (1–2.5 years), companies can withstand market corrections; nervousness increases only if payback periods lengthen.
    • Skepticism: Deep skepticism regarding Anthropic's "AI worm" and "distillation" scare campaigns, viewing them as "psyops" to justify regulatory capture and protect market share.
    • Crisis Management: Suggests CoreWeave should proactively manage risk by switching to spot pricing and canceling contracts to avoid covenant breaches.