Interview, Fireside Chat
Groq Founder, Jonathan Ross: OpenAI & Anthropic Will Build Their Own Chips & Will NVIDIA Hit $10TRN
- Compute and Energy Nexus: The controlling of AI is contingent on controlling compute, which is inextricably linked to energy supply; countries with insufficient energy capacity cannot sustain the required compute infrastructure.
- NVIDIA Valuation Projection: Jonathan Ross predicts NVIDIA could reach a $10 trillion valuation within five years, driven by insatiable demand and a "temporal moat" where competitors face a 3–4 year development cycle versus Grok's 1-year cycle.
- Market Concentration Risk: Approximately 35 to 36 companies currently account for 99% of AI revenue or token spend, creating a "lumpy" market akin to early oil drilling where high-risk "vibe investing" yields disproportionate returns.
- Revenue-Compute Correlation: Ross asserts that if providers like OpenAI or Anthropic were granted double their current inference compute, their revenue would nearly double within one month due to existing rate limits and latency issues.
- Speed as a Product Driver: Product quality and brand affinity are directly correlated with response speed; a 100ms reduction in latency historically yields an 8% conversion rate increase in consumer-facing applications.
- Supply Chain Constraints: The primary bottleneck in the AI sector is not algorithm design but hardware availability; NVIDIA faces a monopsony on High Bandwidth Memory (HBM), creating a 2-year lead time for capacity expansion, whereas Grok offers a 6-month supply timeline.
- Chip Ownership Strategy: While hyperscalers are vertically integrating by building custom chips (e.g., TPU, custom silicon), Ross notes this is a defensive play to secure HBM supply and control over destiny rather than a purely efficiency-driven move, as replicating NVIDIA's engineering success is statistically improbable.
- Grok's Competitive Advantage: Grok has raised $750 million at a ~$7 billion valuation to expand LPU (Low Precision Unit) capacity, allowing it to offer 50x the speed of traditional GPU inference at a lower cost per token by optimizing system-level resource allocation.
- Economic Impact Forecast: Ross anticipates massive deflationary pressure from AI efficiency, leading to reduced work hours, earlier retirements, and the creation of entirely new job categories, rather than net job loss.
- Vibe Coding Evolution: The market for "vibe coding" (AI-generated code via natural language) is expected to mature into a standard skill comparable to literacy, replacing specialized programming roles with generalist AI-assisted development.
- Strategic Positioning: Unlike competitors who face "overheating" risks from rapid capital deployment, Grok aims to align margins with customer value, prioritizing volume and speed over short-term profitability to capture insatiable market demand.
- Open Source Strategy: Ross suggests open-sourcing previous-generation models (e.g., by Anthropic) to prevent users from adopting Chinese models, leveraging prompt compatibility as a retention strategy while competitors focus on proprietary, closed systems.
- European Energy Gap: Europe risks economic stagnation (reduced to a "tourist economy") if it fails to accelerate energy permitting; Ross cites Norway's potential to power the entire US with wind/hydro and urges hyperscalers to locate data centers near cheap renewable sources.
- China vs. US Dynamics: While China can subsidize high-energy inference domestically, the US maintains a strategic advantage in the "away game" (allies in Europe, India, Japan) due to superior chip energy efficiency.
- Market Valuation Logic: Ross distinguishes between "weighing machines" (value-driven AI firms like PE-backed compute providers) and "popularity contests" (hype-driven sectors), arguing current AI valuations reflect genuine productivity gains rather than a bubble.
- Future Hardware Cycles: Grok operates on a one-year silicon iteration cycle (V2, V3, V4), contrasting with NVIDIA's 3–4 year cycle, mitigating the risk of older hardware becoming obsolete before ROI is realized.
- Labor Shortage Prediction: The AI boom is predicted to create a net shortage of talent rather than a surplus, as automation increases productivity and creates new industries that cannot yet be defined.
- Grok's Business Model: Grok generates positive margins on hardware sales and selective software models, focusing on long-term cash flow and low volatility rather than maximizing short-term profit per unit.
- Investment Philosophy: Ross advises investors to be "greedy where there is a moat," warning that pre-moat startups (those without a competitive barrier) should be valued significantly lower, noting that most AI startups lack a defensible advantage against incumbents.
- Galileo Analogy: Ross concludes with a long-term positive outlook, comparing LLMs to the telescope: initially making humanity feel small, but ultimately revealing the vastness of intelligence and leading to a more beautiful understanding of the universe.