Coinbase CEO’s Top 3 Crypto Trends for 2026 + More from Davos!

Summary

This All-In Davos episode combines interviews with Brian Armstrong of Coinbase, Andrew Feldman of Cerebras, and Jake Loosararian of Gecko Robotics. The crypto segment emphasizes regulated stablecoins, on-chain assets, prediction markets, and capital formation; the AI segment treats inference speed, memory, power, cooling, and chip geopolitics as system constraints; and the robotics segment argues that paid industrial inspection can create the proprietary physical-world data needed for useful industrial AI. Across all three interviews, the shared frame is a shift from technology narratives toward deployment, regulation, infrastructure, and measurable ROI.

Key Claims

• Armstrong identifies on-chain assets, prediction markets, and stablecoin payments as the three leading crypto trends, while presenting the GENIUS Act and bank adoption as evidence of a more legible U.S. regulatory path.
• USDC is presented as the largest regulated stablecoin, while Tether is credited with real emerging-market utility but described as not compliant with the current U.S. framework.
• Stablecoins are framed as especially useful for B2B cross-border payments, where settlement delay and foreign-exchange fees create a concrete alternative-rail opportunity.
• Armstrong argues that tokenization can widen investor access and reduce back-office and settlement friction, but says private-company shares should be tokenized with company permission because employee liquidity, vesting, disclosure, and issuer control still matter.
• Feldman presents Cerebras as a wafer-scale inference system whose product value comes from reducing user-visible latency in deep research and coding, while power, memory, transport, cooling, and construction remain system bottlenecks.
• Feldman says data centers are increasingly measured by power rather than floor area and argues that developers need to protect local ratepayers and invest in host communities rather than treating power access as a purely technical problem.
• Loosararian argues that industrial robots create value by collecting otherwise missing data from ships, bridges, refineries, dams, and other assets, then using that data to extend asset life, increase output, and reduce hazardous human work.
• Gecko Robotics is described as moving from inspection and diagnosis toward monitored repair and manufacturing actions, including automated welding, while humans remain in the loop through supervision and teleoperation.

Key Quotes

“everything exchange” - Armstrong’s label for a future exchange spanning crypto, equities, and other on-chain assets.

“calculation speed, memory, and transport” - Feldman’s three-part computer-architecture frame.

“voice or exit” - Armstrong’s frame for trying to change California or helping builders relocate.

Connections

• Brian Armstrong, Coinbase, USDC, GENIUS Act, Stablecoins, and Real World Asset Tokenization Risk - regulated crypto, payment, and tokenization branch.
• Kalshi, Polymarket, Prediction Market Legal Boundary, and Cryptocurrency Market Structure - prediction-market and exchange-expansion branch.
• Andrew Feldman, Cerebras, Low-Latency Inference Chip, Memory Wall, and Data Center Power Bottleneck - inference architecture and physical infrastructure branch.
• OpenAI, Cognition, Data Center Thermal Management, and Electric Grid Modernization Bottleneck - customer, cooling, and grid context.
• Jake Loosararian, Gecko Robotics, Industrial Inspection Robotics, Defense Robotics Maintenance, and Physical World Data Flywheel - industrial inspection, defense, and field-data branch.
• NewLimit - Armstrong’s epigenetic-reprogramming company discussed as a separate biotech investment.

Contradictions

• No settled contradiction identified. Armstrong’s regulatory and political account, Feldman’s performance and demand claims, and Loosararian’s deployment and roadmap claims are participant perspectives rather than independent audits.
• The optimistic tokenization case is qualified by Real World Asset Tokenization Risk: faster distribution and settlement do not by themselves resolve asset quality, valuation, disclosure, issuer permission, or legal enforceability.
• The robotics roadmap extends rather than overturns the current Industrial Inspection Robotics synthesis: inspection is commercially nearer than autonomous repair, and human supervision remains necessary.