CoinTelegraph
Coinbase CEO Brian Armstrong sees AI demand as "near infinite" but expects a dramatic cost collapse to reshape how that demand is served — with 80% of workloads migrating to models that cost 99% less within the next 12 to 18 months.
Why it matters
Armstrong's framing cuts against the prevailing narrative that AI infrastructure spend will compound indefinitely at current price points. If the bulk of enterprise and consumer workloads shift to radically cheaper inference models, the economics of AI compute change fundamentally — with implications for data centre capex, GPU demand, and the energy grids being built to support them. The CEO's explicit callout of energy and compute as the binding constraints signals that the bottleneck is physical, not algorithmic.
Market impact
For crypto and fintech investors, the signal is layered: cheaper AI models lower the cost of building AI-native financial products, which is net positive for platforms like Coinbase that are embedding AI into their stack. At the same time, Armstrong's energy constraint thesis aligns with a broader macro trade around power infrastructure and compute scarcity — two themes that have driven institutional interest in Bitcoin mining equities and energy-adjacent tokens over the past year.
CoinDesk
TheBlock
WuBlockchain