@fchollet
Post
The most reliably predictable trend of the next 100 years: every year, humanity will use significantly more computing power than the previous year.
Someone should start an ETF based on that thesis (it's not just $NVDA and $AMD, it's cloud services, the data center industry, nuclear power...)
Explanation
What it says François Chollet’s thesis is that the safest century-scale technology bet is relentlessly rising global compute consumption: each year, humanity will use materially more computing power than the year before. He suggests expressing that thesis through a broad investment basket—not just GPU makers like Nvidia and AMD, but cloud infrastructure, data centers, and nuclear power.
Context This is a macro investment idea, not a detailed forecast. The implicit chain is: more software/AI/digital activity → more computation → more chips, servers, electricity, cooling, networking, and physical infrastructure. No evidence, valuation framework, ETF construction, or distinction between compute demand and investor returns is provided.
Why it matters The useful insight is broader than “buy AI stocks”: if compute demand compounds for decades, bottlenecks and value capture may migrate across the entire compute supply chain, especially toward scarce infrastructure and energy. But rising compute usage does not imply every supplier earns excess returns; competition, efficiency gains, regulation, overbuilding, and valuation can overwhelm demand growth. The post is best saved as a durable thematic lens, not an actionable portfolio thesis.