Satya Nadella defines a transformational technology as one that delivers 5-10% GDP growth above inflation-adjusted GDP in a Feb 2025 interview.
Jul 2026, J.P. Morgan forecasts US GDP growth of 1.5% – 2%. Last year was 2.1% as well.
So how transformative is AI in creating wealth?
We still haven’t seen a meaningful contribution from large language models to economic growth.
> Railroads contributed to 25% of GDP growth between 1830 and 1860.
> WWII mobilization reported 17%+ in 1941, 1942, 1943.
Even in times of military action abroad, GDP progresses alongside inflation, or even lower.
AI is an operational efficiency framework that differentiates from previous booms in several ways:
1. GDP growth is traditionally tied to the PRODUCTION of goods, logistics and transportation, and local services (but rarely digital solutions).
2. Despite disappearing jobs, new roles were created fast, and human labor was still poured into new industrialized jobs, with better pay and opportunities at scale.
3. Previous transformations had local moat. Building railroads, the steam engine, cars, electricity, telecommunication took years or decades for other countries and continents to catch up. This competitive advantage increased production and export because of this leverage.
4. AI is largely available and accessible globally. This isn’t the definition of moat if ChatGPT is used across the board, and so are Gemini, Claude, Grok, Mistral, and all open source models.
5. Demand isn’t equal:
a. Railroads provided opportunities to build cities, warehouses, move military goods, facilitate trade between locations. It unlocked plans that took a decade possible in 2 years, and demand was able to accommodate the increased production.
b. Cheaper content, websites, posts hit a finite number of views, impressions, clicks since consumer behavior remains the same.
6. Rapid GDP growth requires massive increases in labor productivity, workforce expansion, and heavy capital or infrastructure investment. AI shortens workforces and undercuts projects, resulting in a zero sum workflow.
7. Existing recurring supply remains largely the same and a finite number of operations shifts between providers, often into a “race to the bottom” equation.
Robots have a better chance to lead to a true industrial revolution if one or two countries manage to build a fleet fast, increase productivity significantly, and grow manufacturing + trade globally before everyone catches up.
AI doesn’t demonstrate a clear path to GPT growth in the coming 5 years.
