Market Lab Report - AI Productivity: Why the Boom Is Coming (Even If GDP Looks Slow Right Now)

## Quick Summary
U.S. GDP growth in Q4 2025 was only 0.7%—much lower than the expected 1.4% and way below the 3%+ numbers from earlier quarters. People are asking: “Where are the AI productivity gains everyone promised?”

The answer: AI is improving extremely fast (capabilities roughly doubling every 3–7 months), much quicker than the internet did in the 1990s. That means the big productivity jump—and higher GDP—is still coming. It just takes time for companies to use the new technology well. The internet's productivity surge finally arrived around the late 1990s (roughly 1995–2005), even continuing after the dot-com stock bubble burst in 2000–2002. AI could do it faster because it’s advancing so quickly.

## Why GDP Looked Weak in Q4 2025
- A federal government shutdown cut spending sharply → subtracted about 1% from growth.
- Exports fell, consumer spending slowed, and high interest rates plus new tariffs hurt.
- Full-year 2025 GDP growth was 2.2%—decent, but not exciting.

AI spending (data centers, chips, etc.) helped a little, but heavy imports of equipment meant the net boost to GDP was small.

## AI Is Getting Better Much Faster Than the Internet Did
- AI task performance (how long a task AI can handle on its own) doubles every 3–7 months (source: METR research).
- In recent years it has accelerated, with doubling around every 7 months over the past 6 years, and faster in some periods.
- Example: Today’s best models can do 50-minute human tasks with 50% success. A few years ago, they could only do minutes-long tasks.

This speed means AI could start automating much bigger projects (month-long software work) within a few years.

## The 1990s Internet Comparison
- In the 1990s, computers and the internet were everywhere, but productivity barely moved for years (the “Solow paradox”).
- The real productivity boom finally arrived around the late 1990s (roughly 1995–2005), with growth rates doubling even as the dot-com stock bubble burst in 2000–2002.
- AI is following a similar pattern: early excitement, then quiet progress, then a big lift.

Key differences that favor AI:
- 70–78% of companies already use AI tools.
- ChatGPT reached 100 million users in months (internet took years to grow that fast).
- AI helps all kinds of knowledge work, not just consumer internet.

## Why the Productivity Gains Might Take Longer (Strong Counterarguments)
Even though AI improves quickly, several real barriers can slow the economy-wide payoff:
- Measuring productivity is hard—especially for “knowledge work” like writing code or making decisions.
- Companies need to retrain people, change processes, and buy new tools → this takes years.
- Early AI use can sometimes make workers slower (debugging mistakes) before it makes them faster.
- If too much money goes into overhyped projects (like too many data centers), some investments may not pay off quickly.
- Gains might stay in tech companies and not spread to the whole economy right away.

These are legitimate reasons for delay, but they don’t stop the long-term trend—they just stretch the timeline.

## What Could Happen Next
- Short term (2026): AI adds a small boost (0.1–0.2% to GDP) mostly from building data centers.
- Long term: PwC forecasts AI could boost global GDP by an additional 15 percentage points by 2035 (in an optimistic scenario). McKinsey sees $2.6–4.4 trillion added annually globally by the 2030s. If the doubling trend continues, the productivity surge could be bigger and faster than the 1990s internet boom.

## Bottom Line
The weak Q4 GDP number doesn’t mean AI is failing. It means we’re still in the early stage—building infrastructure and learning how to use it. With AI getting twice as capable every few months, the productivity explosion is coming. When it hits, GDP should rise sharply.

In the meantime, major averages and leading stocks will telegraph through price/volume action and constructive chart patterns when the market has reached a major low. As always, we continue to send actionable entry points (long or short) to members.

**Further Reading**  
- METR on AI task doubling (2025 blog): https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks  
- PwC AI economic impact forecast (2025 update): https://www.pwc.com/gx/en/news-room/press-releases/2025/ai-adoption-could-boost-global-gdp-by-an-additional-15-percentage.html  
- Brookings on 1990s productivity trends and the internet: https://www.brookings.edu/articles/the-internet-and-the-new-economy