The AI Productivity J-Curve: Why the Economic “Harvest” Is Still to Come

by Koo Ping Shung

Many may know I studied Economics during my undergrad years. It was a topic close to me and trained up my model thinking, and mathematics which is transferable to the study of Artificial Intelligence. With a idea spark from my co-trainer, mentor and good friend from Eric Sandosham, I decide to re-kindle my study of Economics, this time at the intersection with Artificial Intelligence. This learning and research is possible with the help of Google Gemini. :)

This is my first article, on the intersection between Economics and AI. Please share your feedback after reading! :)

Introduction

Have you ever wondered why, despite the daily headlines about generative AI’s brilliance, national productivity numbers still look… well, flat?

This isn't a failure of the technology. It’s possibly a phenomenon economists call the Productivity J-Curve. As we navigate 2026, understanding this curve is the difference between seeing a "bubble" and seeing the most significant economic restructuring of our lifetime.

1. The Origin: Solving the "Solow Paradox"

In 1987, Nobel laureate Robert Solow famously quipped, "You can see the computer age everywhere but in the productivity statistics." This became known as the Solow Paradox.

Decades later, economists Erik Brynjolfsson, Daniel Rock, and Chad Syverson formalized a solution: the Productivity J-Curve. They argued that when a "General Purpose Technology" (GPT) like AI arrives, it creates a temporary dip in measured productivity.

Why? Because companies are pouring massive capital into intangible assets—restructuring workflows, retraining staff, and developing new business models. These investments aren't totally captured in traditional GDP math today, but they are the "coiled spring" that powers growth tomorrow and thereafter.

2. History Repeats: From Steam to Silicon

We have seen this "dip and surge" pattern with every major technological leap:

  • The Steam Engine (1700s): It took 40 years for steam to move British GDP. The economy had to wait for the invention of the factory system and the railway network.
  • Electricity (1890s–1920s): Electricity entered factories in the 1890s, but productivity actually slowed down. It wasn't until managers threw out the "one big motor" blueprint and redesigned factories for individual electric motors—creating the assembly line—that the J-curve turned upward.
  • The Computer (1970s–1990s): We spent twenty years buying PCs before the "Productivity Miracle" of 1995 arrived. The missing link? The Internet and relational databases that finally allowed those PCs to talk to each other.

3. The Current AI "Dip"

Today, we are at the bottom of the "J." Organizations, especially the Big Tech, are borrowing and spending billions on GPUs and AI talent, but they are undergoing painful, slow restructuring at the same time.

Business owners on AI adoption is thinking of this daily: companies aren't just "buying a chatbot"; they are redefining what a "worker" should be doing, together with Artificial Intelligence. This period of "unmeasured investment" is why we haven't seen the 4% GDP growth yet—but history seems to promises it is coming, if the Productivity J-Curve is anything to go with.

4. The 4 Cs: The Great AI Divide

While the J-curve is a historical certainty, the height of the upward swing is not guaranteed for everyone. There is a widening gap between those who can ride the curve and those who will be left behind. This gap is defined by the 4 Cs Framework:

  • Connectivity: High-speed infrastructure and stable energy to run AI systems.
  • Compute: Access to the chips and data centers that power the models.
  • Context: The availability of local, relevant data that makes AI useful for specific regions or languages.
  • Competency: A workforce with the digital literacy and specialized skills to actually use AI (what I call the "AI-skilled human-in-the-loop").

5. The Future Trajectory: Two Worlds

If we continue on our current path, the 4 Cs will create a "Great Divergence." And even redefine the definition of "Developed" and "Developing", if we still want to stick to that labelling.

  • Developed Nations with high 4C scores will accelerate up the J-curve, reaping massive GDP gains as they automate cognitive tasks.
  • Developing Nations without the 4 Cs risk falling into a "stagnation trap," where their traditional labor advantage is eroded by AI-driven efficiency in the West.

Imagine a world where the "AI haves" see a 7% annual GDP boost, while the "AI have-nots" struggle with achieving 1% growth. That is the trajectory we must correct.

6. A Call to Action for Governments

The J-curve is inevitable, but its benefits are not. To my colleagues in policy and governance: the time to invest in the 4 Cs is not "later"—it is now. Wait-and-see is no longer a viable strategy. Governments must treat Compute and Competency as critical national infrastructure, much like roads and electricity in the 20th century. We need policies that don't just regulate AI risks, but actively build the 4C foundations that allow every citizen to participate in the "harvest" phase of the J-curve.

Let's stop asking when the AI productivity will show up and start building the foundations that will make it happen.

Thoughts? Share them with me on LinkedIn!