3.54% annual productivity growth would erase Japan's IT talent shortage

Discussions of Japan's IT talent shortage usually cite a shortfall of up to 790,000 people by 2030. That figure comes from a METI survey published in April 2019.

Supply reaches 1.13 million by 2030. The gap against demand is 450,000 in the mid scenario and 790,000 in the high scenario.

That number is quoted everywhere. A different line, on page 3 of the same report, is quoted almost nowhere.

At 3.54% annual labour productivity growth, supply and demand balance.

What 3.54% represents

The model's baseline is 0.7% per year. So 3.54% is roughly five times the assumed rate.

Five times sounds unreachable. But it is not a claim about the whole economy. It concerns the productivity of IT personnel specifically.

And the report was written in 2019, before generative AI entered professional work.

Set against the current numbers

The 2026 White Paper on Information and Communications puts Japanese usage of generative AI for code generation at 7.0%, against 35.4% in the United States.

IPA's DX Trends 2025 reports that 3.8% of Japanese companies consider their DX talent adequate by quality, against 52.9% in the US. More than 80% report a shortage, and 19.4% conduct no hiring at all.

Line those up and the situation resolves. There are not enough people. Hiring is not happening. And the lever that raises output per person is in use at 7.0%.

The reason 3.54% matters is that METI's own projection shows the talent shortage is not a problem only solvable through recruitment.

How to use the number

This is not a proposal to plan for 3.54% company wide. The value of the figure is that it changes the frame of the internal discussion.

Talent shortage conversations become recruitment conversations. Recruitment conversations become budget and market conversations, and usually stop there. As covered earlier, they also stall because the requirement cannot be written.

Reframed as productivity, the conversation moves back inside the company. Inside the company, things can be decided.

Concretely it becomes these questions.

• How much time does the current team spend building things that should not be built
• How much time is spent waiting on approvals and rework
• Of the time spent on routine implementation, what share could be automated

All three are measurable. The total is frequently larger than expected.

790,000 invites a conversation about hiring. 3.54% invites a conversation about design. The second one covers more ground you actually control.

The limits of the figure

The projection is from 2019. The technology assumptions have changed and so have the demand side estimates. It cannot be applied directly to today.

The line still has value, because it formalises a talent shortage as a productivity problem. A government projection is showing that the same problem has a solution in a different variable.

And unlike recruitment, raising productivity does not depend on an external market.