Berkeley and London study finds developer productivity gains jump 32% in market valuations before official statistics

When economist Robert Solow delivered his famous 1987 remark about the computer age, noting that its impact was visible everywhere except in productivity statistics, he was describing a time lag that took years before the information revolution appeared in official figures. Today, as annual AI investments approach one trillion dollars less than four years after the launch of large language models, economists are seeing the same pattern as they look for concrete evidence of employee efficiency within firms, yet when we review financial market records and job postings we find that the signal was captured early, revealing that the software engineering sector has become the first arena where these gains are translated into explicit numbers.

A research paper authored by Alex Bloomfield of the University of California, Berkeley, and Jonathan Hazel of the London School of Economics, with other economists, set out to measure this effect by comparing company stock returns to an AI-focused index between November 2022 and December 2025. The authors linked returns that exceeded or fell short of expectations to the magnitude of firms’ spending on programmer salaries after isolating other influencing factors.The paper concluded that artificial intelligence raised market estimates of the present value of software engineers’ productivity by a sustained increase of 32.6 %.It represents a qualitative jump that reveals how investors price the future return of application and system development.

The study’s findings are based on examining companies for which software is a core driver of their business model without being the sold commodity, such as the e-commerce platform eBay, the electronics manufacturer Sonos, and the travel platforms Expedia and Airbnb. To isolate the productivity factor precisely, the study excluded software-selling firms and semiconductor-chain suppliers. The same pattern recurs in an independent analysis that included consumer giants in Asia such as Kobang, a Korean consumer giant, and Garab, a Southeast Asian consumer giant, when comparing their performance to the Robo Global AI Index and using the proportion of developer vacancies as an alternative wage indicator, while Sony’s case reflected an additional complexity that may cause its true reliance on software engineering to exceed the initial estimate of 12%.

These figures reset the historic phrase coined by investor Mark Andreessen in 2011 when he declared that software is eating the world, as technology-linked companies now account for roughly 95 % of the market value of the ten largest global firms. However, writing code is undergoing a fundamental shift that ends the era of humans monopolising every line of code, after Andreessen explained that delegating 90 % of code writing to AI agents gives engineers the ability to multiply their output by tenfold, moving the workload from routine typing to design, supervision and the development of complex pathways.

This shift in pricing developer productivity puts tech firms and development teams in the Gulf, Egypt and the Levant facing entirely different calculations for project costs and implementation speed.In work environments where banks, fintech companies and regional logistics services race to digitize their internal infrastructure, managing teams based on the number of programmers or traditional work hours becomes an inadequate metric. The constant rise in engineer efficiency by more than thirty percent means that a team that employs agents in the coding cycle completes its tasks with a smaller size and lower cost, while institutions that cling to manual pathways bear higher operating bills and a persistent delay in launching their competitive features. If you want to measure the efficiency of your software architecture today, look at the rate of agent integration rather than the volume of hand-written code.

We see at the end of this scene that the gap in official government statistics does not indicate an absence of gains, but confirms that structural shifts begin in development rooms before being reflected in aggregate indicators. Financial markets have begun to punish slowness and reward companies that reengineer their software teams to accommodate the capabilities of generative models, confirming that the impact of artificial intelligence is no longer dependent on theory but has become a productive reality that precedes official records by years.