US-based software developers are the world's most prolific users of AI coding assistants, and are the trends that researchers believe have the economic implications of the country.
The researcher quartet explores American coders who prefer informative bots in preprint paper that analyzed 80 million code submissions to GitHub between 2018 and 2024.
Researchers – Simone Daniotti, Johannes Wax, Xiangnan Feng, and Frank Neffke have devised a machine learning model to analyze GitHub submissions, discovering 30.1% of Python features sold in the US that were submitted to Github in 2024.
Germany came next with 24.3% ahead of France (23.2 per cent), India (21.6 per cent), Russia (15.4 per cent) and China (11.7 per cent).
The paper claims that when developers use AI in 30% of their code, they will increase by 2.4% per quarter.
“Combining this effect with occupational challenges and wage data will result in an annual value of between $96-14.4 billion in the US,” the author argues.
This estimate is consistent with Microsoft CEO Satya Nadella's claim that around 30% of Microsoft code is written by AI.
The potential economic benefits arising from AI-enhanced commit rates could be even higher, as the authors found a 26% productivity boost in September last year, when productivity bills from other AI surveys are used.
Based on estimates of task completion times from three different randomized controlled trials that found productivity gains of 16.5%, 6.3%, and 26%, respectively, the researchers conclude that 30% of AI use leads to productivity lifts worth between $64 billion and $96 billion per year.
The authors acknowledge that their estimates are limited. For example, it should be noted that focusing on submitting Github codes can lead to miss out on something made by the popular Gitee in China. And they say they don't take into account “the potential reduction in the value of coding tasks due to the additional supply of code through AI.”
Assume that there are other factors that could distort the author's results, such as treating Python as a representative of the impact on software development of other languages, and that AI usage in GitHub's open source projects is repeated in other settings.
But overall, the author says he is bullish on the productivity value of AI. Furthermore, adoption of AI will lead to an increase in experimentation with new software libraries and libraries combinations, leading to an expanded knowledge of developers. It assumes that those libraries actually existed and that they were not dreaming by AI.
Apart from writing code, the economic impact of AI can be more modest. In a paper published last year, Daron Acemoglu, professor at MIT Institute, “The Simple Macroeconomics of AI,” predicted only about 0.7% of AI-driven productivity gains. ®
