TL;DR
Get the latest gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
The Pragmatic Engineer surveyed engineering leaders and senior engineers at 27 workplaces about AI-tool spending and pull-request (PR) output. Separate data from CodeRabbit showed PR counts more than doubled among its measured developers, but the report found that companies generally could not say whether the spending produced a measurable return.
The Pragmatic Engineer surveyed engineering leaders and senior engineers at 27 workplaces about their spending on AI coding tools and changes in software pull-request output, finding that costs varied widely while many companies could not quantify the return. The report also cited platform data showing higher PR counts among developers using AI-related services, but those figures do not by themselves establish that AI caused the increase or improved business results.
In data shared by CodeRabbit, weekly PRs among its surveyed data engineers were up 113%, while monthly PRs rose 127%. A tracked group of 20,000 developers recorded a 128% increase over a year. The report described these as platform observations; it did not establish that the increases were solely attributable to AI tools or that every PR represented equivalent work.
Gitar, an AI code-verification agent, provided data showing different levels of PR output across customer teams. The report said some “AI-native” teams shipped more than 60 PRs per developer per week. Its author suggested team norms and company culture may help explain differences, rather than treating output as a direct measure of individual productivity.
Spending also varied substantially. The report described a two-person .NET business spending $3,000 per developer monthly, while a publicly traded company with 1,500 engineers capped monthly spend at $150 per developer. A high-spending trading firm reported total AI costs of $8 million in September, up from $500,000 in January, and average spending of $1,600 per developer per month. These are examples reported by the publication, not a representative industry-wide cost estimate.
Higher Output Does Not Prove ROI
The report highlights a measurement problem for engineering executives: tool bills and PR counts are comparatively easy to track, but neither tells a company whether AI spending has paid off. A higher PR count may reflect more small changes, different review practices, or changes in team workflow. It does not alone show better software, faster delivery of customer value, or lower total costs.
That distinction matters as businesses decide whether to expand access, impose spending limits, or move work to cheaper models. If subsidized subscriptions end, the report says some firms expect costs to rise by five to 10 times unless they optimize usage. Leaders will need to weigh those costs against outcomes such as development time, quality, maintenance burden, and delivery results—measures the report did not provide consistently.
AI coding tools for software engineers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How the Spending Data Was Gathered
The report began with a reader asking how much companies spend on AI per engineer and whether PR throughput changes before and after adoption. Its author said they spoke with engineering leaders and senior engineers at 27 workplaces, anonymizing most company identities at their request, and collected aggregate platform data from CodeRabbit and Gitar.
The publication found no apparent link between company size and per-engineer spend in this limited sample. It also reported that spending peaked for many companies in June and July and had since fallen roughly by half among some API-plan users. The author attributed much of that decline, based on company accounts, to model changes and a shift toward open models. Those findings describe the sources and period covered by the report, not a universal market trend.
“Lots of traders are using Claude. Engineering is learning the importance of rate limiting for traders who accidentally run expensive queries without realizing it.”
— An engineer at the high-spending trading firm, quoted anonymously by The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
Causation and Returns Remain Unmeasured
The available figures do not show whether AI tools caused the reported increase in PRs. CodeRabbit’s data covers developers using its platform, and the report does not provide a comparable control group of non-users or enough detail to assess changes in code quality, rework, or delivery time. PR volume is an activity measure, not a direct measure of value created.
The 27-company spending sample is limited, with most businesses anonymized, and should not be read as a market average. The source also does not provide a consistent accounting of subscription subsidies, API use, or infrastructure costs across every company. It remains unclear how much total expenditure will settle as firms change models and usage policies, and whether higher output will translate into financial gains.
As an affiliate, we earn on qualifying purchases.
Companies Face Cost and Impact Tests
Companies cited in the report are continuing to adjust spending limits, monitor usage, and consider model choices as bills change. Some firms were preparing for subsidized subscriptions to end, while the high-spending trading firm described plans to build internal computing capacity and models. The report does not give a timeline or independently verified cost projections for those plans.
The next useful evidence would connect AI spending to comparable outcomes over time, including delivery speed, software quality, maintenance work, and the full cost of subscriptions and infrastructure. Until companies publish or collect that kind of evidence, the report’s spending totals and PR counts offer an incomplete view of ROI.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI per software engineer?
The report found wide variation among 27 workplaces, from a company limiting spend to $150 per engineer monthly to examples spending $1,600 or $3,000 per engineer monthly. These cases are not an industry-wide average.
Did AI make developers produce more pull requests?
CodeRabbit data showed weekly PRs up 113% among surveyed platform users and a 128% increase in a tracked group of 20,000 developers over a year. The figures show a rise in PR activity, but do not establish that AI alone caused it.
Does the report show that AI spending is profitable?
No. The report said few companies could answer the ROI question. It gathered spending and PR-output information, but did not present consistent evidence linking those measures to financial returns.
Why might AI costs change?
The report cited changes in model use, movement to open models, spending caps, and the possible end of subsidized subscriptions. Some companies said subscription costs could rise substantially if subsidies end.
Source: rss
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
