TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
The Pragmatic Engineer gathered AI spending and pull-request (PR) data from 27 workplaces, alongside aggregate figures from CodeRabbit and Gitar. CodeRabbit’s data showed more than twice as many PRs per developer as a year earlier, but the report found that most companies could not say whether AI spending produced a measurable return.
AI tool spending varies widely across software companies, while available platform data show developers opening more pull requests (PRs), according to a report by The Pragmatic Engineer based on interviews with leaders and senior engineers at 27 workplaces. But the report found that few companies could answer the central question: whether the spending is generating a clear return on investment.
The report combined company-reported spending and PR activity with aggregate data shared by CodeRabbit and Gitar, providers of AI software-development tools. CodeRabbit’s anonymized platform data showed a 113% rise in weekly PRs and a 127% rise in monthly PRs. In a separate comparison tracking the same 20,000 developers over a year, PR counts increased 128%, the report said. These are changes in activity on one platform, not a direct measure of software quality, business value or the effect of AI alone.
Gitar data covering hundreds of customers showed that some teams classified as “AI-native” produced more than 60 PRs per developer per week. The report also described wide differences in company spending: a two-person .NET business reported spending $3,000 per developer per month, while a publicly traded company with 1,500 engineers capped monthly AI spending at $150 per developer. The figures come from the report’s respondents and are not a representative industry-wide survey.
Spending patterns also shifted over time. Respondents said monthly costs peaked around June and July and had roughly halved for some companies on API-based plans. The report associated the decline with a move toward open models and changes in model use. It also found that spending caps were common among medium- and higher-spending companies, while subsidized subscriptions made some firms’ actual costs harder to see.
PR Growth Is Not a Return Measure
The findings capture a problem facing engineering leaders as AI tools become a recurring budget item: usage and output counts are easier to collect than financial returns. A rise in PRs may indicate that teams are producing more code changes, but it does not establish that changes are better, reach customers sooner, reduce staffing needs or lower total development costs.
PR volume can also reflect differences in team practices and company culture. The report’s author noted that teams have different norms for how many PRs engineers submit, and that workers may adapt to those expectations. That makes raw comparisons across workplaces difficult. For employees and executives, the distinction matters: a higher tool bill may be justified if it improves a meaningful outcome, but more PRs alone cannot show that it does.
Costs may change as companies shift between subscriptions, usage-based API plans and open models. Some respondents said subsidies were ending, potentially raising bills by five to ten times unless firms changed their approach. That makes tracking cost per engineer useful, but it still needs to be compared with outcomes such as delivery time, reliability or engineering costs to support an ROI calculation.
AI development tools for software engineers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How the Spending Figures Were Collected
The Pragmatic Engineer said it sought data after an engineer asked how much other companies spend on AI per developer and whether PR throughput changed before and after adoption. The author gathered accounts from engineering leaders and senior engineers at 27 workplaces and added anonymized aggregate platform data from CodeRabbit and Gitar. Most company names were withheld at respondents’ request.
The report described AI spending as increasing at many firms, with Claude and Opus 5.5 frequently named among the main destinations for spending; Cursor, Codex and GitHub Copilot were also mentioned. One quant-trading company reported monthly AI costs rising from $500,000 in January to $8 million in September, or 16 times the January figure. The report said its average was $1,600 per developer per month, while traders were among the largest users. This is one company’s account, not a typical spending level.
The author also reported that some companies’ costs fell after summer peaks, in part as they moved to open models. Because organizations use different tools, pricing arrangements, limits and methods for counting engineers, the spending examples are best read as a snapshot of varied practices rather than a league table.
““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 quant-trading company, quoted by The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
What the Data Cannot Establish
The figures do not show that AI tools caused the increase in PRs. The supplied material gives no controlled comparison separating AI use from changes in staffing, workload, team processes or other factors. It also does not report whether the additional PRs were merged, improved product quality or reduced time and costs.
The 27 workplace accounts are not described as a representative sample, and many company identities are anonymized. The report does not provide a common accounting method for subscriptions, API use, subsidies or employee counts. It is not yet clear how typical the reported spending levels are, or how firms measured their returns. The supplied excerpt also ends mid-sentence in its account of one company’s plans to build computing capacity, so further details on that plan are unavailable here.
As an affiliate, we earn on qualifying purchases.
Companies Face a Measurement Test
The report points to a need for companies to track AI costs alongside outcomes that matter to their operations, rather than treating PR volume as a stand-in for productivity or ROI. Respondents may face new budget decisions as subsidies expire, model choices change and spending caps are reviewed. Some firms are also considering or building their own computing capacity, according to the report.
No shared industry benchmark or follow-up measurement is provided in the source material. The next useful evidence would show how spending changes affect comparable measures—such as delivery speed, defects, maintenance work and total engineering costs—over a defined period. Until then, the report offers a snapshot of rising but uneven spending, with the financial payoff still unresolved.
developer productivity enhancement tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI per software engineer?
The report found a wide range among its respondents. Examples included $150 per developer per month at a 1,500-engineer public company and $3,000 per developer per month at a two-person .NET business. These examples are not an industry average.
Did AI use increase the number of pull requests?
CodeRabbit’s platform data showed weekly PRs rising 113% and monthly PRs rising 127%; a tracked group of 20,000 developers showed a 128% increase over a year. The report does not establish that AI alone caused those changes.
Does a higher PR count prove AI is worth its cost?
No. PR counts measure activity, not whether code is better, delivered faster or less costly. The report says few companies could quantify AI’s return, and it does not provide a direct ROI calculation.
Why did some companies’ AI costs fall after summer?
Respondents said spending peaked around June and July and roughly halved for some firms on API plans. The report linked the decline to model changes and a move toward open models, but the supplied material does not provide a standardized breakdown for all companies.
Source: rss
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
