🔍 Read the full analysis: Evaluating AI Coding: Measure The Spend, Savings, And Outcomes on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A report published by The Pragmatic Engineer on Oct. 20 gathered AI coding budgets and output measures from 27 workplaces, with additional data from CodeRabbit and Gitar. Spending ranges widely, and reported productivity gains are inconsistent; the clearest benefits described are in triage, code research, prototyping and backlog work. The evidence does not establish a single return on investment, so companies are being urged to track accepted changes, rework and delivery outcomes rather than tool usage alone.
A report published Oct. 20 by The Pragmatic Engineer found that companies’ AI coding budgets range from about $150 to $3,000 per developer per month, while reported effects on output vary from no change to large increases. The findings, based on interviews at 27 workplaces and vendor data, arrive as companies set 2027 budgets but often lack a clear way to connect tool spending with delivered results.
The report’s author, Gergely Orosz, gathered accounts from engineering leaders and senior engineers at 27 workplaces, alongside aggregate figures from AI code-review companies CodeRabbit and Gitar. The source article says its cited material comes from the freely available portion of Orosz’s report. Spending varies substantially even among organizations of comparable size, and some employers impose monthly caps that workers can reportedly bypass.
Output figures do not tell one consistent story. CodeRabbit says developers on its platform produced 113% more pull requests per week than a year earlier. A Microsoft study of tens of thousands of engineers during an early-2026 rollout found adopters merged roughly 24% more pull requests over four months than they otherwise would have. Microsoft’s researchers cautioned that a merged pull request is not necessarily valuable work.
Companies interviewed reported outcomes ranging from flat pull-request counts to increases of five to ten times. In one case, pull-request frequency remained flat while the changes became about three times larger. The source article also cites tests by Veracode in which 44% of AI coding tasks produced a known vulnerability, underscoring the need to track quality and security as well as volume.
Infographic 01 · Where the money goes
From a $150 cap to $3,000 a month per developer.
Anecdotal data from 27 workplaces, mostly anonymized. It shows how wide the range is, not an industry average.
Spend per developer per month
Examples from the free part of the article · scale to $4,000
actual or typical spendbudget cap
What changed in 2026
Why the 2027 budget needs a range
- Peak in June and July: spend has roughly halved since for companies on API plans (Opus 5.5 and open models).
- Caps are common at medium and high spenders, and most people say they are easy to get around.
- Subsidies are ending: some companies must optimize or see the bill rise 5–10×.
- GitHub Copilot bills by token since 1 June 2026, at each model’s API rate, with no fallback to a cheaper model.
- Company size does not predict spend.
ROI is “a question that few companies seem able to answer – at least for now.”Gergely Orosz, The Pragmatic Engineer, 20 Oct 2026
Sources: The Pragmatic Engineer, “What’s the ROI of software engineers’ spending on AI?” (20 Oct 2026, free part); GitHub (27 Apr 2026).
Powered by Thorsten Meyer AI
Budgeting Beyond Tool Usage
The findings matter because AI coding bills can rise without a clear record of business value. The source article says some discounted subscriptions are due to expire, potentially raising costs for affected companies by five to ten times if they do not change plans or usage. It also reports that API-plan spending peaked in June and July before roughly halving, a pattern Orosz attributes to the availability of Opus 5.5 and a shift toward open models.
Higher output counts alone can obscure costs elsewhere. Larger pull requests may take longer to review, and security or rework can erase time saved during code generation. The article recommends tracking cost per accepted change, rework, change-failure rates and whether teams close more work than enters their backlog. Those measures connect AI use with delivery, rather than treating tokens or generated code as the result.
The source describes clearer gains in particular tasks: customer-issue triage, understanding unfamiliar code, prototypes and backlog cleanup on straightforward, uniform technology stacks. These examples suggest returns may depend on the work and the organization, not simply on which coding assistant is used. They are reported cases, not proof that the same savings will occur across companies.
As an affiliate, we earn on qualifying purchases.
Why AI Coding Bills Are Changing
Companies are reviewing costs as usage-based pricing becomes more prominent. The source article says GitHub moved all Copilot plans to usage-based billing on June 1, charging for tokens at each model’s API rate and removing the fallback to a cheaper model. GitHub said the change reflected growing agent use and its higher computing and inference demands.
Spending examples in the report range from a publicly traded company with 1,500 engineers and a $150 monthly cap per developer to a two-person European .NET agency spending about $3,000 per developer. One quant trading firm’s monthly AI costs reportedly rose from $500,000 in January to $8 million in September. The source says traders, rather than engineers, were its largest spenders; the firm reportedly evaluates the cost against each desk’s profit and loss.
Research cited in the source is mixed. METR initially found experienced open-source developers took 19% longer with AI in early 2025. A February follow-up estimated an 18% speedup for returning developers and a 4% change for new developers, but both estimates had confidence intervals that included zero. METR also described the results as unreliable, citing task selection by participants. These findings differ in setting and method from company and vendor reports.
As an affiliate, we earn on qualifying purchases.
What the Productivity Numbers Miss
The available figures do not establish a standard, independently verified productivity gain across software teams. Vendor data describes users of a vendor’s platform, while company reports may use different definitions of output. Pull-request counts also do not measure whether changes are useful, safe or maintainable.
The source article does not provide enough detail to compare all 27 workplaces on a common basis, including differences in team size, task complexity, tool use and measurement periods. It also relays a wide spread of reported output changes without a consistent underlying metric. The 44% Veracode result is a finding from its own test, not a measure of vulnerability rates in all AI-written code.
It remains unclear how much reported time saved becomes faster delivery or lower costs after review, rework, security checks and subscription charges. The source’s estimate that individual output is probably up by double digits is the author’s interpretation, not a settled finding across the research cited.
As an affiliate, we earn on qualifying purchases.
What Companies Should Track Next
Companies setting 2027 engineering budgets will need to compare AI expenses with outcomes over time, particularly as discounts end and billing structures change. The source article advises measuring costs against accepted changes, rework, production failures and the flow of completed versus incoming work. It does not identify a common industry standard or a deadline for adopting one.
Further evidence will depend on teams reporting results with clear baselines and comparable periods, while separating task-level time savings from overall delivery performance. Until then, the report’s examples can help organizations identify where to test AI coding tools, but they do not establish a universal return on investment.
AI coding budget management software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much do companies spend on AI coding tools?
The source article reports budgets ranging from about $150 to $3,000 per developer per month among the workplaces discussed. These are examples from the report, not a complete market-wide estimate.
Do the reported figures prove AI coding makes developers more productive?
No. Results vary by source and workplace, and pull-request counts do not necessarily measure the value or quality of delivered work. METR’s controlled estimates cited in the article were uncertain, with confidence intervals that included no change.
Where did companies report the clearest benefits?
The source describes reported gains in customer-issue triage, code archaeology, prototyping and backlog cleanup, particularly on simple, consistent technology stacks. These are company examples rather than guarantees for other teams.
What should engineering leaders measure?
The article recommends tracking cost per accepted change, rework, change-failure rates and whether completed work exceeds incoming work. These measures can help distinguish tool activity from useful delivery.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
