🔍 Read the full analysis: How Much Value Does AI Coding Add For Your Team? 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
Engineering teams in a survey of 27 workplaces reported AI coding budgets ranging from $150 to about $3,000 per developer per month, while reported output changes ranged from flat to several-fold. The evidence points to clear gains in some tasks, but more pull requests do not necessarily mean more delivered value, and many companies cannot yet quantify returns.
Engineering teams are spending $150 to about $3,000 per developer each month on AI coding tools, but many still cannot say how much business value that spending produces, according to a report by Thorsten Meyer published October 23. The report draws on a survey of 27 workplaces and productivity data from companies and vendors, which show uneven results: some teams report more merged changes or faster task completion, while others see little change in pull-request counts.
The spending figures come from interviews and information gathered by The Pragmatic Engineer, which published its account on October 20. The sample spans a publicly traded company that caps monthly spending at $150 per developer and a two-person European agency, Ix Technologies, that spends about $3,000 per developer. Company size alone did not explain where budgets landed, Meyer wrote. Some organizations use caps, but participants said those limits can be bypassed.
Output measures also differ. CodeRabbit, an AI code-review vendor, reported that developers on its platform created 113% more pull requests per week than a year earlier. A Microsoft study covering tens of thousands of engineers during an early-2026 rollout found adopters merged about 24% more pull requests than they otherwise would have over four months. Microsoft’s researchers cautioned that a merged pull request is not the same as delivered value.
Among the 27 workplaces, reported changes ranged from flat pull-request frequency to five- or tenfold increases. At one company, pull-request frequency stayed flat while the changes grew to three times their previous size. These are different measures and cannot be treated as a single productivity result. The report also cites a Veracode test in which 44% of AI coding tasks produced a known vulnerability; the source does not provide enough detail here to generalize that rate to all AI-written code or teams.
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
Why More Code May Not Mean More Value
The figures matter as companies set 2027 engineering budgets and face the possibility of higher bills. Meyer’s report says some discounted subscriptions are due to expire, while GitHub moved Copilot plans to usage-based billing on June 1. In a statement explaining the change, GitHub said agentic use was becoming more common and brought higher computing and inference demands. A team can therefore face rising costs even when its measured output does not rise in step.
Counting generated code or pull requests can reward volume without showing whether work was useful, safe, or accepted by customers. The report points to measures closer to outcomes: cost per accepted change, rework, change-failure rates, and whether teams close more user stories than arrive. These measures can also reveal costs that a simple output count misses, such as added review work or defects.
The evidence suggests the returns depend on the work and the surrounding process, not simply on whether a team buys a particular tool. AI may shorten research and routine work while leaving complex legacy changes difficult to review. Organizations that track both spending and delivery quality are better placed to decide whether to expand, limit, or redirect a budget; the available examples do not establish one return rate that applies across teams.
As an affiliate, we earn on qualifying purchases.
Evidence Behind the Budget Debate
The Pragmatic Engineer’s October 20 report combined interviews with engineering leaders and senior engineers at 27 workplaces with aggregate information from CodeRabbit and another code-review vendor, Gitar. Meyer says the details he used came from the report’s freely readable material. These sources have different methods: workplace accounts are not controlled experiments, and vendor-platform data reflects teams already using those products.
Controlled research cited by Meyer is mixed. METR’s early-2025 study found experienced open-source developers took 19% longer when using AI tools. A February follow-up estimated an 18% speedup for returning developers and a 4% speedup for new developers, but both estimates had confidence intervals that included zero. METR also said the data was unreliable because 30% to 50% of participants held back tasks they did not want to do without AI.
In a METR survey of 349 technical workers, the median respondent reported a 1.4-to-2-times change in the value of their work. METR noted that participants in its 2025 study had overestimated AI’s effect by 40 percentage points. Those self-reports, vendor metrics, company comparisons, and experiments measure different things, so they should not be combined as if they were a single estimate of productivity.
“Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.”
— GitHub, explaining its Copilot billing change as quoted in Meyer’s report
As an affiliate, we earn on qualifying purchases.
Returns Remain Hard to Compare
The evidence does not establish a typical return on investment for AI coding tools. The 27-workplace sample is not described as representative of all engineering teams, and its reported results rely on varied measures. Vendor platform data, company anecdotes, employee surveys, and controlled trials each have limitations and do not show the same outcome.
It is also unclear how much reported output translates into features customers use, fewer incidents, lower operating costs, or faster delivery. The report gives examples of security vulnerabilities and larger changes, but does not provide a common measure of downstream review, rework, or defect costs across the companies. Meyer’s broad assessment that individual output is probably up by double digits is his interpretation, not a result established uniformly by the cited studies.
best AI coding tools for developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Track Outcomes as Budgets Renew
Companies renewing or setting budgets can compare spending with delivery and quality measures rather than relying on tool usage or generated-code totals. The report recommends tracking cost per accepted change, rework, change-failure rates, and completed stories against incoming work. Results should be compared across similar tasks and time periods, with review and security costs included.
As discounts expire and usage-based plans expose more of the underlying model costs, organizations may change spending caps, model choices, or the tasks they assign to AI tools. The source does not identify a common timetable for those decisions or a standard measurement system already adopted across the 27 workplaces. The next meaningful evidence will be whether teams can show that measured savings or delivery gains persist as costs and workflows change.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The workplaces described in the report ranged from a $150 monthly cap per developer to spending of about $3,000 per developer each month. These are examples from a 27-workplace survey, not an industry-wide average.
Does AI coding increase developer productivity?
Some sources report more pull requests or faster completion of particular tasks, while other teams report flat counts. Controlled METR results cited in the report are uncertain, so the evidence does not establish a single productivity gain that applies to all teams.
Why are pull-request counts not enough to measure value?
A higher count does not show whether changes are useful, safe, or delivered to customers. Microsoft researchers cited in the report said that a merged pull request is not the same as the value it delivers; teams also need to track rework, failures, and outcomes.
Where does the report say AI coding tools help most?
Examples in Meyer’s report include issue triage, code archaeology, prototypes, and backlog cleanup, particularly on simple, uniform technology stacks. These are reported experiences, not guarantees for every team.
What should teams measure to judge their return?
The report suggests measuring cost per accepted change, rework, change-failure rates, and whether completed user stories outpace incoming work. It does not offer a standardized calculation that has been validated across all organizations.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
