🔍 Read the full analysis: AI Coding: Is The Investment Paying Off? on ThorstenMeyerAI.com
Get business pricing on tech for your team
- 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, but many could not say what the spending returned. Productivity measures vary sharply across companies and studies, while reported gains are clearest for tasks such as triage, code research and backlog cleanup. The evidence does not establish a single industry-wide return on investment.
A survey of 27 workplaces found that monthly AI coding budgets range from about $150 to $3,000 per developer, while many engineering leaders still cannot quantify what that spending delivers. The findings, reported by Gergely Orosz in The Pragmatic Engineer on October 20, arrive as companies set 2027 budgets and weigh rising or changing subscription costs against uneven evidence of productivity gains.
The survey drew on conversations with engineering leaders and senior developers, alongside aggregate figures from AI code-review companies CodeRabbit and Gitar. The report describes substantial differences in spending, with company size alone not explaining where budgets land. Some organizations set strict caps; others spend far more, including a two-person European .NET agency reporting about $3,000 per developer each month.
Output measures also differ. CodeRabbit reported that developers on its platform created 113% more pull requests per week than a year earlier. Microsoft, studying tens of thousands of engineers during an early-2026 rollout of Claude Code and GitHub Copilot CLI, found adopters merged about 24% more pull requests over four months than they otherwise would have. Microsoft cautioned that a merged pull request does not by itself measure the value delivered.
Other companies in the survey reported flat pull-request counts, while some saw much larger changes. The source report also cites a Veracode test in which 44% of AI coding tasks produced a known vulnerability. That result underscores that higher output counts do not establish better or safer software. The available material does not provide enough detail about the test methods to generalize the figure to all AI-generated code.
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
Budgets Need Outcome Measures
The findings matter because software companies are making spending decisions without a consistent way to connect AI tool costs to business results. A rise in pull requests may reflect faster work, but it can also mean larger review burdens or more changes that need rework. Output volume is not the same as delivered value, and the reported security result points to a potential cost if review does not keep pace.
The source report points to practical measures companies could track, including cost per accepted change, rework, change failure rates and whether completed user stories outnumber incoming work. Those measures would not make every team’s work directly comparable, but they can help organizations test whether a tool is improving outcomes rather than merely increasing usage. Returns appear to depend on the task, codebase and review process, not just the software subscription.
Some companies are also approaching a cost change: the source says several reported that discounted subscriptions are due to expire, potentially raising bills five- to tenfold if they do not optimize. That is a reported risk, not a confirmed increase across the market. For budget owners, the immediate question is how to evaluate tools before subsidies end and new spending plans are approved.
As an affiliate, we earn on qualifying purchases.
Mixed Results Across Teams
The source report describes spending on API plans as having peaked in June and July and then roughly halved for some companies, which Orosz attributed to the arrival of Opus 5.5 and a shift toward open models. Those figures are based on the report’s company conversations; they are not a market-wide spending series. The account also says spending caps are common and, according to some interviewees, easy to work around.
Billing policies are changing alongside usage. GitHub said it moved 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. The company explained the shift by saying, “Agentic usage is becoming the default” and brings higher compute and inference demands. The change makes the relationship between tool use and costs more direct for customers.
Research findings cited in the report do not point to one settled productivity estimate. METR’s early-2025 study found experienced open-source developers took 19% longer with AI. A February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but both confidence intervals included zero; METR also said the data was unreliable because many participants withheld tasks they preferred to do without AI. Self-reported improvements were higher, but METR has warned that participants in its 2025 survey overestimated AI’s effect.
““A question that few companies seem able to answer – at least for now.””
— Gergely Orosz, The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
The survey does not establish an industry-wide return on investment, and its 27 workplaces are not presented as a representative sample of all software companies. Vendor platform figures, internal company studies and controlled experiments measure different groups and outcomes, so their results cannot be treated as directly comparable. The size of any typical productivity gain remains unsettled.
It is also unclear how much reported output growth becomes customer value after review, testing, rework and maintenance. The Veracode result cited in the source does not, by itself, show how frequently vulnerabilities reach production. Nor does the source provide a common accounting method for comparing tool costs, engineering time and business outcomes across teams.
Some workplace examples are individual company accounts, not controlled trials. Results from a small agency using one main technology stack may not transfer to teams maintaining complex legacy systems. Claims about expiring discounts and possible five- to tenfold bill increases also apply to companies described in the report, not necessarily every customer.
As an affiliate, we earn on qualifying purchases.
Companies Face Budget Decisions
Companies are now setting 2027 engineering budgets, and the source report says some will need to decide whether to renew discounted subscriptions, impose spending limits or shift models. The next useful evidence will come from teams comparing costs with measures tied to accepted work and delivery quality, rather than relying on token use or pull-request totals alone.
Further updates may clarify how companies handle changing billing plans and whether their measured results persist across different tasks and codebases. For now, the report offers no single verdict that AI coding pays off everywhere: it documents reported gains in particular settings, costs that vary widely and substantial gaps in how organizations measure returns.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The 27-workplace survey described monthly budgets ranging from about $150 to $3,000 per developer. The figures are examples from surveyed workplaces, not a market-wide average.
Has AI coding increased developer output?
Some reported measures rose: CodeRabbit said pull requests per week on its platform increased 113% year over year, while a Microsoft study found adopters merged about 24% more pull requests over four months. Other surveyed companies reported flat counts, and these measures do not by themselves show business value.
Does the evidence show that AI coding is profitable?
No single conclusion is established by the source material. Many companies reportedly cannot quantify returns, and studies use different methods and outcomes. The report recommends tracking costs against accepted changes, rework, failures and completed work.
What kinds of work appear to benefit most?
Company accounts in the report describe gains in issue triage, code archaeology, prototyping and backlog cleanup, particularly on simple, uniform technology stacks. These are reported examples rather than proof that every team will see similar results.
What safety concerns did the report raise?
It cited a Veracode test in which 44% of AI coding tasks produced a known vulnerability. The source does not provide enough methodological detail to apply that figure to all AI coding or to determine how many vulnerabilities reached production.
Source: ThorstenMeyerAI.com
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
