🔍 Read the full analysis: Is AI Coding Paying Off? Where The Money Burns, Where It Pays, And How To Measure It on ThorstenMeyerAI.com
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
TL;DR
AI coding budgets reported across 27 workplaces range from $150 to about $3,000 per developer per month, but many companies cannot link that spend to delivered value. Accounts point to gains in triage, code research and simple-stack backlog work, while output measures vary and security, review and rework costs remain concerns.
The Pragmatic Engineer’s Gergely Orosz gathered accounts from engineering leaders and senior engineers and included aggregate figures from AI code-review services **CodeRabbit and Gitar**. The publicly available portion describes monthly spending caps, changes in API costs and examples of teams using AI for software work. It does not establish a single industry-wide return on investment.
Reported output measures conflict. CodeRabbit says developers on its platform created **113% more pull requests per week** than a year earlier, a vendor measure covering teams already using its review tool. A Microsoft study of tens of thousands of engineers during an early-2026 rollout found adopters merged about **24% more pull requests** over four months than they otherwise would have, according to the source report. Other companies surveyed reported flat PR counts, while one said PRs became larger.
The source also cites security and productivity research as reasons not to treat activity as value. Veracode’s latest test found **44% of AI coding tasks produced a known vulnerability**. METR’s February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but said both estimates had confidence intervals that included zero and described the data as unreliable. These findings use different methods and populations, so they cannot be read as one comparable measure of productivity.
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
How Companies Can Judge AI Spend
The budget question is moving beyond whether developers use AI to whether the work it produces is **accepted, safe and useful**. A higher PR count can reflect faster delivery, but it can also mean larger review burdens or more changes that need rework. Microsoft’s study authors, as quoted in the source report, cautioned that **a merged pull request is not the same as delivered value**.
The article recommends tracking **cost per accepted change, rework, change failure rate and stories closed compared with stories arriving**. Those measures connect spending with engineering outcomes more directly than token use or raw output volume. They also help expose costs that can be hidden when developers spend less time typing but reviewers, security teams or maintainers must do more work.
That distinction matters as discounts expire and usage-based billing becomes more common. Companies that cannot compare tool costs with quality and delivery measures may face rising bills without knowing which teams or tasks benefit. The source reports that some companies expect discounted subscriptions to end soon and could see bills rise five- to tenfold if they do not optimize; that figure is a reported risk, not a measured outcome across the surveyed workplaces.
As an affiliate, we earn on qualifying purchases.
Budgets Shift as Billing Changes
The Pragmatic Engineer report says spending on API plans peaked in June and July and had roughly **halved by the time of its reporting** for companies on those plans. Orosz attributed the fall to the availability of Opus 5.5 and a move to open models. Spending caps are common, the report says, but respondents described them as easy to get around.
At the high end, a two-person European .NET agency, Ix Technologies, reported spending about **$3,000 per developer per month**. A publicly traded company with 1,500 engineers capped spending at $150. A quant trading firm’s AI bill reportedly grew from $500,000 a month in January to $8 million in September; its largest spenders were traders, and the firm evaluates AI as part of desks’ profit and loss rather than relying on a standard engineering productivity measure.
Billing changes add pressure to reassess budgets. The source says GitHub moved all Copilot plans to usage-based billing on June 1, charging tokens at each model’s API rate and removing the fallback to a cheaper model. GitHub explained the change by saying agentic usage was becoming the default and carried higher compute and inference demands.
““A merged PR is not the same as the value it delivers.””
— Microsoft study authors, as quoted in The Pragmatic Engineer report
Where the Evidence Falls Short
The survey and company examples do not provide a standardized comparison across teams. Differences in codebases, task difficulty, staffing, tool selection and accounting methods make the reported spending and output figures difficult to compare directly. Vendor figures, including CodeRabbit’s platform data, describe selected users and are not independent estimates of industry-wide gains.
Even controlled research remains unsettled. METR’s estimates had uncertainty ranges that included no effect, and the group said participant behavior made the data unreliable. The source report also cites a METR survey in which workers reported a median 1.4-to-2-times change in the value of their work, while METR noted that participants in its 2025 study had overestimated AI’s effect by 40 percentage points. Self-reported value and measured delivery are not interchangeable.
The source material ends as it begins describing additional examples of where AI coding pays, so it does not provide a complete account of every case or a definitive overall ROI estimate. It also does not establish how widespread the cited security finding is across production code or how often detected vulnerabilities were fixed before release.
What Budget Owners Should Track
Companies setting **2027 engineering budgets** will need to compare AI costs with outcomes over time, rather than infer returns from licenses, token consumption or PR volume alone. The measures suggested in the report—cost per accepted change, rework, change failures and the balance between stories completed and incoming—can help teams assess whether faster work results in better delivery.
For a useful comparison, organizations would need to define those measures consistently and account for review and security work as well as development time. The source report does not announce a common industry standard or a scheduled follow-up study. It remains unclear which employers will publish comparable results, or how many will reduce, expand or restructure their tool budgets as discounts end and usage-based charges take effect.
Key Questions
How much are companies spending on AI coding tools?
The Pragmatic Engineer survey reported budgets ranging from **$150 to about $3,000 per developer per month** across 27 workplaces. The figures are reported examples, not an industry-wide average.
Does the report show that AI coding increases productivity?
It presents mixed evidence. CodeRabbit reported more weekly PRs on its platform, and a Microsoft study found about 24% more merged PRs among adopters over four months. Other surveyed teams reported flat PR counts, and METR’s speed estimates were uncertain.
Where did teams say AI coding helped most?
The source describes reported gains in **triage, code archaeology, prototypes and backlog cleanup**, especially for simpler, more uniform software stacks. These are company accounts, not proof that all teams will see the same results.
What should companies measure to judge return on investment?
The report recommends tracking **cost per accepted change, rework, change failure rate and stories closed compared with stories arriving**. These measures connect tool spending to delivery and quality more directly than usage volume alone.
What is still unknown about AI coding returns?
There is no consistent, comparable estimate of overall industry ROI in the source material. Results vary by team and task, and the available studies use different methods; security, review and rework costs also need clearer measurement.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
