🔍 Read the full analysis: The Hidden Costs And Potential Payoffs Of AI Coding 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
A report published October 20 by The Pragmatic Engineer found AI coding budgets at 27 workplaces ranged from $150 to about $3,000 per developer per month, while reported productivity results varied sharply. Some teams cite faster triage, prototyping and backlog work, but higher pull request counts do not establish business value, and security and review costs remain concerns.
Engineering teams surveyed at 27 workplaces report spending from $150 to about $3,000 per developer per month on AI coding tools, while their productivity results range from flat pull-request counts to sharply higher output. The findings, published October 20 by The Pragmatic Engineer, land as companies set 2027 budgets and face uncertainty about whether the spending is producing useful, accepted software changes.
The survey, by The Pragmatic Engineer editor Gergely Orosz, drew on conversations with engineering leaders and senior engineers at 27 workplaces, alongside aggregate figures from code-review vendors CodeRabbit and Gitar. The source article says its cited material comes from the freely available portion of Orosz’s report. Spending levels varied widely without a clear relationship to company size: one publicly traded company with 1,500 engineers set a $150 monthly cap per developer, while European agency Ix Technologies reported spending about $3,000 per developer.
Reported output measures do not point to a single productivity result. CodeRabbit said developers on its platform created 113% more pull requests per week than a year earlier; that is vendor data from teams already using its AI review product. Microsoft’s study of 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. The study’s authors cautioned that a merged pull request is not necessarily valuable work.
At other surveyed companies, pull-request frequency was flat, or the reported gains involved larger changes rather than more frequent ones. The source also cites METR research with mixed results: its early-2025 study found experienced open-source developers took longer with AI, while a February follow-up estimated modest speedups whose confidence intervals included zero. METR described the follow-up data as unreliable, in part because some participants avoided tasks they preferred to do without AI.
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 AI Coding Bills Need Outcome Measures
The gap between tool usage and delivered value matters because more code or more pull requests can also mean more review, rework or risk. Companies that budget by subscriptions or token consumption may not know whether their investment helps teams ship useful changes, clear backlogs or improve customer outcomes. The report recommends tracking measures such as cost per accepted change, rework, change failure rate and whether completed user stories outpace incoming work.
Security is another cost to account for. The source reports that Veracode’s latest test found 44% of AI coding tasks produced a known vulnerability. That figure describes the tasks in the test, not the share of all AI-written software that is vulnerable. Even so, it underscores why review and security capacity belong in any assessment of returns, rather than treating faster code generation as a complete productivity measure.
Some organizations have a direct financial yardstick. The report describes a quant trading firm whose AI spending rose from $500,000 per month in January to $8 million in September, with traders as its largest spenders. The firm evaluates AI as part of each desk’s profit-and-loss picture, according to the source. Most engineering teams lack an equivalent scoreboard, making their return harder to establish.
As an affiliate, we earn on qualifying purchases.
Budgets Shift as Billing Changes
Spending on API plans reportedly peaked in June and July and had roughly halved by the time of Orosz’s reporting, which he attributed to the arrival of Opus 5.5 and a move toward open models. Several companies also told him that discounted subscriptions were due to expire, potentially raising their bills by five to ten times unless they changed plans or usage. These are accounts from the surveyed companies, not a forecast that applies to every organization.
Billing models are changing alongside usage. 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 use was becoming the default and brought higher computing and inference demands. The shift can make costs more sensitive to how much developers use tools and which models they select.
The reported gains also depend on the work and codebase. Survey participants cited clear benefits in customer-issue triage, understanding older code, rapid prototypes and backlog cleanup on simple, uniform technology stacks. One SaaS company’s staff engineer said triage that could take a day sometimes became minutes of agent work plus up to an hour of human checking. Such examples are accounts from individual teams; they do not establish a general productivity rate.
“Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.”
— GitHub, explaining its Copilot billing change as quoted by ThorstenMeyerAI.com
How Much Output Becomes Value
The evidence does not yet settle how much AI coding raises overall business value. Vendor platform figures, company reports, controlled experiments and worker self-reports measure different things and do not provide a directly comparable estimate. METR’s follow-up results were uncertain, and the source says its survey participants had previously overestimated AI’s effect by 40 percentage points.
It is also unclear how representative the reported cases are across organizations, programming languages and complex legacy systems. The source says AI appears less useful for some complicated legacy work, but the material provided does not quantify that difference. Nor does the cited vulnerability test establish the real-world rate of security flaws in deployed AI-assisted code.
Companies Prepare 2027 Budgets
Engineering leaders are setting 2027 budgets and will need to decide whether to renew discounted subscriptions, adopt usage-based plans, set spending limits or shift to other models. The report’s findings point to measuring completed and accepted work, rework and failures alongside tool costs, rather than relying on activity totals alone.
The next useful evidence will come from company-level tracking that connects AI use with delivery quality, security and cost over time. The source does not identify a common measurement standard or give a timeline for one. Until teams publish comparable results, the size and durability of AI coding’s returns will remain difficult to establish.
Key Questions
How much are companies spending on AI coding tools?
The workplaces in The Pragmatic Engineer’s survey reported budgets ranging from $150 to about $3,000 per developer per month. These examples do not establish an industry-wide average.
Does AI coding increase developer productivity?
Some reported measures show higher output, including more merged pull requests at Microsoft and more weekly pull requests on CodeRabbit’s platform. Other workplaces reported flat pull-request counts, and METR’s experimental findings were uncertain. The evidence does not establish one productivity gain that applies to all teams.
Where do teams report the clearest benefits?
Teams described gains in customer-issue triage, code archaeology, prototypes and backlog cleanup, especially on simple, consistent technology stacks. These are reported examples, not proof that the same results will occur in other organizations.
What should companies track to judge AI coding’s return?
The report suggests tracking cost per accepted change, rework, change failure rate and whether completed user stories exceed incoming work. Pull-request counts alone do not show whether changes deliver value.
What remains unknown about AI-generated code security?
The cited Veracode test found a known vulnerability in 44% of the AI coding tasks it tested. The source does not establish that this is the vulnerability rate for all AI-assisted code or deployed software.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
