🔍 Read the full analysis: How To See Where AI Coding Pays—and Where Money Burns 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 based on interviews at 27 workplaces finds AI coding spending ranges from about $150 to $3,000 per developer per month, with no clear link between company size and budget. Reported output gains vary sharply, while specific tasks such as triage and code archaeology appear more consistently useful; the source says many companies still lack a way to connect spending to delivered value.
A review of 27 workplaces found AI coding budgets ranging from about $150 to $3,000 per developer per month, while reported productivity results ranged from flat pull-request counts to much larger gains. The findings, reported by The Pragmatic Engineer on October 20 and summarized by Thorsten Meyer on October 23, come as companies set 2027 budgets and try to determine whether AI tools are producing value beyond more code.
The Pragmatic Engineer’s Gergely Orosz interviewed engineering leaders and senior engineers at 27 workplaces and included aggregate data from AI code-review vendors CodeRabbit and Gitar. The freely available portion of the report describes a wide range of spending: one publicly traded company with 1,500 engineers caps use at $150 per developer monthly, while European agency Ix Technologies, a two-person .NET firm, spends about $3,000 per developer. Company size did not predict spending in the examples.
Output measures also conflict. CodeRabbit reported that developers on its platform created 113% more pull requests per week than a year earlier. Microsoft’s 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. Those are different measures and settings; Microsoft’s authors cautioned that a merged pull request does not necessarily represent delivered value. Several companies interviewed reported flat PR counts, and one said PRs became three times larger.
The report points to more task-specific gains in customer-issue triage, code archaeology, prototyping and backlog cleanup, particularly on uniform technology stacks. One SaaS company’s staff engineer said issue triage that could take a day sometimes became minutes of agent work followed by up to an hour of human review. Ix Technologies said its C# and Blazor stack helped it work through a long-standing backlog. These are company accounts, not controlled comparisons. The material also cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability, highlighting the need to account for review and security work.
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 Value Measures
The budget question is becoming more immediate as some companies’ discounted subscriptions approach expiration and usage-based plans can make costs less predictable. The source says spending on API plans peaked in June and July and had roughly halved by the time of its report, attributing the decline to a move to open models and the availability of Opus 5.5. Those explanations are reported by Orosz, rather than established as a universal market trend.
Without a connection between usage and outcomes, a rising token bill or a higher PR count can create a false sense of progress. The report recommends tracking cost per accepted change, rework, change-failure rate and stories closed compared with stories arriving. These measures can help teams see whether faster drafting translates into maintainable software and work completed, rather than simply more code awaiting review.
The cost of oversight matters too. If review and security checks do not keep pace with generated changes, apparent productivity may shift effort downstream. Veracode’s reported vulnerability result does not show that every AI-assisted change is unsafe, but it is a reason to include defects and review effort in evaluations rather than treating output volume as a standalone return.
As an affiliate, we earn on qualifying purchases.
Evidence Behind the Spending Debate
Orosz’s October 20 report arrives as engineering organizations plan budgets for 2027. The source says spending caps are common and can be bypassed, while some companies face the end of discounted subscriptions. It also reports a quant trading firm whose AI spend rose from $500,000 a month in January to $8 million in September, with traders accounting for the largest spending. That firm used desk profit and loss as its measure; the example is not representative of software teams generally.
Other research cited in Meyer’s account points to uncertainty about productivity. 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 confidence intervals included zero. METR described the follow-up data as unreliable in part because 30% to 50% of participants held back tasks they did not want to do without AI. Survey responses can also diverge from observed outcomes: METR said participants in its 2025 survey overestimated AI’s effect by 40 percentage points.
Billing changes add pressure to measure use. The source says GitHub moved 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 is becoming the default, and it brings significantly higher compute and inference demands.” The policy change makes consumption more visible as a cost, but it does not by itself establish whether the work generated is worthwhile.
““Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.””
— GitHub, explaining its Copilot billing change
As an affiliate, we earn on qualifying purchases.
Productivity Gains Remain Uneven
The evidence does not establish a single productivity gain that applies across companies. Vendor platform data, Microsoft’s study, workplace interviews and METR experiments cover different users, tasks and outcome measures. Some companies report substantial increases, while others see flat PR counts or larger changes without more PRs. The comparison basis varies, so the figures should not be treated as directly interchangeable or as a forecast for every team.
It is also unclear how much reported speed turns into shipped features, fewer defects or reduced total engineering costs. The source offers examples of faster triage and backlog work, but does not provide a common accounting method that compares those benefits with model charges, human review, rework and security remediation. The Veracode figure concerns tasks in its test and should not be read as a rate for all AI-generated code.
AI-powered developer productivity tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Track Results Through 2027
As companies finalize 2027 plans, the next step described in the report is to evaluate AI coding by outcome rather than tool consumption. Teams can compare spending with accepted changes, rework, production failures and the rate at which work is completed versus newly arriving. The source does not identify a common industry standard or a date when one will emerge.
For organizations whose discounts are ending or whose billing has shifted to usage-based pricing, budgets may change as actual consumption becomes clearer. Whether those costs produce a return will depend on each team’s codebase, tasks, review capacity and chosen measures. Further evidence would need to show consistent outcomes across those settings.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The workplaces in the report ranged from a $150 monthly cap per developer to about $3,000 per developer per month. These are examples from 27 workplaces, not a market-wide average.
Does AI coding increase developer output?
The evidence cited is mixed. CodeRabbit reported 113% more weekly pull requests on its platform year over year, and Microsoft reported about 24% more merged PRs among adopters over four months. Other companies reported flat PR counts, and the measures do not establish the same thing as delivered value.
Which tasks appear to benefit most?
The report describes gains in triage, code archaeology, prototyping and backlog cleanup, especially on simple, uniform technology stacks. Those examples are reported experiences rather than controlled findings that apply to every team.
What should companies measure to assess return?
The source recommends tracking cost per accepted change, rework, change-failure rate and stories closed versus stories coming in. Such measures capture more than usage or the number of pull requests created.
What remains uncertain about AI coding’s return?
There is no consistent result across the studies and companies cited, and the report does not establish how much faster work becomes better or cheaper software after review, rework and security costs are counted.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
