🔍 Read the full analysis: Track The Spend, Find The Value: Measuring 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 by The Pragmatic Engineer on October 20 surveyed engineering leaders and senior engineers at 27 workplaces about AI coding costs and results. Spending varies widely, and productivity measures conflict: reported wins include faster triage and backlog work, while PR counts and controlled study results are mixed. The report recommends measuring accepted changes, rework and delivery outcomes rather than tool usage alone.
Engineering teams surveyed by The Pragmatic Engineer reported monthly AI coding budgets ranging from about $150 to $3,000 per developer, while many said they could not quantify what the spending delivered. The report, published October 20, draws on interviews at 27 workplaces and vendor data, offering a timely account as companies set 2027 budgets and some discounted subscriptions near expiry.
The spending figures vary sharply, including within companies that do not fit a simple size-based pattern. One publicly traded company with about 1,500 engineers caps spending at $150 per developer per month. European agency Ix Technologies, with two developers, spends about $3,000 per developer monthly. Other firms reported caps from $1,000 to $4,000, with typical monthly use around $1,000, according to the report.
One quant trading firm with roughly 2,000 developers said its AI bill increased from $500,000 a month in January to $8 million in September. The report says the firm’s largest spenders are traders rather than engineers and that the business evaluates AI as part of each desk’s profit and loss. The figures are an example from one firm, not a measure of industry-wide spending.
Evidence about output is uneven. 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 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. Microsoft’s researchers cautioned that a merged pull request does not by itself show the value delivered. At several surveyed companies, PR counts stayed flat even as individual PRs grew larger.
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 central issue is not simply how much teams spend, but whether the added activity improves software delivery. More code or more pull requests can mean faster progress, but can also create more review work, rework or security exposure. The report cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability; that result is a finding from the cited test, not a claim that 44% of production code is vulnerable.
The report suggests tracking cost per accepted change, rework, change failure rates and whether completed stories outpace incoming work. Those measures connect tool costs to work that teams accept and maintain. They also help distinguish faster generation from useful, reliable delivery, a distinction that matters as discounted subscriptions end and usage-based charges expose companies to higher bills.
AI coding budget tracking software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Why AI Coding Bills Are Shifting
The October 20 report by Gergely Orosz combines interviews with engineering leaders and senior engineers at 27 workplaces with aggregate information from CodeRabbit and Gitar. The source article says some material is paywalled and that its cited information comes from the freely readable portion. Its findings are a snapshot of a varied set of organizations, not a controlled industry census.
Orosz reported that spending among companies on API plans peaked in June and July and had roughly halved since, attributing the decline to Opus 5.5 and a move to open models. The article also says spending caps are common but can be circumvented, while some companies expect discounted subscriptions to expire. Separately, GitHub changed Copilot plans on June 1 to usage-based billing, charging model tokens at API rates and removing the fallback to a cheaper model. GitHub said: “Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.”
Research cited in the report offers a less uniform picture than company and vendor output figures. METR’s early-2025 study found experienced open-source developers took 19% longer with AI. Its February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but both confidence intervals included zero. METR described the data as unreliable, in part because 30% to 50% of developers withheld tasks they did not want to attempt without AI.
““A question that few companies seem able to answer – at least for now.””
— Gergely Orosz, The Pragmatic Engineer
developer productivity measurement tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Productivity Gains Remain Uneven
The survey does not establish a single productivity effect that applies across engineering teams. Results range from flat PR counts to substantially higher reported output, and vendor platform data may reflect teams already inclined to use AI review tools. PR volume also does not reveal how much work is accepted, how much rework follows or whether software reliability changes.
METR’s estimates carry wide uncertainty, with confidence intervals that include no effect in its February follow-up. The report also notes a survey of 349 technical workers in which the median self-reported change in work value was 1.4 to 2 times, while METR said participants in its 2025 research overestimated AI’s effect by 40 percentage points. The source excerpt ends during its discussion of a DORA report and Stanford research, so it does not provide enough detail to report that finding fully. No common accounting method or industry-wide return-on-investment figure is established by the material cited.
As an affiliate, we earn on qualifying purchases.
Companies Face 2027 Budget Tests
Engineering and finance leaders will need to decide whether to renew subscriptions, set usage limits or move to different models as they prepare 2027 budgets. The report indicates that some company discounts are due to expire, but it does not give a single timetable or identify all affected firms. Usage-based pricing means actual costs can vary with model choice and how heavily teams use agentic features.
The next useful evidence will come from teams comparing tool costs with accepted work and downstream quality over time. Companies can track changes in rework, delivery failures and completed work alongside spending, rather than treating activity counts as a complete measure. The report does not specify a shared measurement standard or announce a follow-up study, so whether firms adopt comparable measures remains open.
software project management tools for AI development
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The workplaces in The Pragmatic Engineer’s report described budgets from about $150 to $3,000 per developer per month. These are examples from 27 workplaces, not an industry-wide average.
Does the report prove AI coding increases productivity?
No single effect is established. Vendor and company data show higher pull-request output in some settings, while other companies reported flat counts and METR studies produced uncertain or conflicting estimates.
What should companies measure instead of AI usage?
The report points to cost per accepted change, rework, change failure rates and whether completed stories outpace incoming work. These indicators connect spending with delivery and quality.
Why could AI coding costs rise?
Some discounted subscriptions are reportedly nearing expiration, and GitHub moved Copilot plans to usage-based billing on June 1. Costs can depend on token use and model pricing.
What is still unknown about the returns?
The cited material does not establish a common return-on-investment figure, a consistent measurement method or a productivity result that applies across companies. It also does not show whether higher output consistently improves software quality.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
