🔍 Read the full analysis: AI Coding’s Financial Trade-Offs: Costs, Value, And Measurement on ThorstenMeyerAI.com
Prime made for students and young adults
- Fast, free delivery for dorm and study essentials
- Prime Video and Amazon Music included
- Member-only deals
TL;DR
A survey of engineering leaders and developers at 27 workplaces found AI coding costs ranging from $150 to about $3,000 per developer monthly, while reported productivity results varied widely. The source points to practical gains in triage and backlog work, but says many companies lack measures linking tool use to delivered value.
A survey of 27 workplaces found monthly AI coding budgets ranging from $150 to about $3,000 per developer, while companies reported mixed changes in output and often could not say what their spending delivered. The findings, published October 20 by The Pragmatic Engineer, arrive as engineering teams prepare 2027 budgets and some discounted subscriptions approach renewal.
The survey by The Pragmatic Engineer editor Gergely Orosz drew on interviews with engineering leaders and senior developers, as well as aggregate information from AI code-review vendors CodeRabbit and Gitar. The source article says it relies on the freely readable portion of Orosz’s report. Spending limits and actual costs varied sharply: one publicly traded company with 1,500 engineers caps monthly spending at $150 per developer, while a two-person European .NET agency, Ix Technologies, spends about $3,000 per developer. Other companies reported caps of $1,000 to $4,000, with typical spending around $1,000.
Output measures did not line up across organizations. CodeRabbit said developers using its platform produced 113% more pull requests per week than a year earlier; that figure reflects a vendor’s users, not a controlled comparison across all engineering teams. A Microsoft study covering 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. The study’s authors cautioned that a merged pull request does not necessarily represent value delivered.
Other companies reported flat pull-request counts, larger changes rather than more frequent ones, or increases ranging from roughly twofold to five- or tenfold. The report also cites mixed experimental evidence: METR’s February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but both estimates had confidence intervals that included zero. METR described the data as unreliable, in part because some participants avoided tasks they did not want to attempt 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
The Cost of Measuring Output
The findings matter because higher tool usage or more pull requests do not, by themselves, show that a company is getting more useful work for its money. Larger changes can increase review burdens, and output counts do not reveal whether software is reliable, secure, or helping customers. Teams making budget decisions need to distinguish activity from results they value.
The source describes clearer returns in work such as customer-issue triage, code archaeology, prototyping and backlog cleanup, especially on simple, consistent technology stacks. A SaaS company with about 600 engineers reported little change in merged-pull-request velocity but said customer issue triage could shrink from a day to minutes of agent work plus up to an hour of human checking. A multinational team reported closing three times as many user stories per month, while Ix Technologies said its uniform C# and Blazor stack helped it work through a long-standing backlog.
Those cases are company accounts, not proof that the same gains will apply elsewhere. Results may differ with codebase complexity, review practices and the nature of the work. The report warns that costs can rise when token use becomes a target, pull requests grow without improving, or security and review capacity fail to keep pace. It cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability; the source does not provide further detail here about the test design, so the figure should not be treated as a universal rate for AI-generated code.
Popular right now
Ninja Foodi DZ201 8-Quart DualZone 6-in-1 Air Fryer
Dyson V9 Motorbar Cordless Vacuum
De'Longhi Classic Signature Espresso Machine with Adjustable Milk Frother
As an affiliate, we earn on qualifying purchases.
Why Budgets Are Being Revisited
Orosz’s October 20 report comes as companies decide what to fund in 2027. According to the source article, API-plan spending peaked in June and July and had roughly halved by the time of reporting for some companies. Orosz attributed that decline to the use of Opus 5.5 and a shift toward open models. The source does not provide a common spending dataset for all companies, so this should be read as a reported pattern, not an industry-wide measurement.
Some companies also said discounted subscriptions were due to expire, potentially raising their bills by five to ten times if they made no changes, according to the report. GitHub changed Copilot billing on June 1, moving all plans to usage-based charges tied to each model’s 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.” These pricing shifts make it harder for teams to forecast costs using a simple per-seat subscription.
““a question that few companies seem able to answer – at least for now.””
— Gergely Orosz, The Pragmatic Engineer
Returns Still Differ by Team
The available figures do not establish a single, reliable productivity gain across engineering teams. The results come from different methods and populations: vendor platform data, a Microsoft study, company interviews and experiments with limitations. They measure different things, and some do not establish whether additional output led to better software or greater business value.
It also remains unclear how often reported gains persist after discounts expire, how AI affects maintenance and security costs over time, and whether teams can absorb larger or more frequent changes without adding review work. The source article offers no shared accounting method for comparing total tool costs with accepted changes, rework or customer outcomes. Its reported vulnerability figure also lacks enough methodological detail in the material provided to assess how broadly it applies.
Track Results Before Renewals
Companies approaching subscription renewals and setting 2027 budgets will need to decide whether to keep, limit or change their AI coding plans. The source recommends measuring cost per accepted change, rework, change failure rate, and stories closed versus stories arriving, rather than treating token consumption or pull-request counts as evidence of return.
Those measures can help teams compare tools and workflows with their own prior results, but the source does not identify a common industry benchmark or a forthcoming study that will settle the question. For now, companies’ own operational data and the conditions of their codebases remain central to judging whether AI coding costs are paying off.
Key Questions
How much are companies spending on AI coding tools?
The report describes monthly budgets ranging from $150 to about $3,000 per developer among surveyed workplaces. Other companies reported caps of $1,000 to $4,000, but the figures are not a representative industry-wide average.
Does AI coding increase developer productivity?
The evidence in the report is mixed. Some studies and companies reported more pull requests or completed work, while others saw flat counts. METR’s estimates had confidence intervals that included zero, and its researchers described the data as unreliable.
Where did companies report the clearest gains?
The source describes practical gains in triage, code archaeology, prototyping and backlog cleanup, particularly in simple, uniform codebases. These are reported company experiences, not guaranteed outcomes for every team.
How should engineering teams measure AI coding’s value?
The source recommends tracking measures tied to outcomes, including cost per accepted change, rework, change failure rate and stories closed compared with stories coming in. Pull-request volume and token use alone do not show whether the work delivered value.
Source: ThorstenMeyerAI.com
