🔍 Read the full analysis: Is AI Coding Worth The Cost? Where It Pays And How To Measure It on ThorstenMeyerAI.com
Get movie nights 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
Engineering teams in a survey of 27 workplaces reported AI coding budgets ranging from $150 to about $3,000 per developer per month, but many could not quantify the return. Reported results vary: some teams produce more pull requests or clear backlogs faster, while other measures show little change and controlled-study results remain uncertain. The source recommends measuring accepted work, rework and delivery outcomes—not tool usage alone.
The analysis draws on an October 20 report by Gergely Orosz of The Pragmatic Engineer, based on interviews with engineering leaders and senior engineers at 27 workplaces, plus aggregate figures from code-review vendors CodeRabbit and Gitar. The cited evidence is mixed: CodeRabbit reported weekly pull requests on its platform were up 113% year over year, while a Microsoft study found adopters merged roughly 24% more pull requests over four months. Several other companies in the survey reported flat pull-request counts.
Those measures do not establish that teams delivered more valuable software. Microsoft’s study authors cautioned that a merged pull request is not the same as delivered value. One company in the survey kept its PR frequency flat while average PR size tripled. The source also reports that a quant trading firm’s AI spending rose from $500,000 a month in January to $8 million in September, with traders—not engineers—accounting for the highest spending. Its measure is desk-level profit and loss, unlike the many companies that lack a comparable scoreboard.
The analysis identifies customer-issue triage, understanding unfamiliar code, prototyping and backlog cleanup on simple, consistent software stacks as areas where teams reported practical gains. At one SaaS company of about 600 engineers, merged-PR velocity barely changed, but a staff engineer said customer-issue triage could shrink from a day to minutes of agent work plus up to an hour of human review. The source also cites a small C# and Blazor agency that says it is clearing a decade-old backlog, and a multinational team that reports closing three times as many user stories each month.
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 Budgets Need Outcome Measures
Companies are setting 2027 engineering budgets while AI coding bills may change quickly. The source says some employers’ discounted subscriptions are due to expire, and that several could face bills five to ten times higher unless they optimize. It also reports that spending on API plans at some companies has roughly halved from June and July peaks, which Orosz attributes to the release of Opus 5.5 and a move to open models.
Without measures tied to useful work, higher tool usage or more pull requests can be mistaken for productivity. The analysis proposes tracking cost per accepted change, rework, change-failure rates, and stories closed compared with stories arriving. These measures can help teams distinguish cheaper or faster work from output that creates additional review, security or maintenance costs. They are proposed metrics, not evidence that every team has adopted them or that AI tools will improve them.
As an affiliate, we earn on qualifying purchases.
Evidence Behind the Spending Debate
The comparison spans different methods and populations, so the figures are not directly interchangeable. CodeRabbit’s numbers describe activity on its own platform, while Microsoft examined tens of thousands of engineers during an early-2026 rollout of Claude Code and GitHub Copilot CLI. The 27-workplace report includes company accounts and vendor aggregates, rather than a single controlled test.
Controlled research also gives a less settled picture. 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 both confidence intervals included zero. METR said the data was unreliable, in part because 30% to 50% of participants withheld tasks they did not want to do without AI. In a survey of 349 technical workers, the median reported a 1.4-to-2-times change in work value; METR noted participants in its 2025 survey overestimated AI’s effect by 40 percentage points.
Pricing is also changing. 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 described the change as a response to increased agent use and compute demands.
“A question that few companies seem able to answer – at least for now.”
— Gergely Orosz, The Pragmatic Engineer
As an affiliate, we earn on qualifying purchases.
What the Productivity Data Cannot Show
The available figures do not establish a single productivity gain that applies across software teams. Studies use different tools, time periods, worker groups and measures, and reported pull-request growth does not show whether code was useful, safe or costly to maintain. METR’s follow-up estimates were statistically uncertain, while company and vendor reports may reflect selected teams or self-reported results.
The source also cites Veracode’s latest test, in which 44% of AI coding tasks produced a known vulnerability. That result is a test finding, not a measure of the share of AI-written code in production that is vulnerable. The supplied material does not specify enough about the test’s task mix or conditions to generalize the figure to all teams. It also does not provide a common dataset that links each company’s tool spending to accepted work, defects and business outcomes.
AI-powered software development tools
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Teams Can Test Returns
The source does not identify a universal benchmark or a forthcoming study that will settle the return-on-investment question. Its immediate recommendation is for companies to compare AI-related costs with outcomes they already track, including accepted changes, rework, delivery failures and backlog movement, and to account for the human review and security work required.
As discounted plans expire and usage-based pricing puts more spending directly against model consumption, teams will need to review budgets alongside those outcome measures. Whether AI coding pays for a particular company remains an open question until it can show that the work delivered improved without shifting the cost into review, defects or maintenance.
programming productivity enhancement tools
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 27-workplace survey cited by Thorsten Meyer reported budgets ranging from $150 to about $3,000 per developer per month. Spending varies widely, and the figures do not establish a typical cost for all companies.
Do AI coding tools increase developer productivity?
The reported results are mixed. Microsoft found adopters merged roughly 24% more pull requests over four months, while some surveyed companies reported flat counts. METR’s follow-up estimates were uncertain, so the evidence does not support one productivity figure for every team.
Where does the source say AI coding is most useful?
The analysis highlights issue triage, code archaeology, prototyping and backlog cleanup, particularly on simple, uniform software stacks. These examples are reported experiences, not a guarantee of similar gains elsewhere.
What should companies measure to judge whether AI coding pays?
The source recommends tracking cost per accepted change, rework, change-failure rates and stories closed compared with stories arriving. Those measures can provide more useful evidence than tool usage or raw pull-request volume alone.
Does the 44% vulnerability figure mean 44% of AI-written production code is unsafe?
No. The source says Veracode found known vulnerabilities in 44% of AI coding tasks in its latest test. It does not establish that the same share of AI-generated production code is vulnerable.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
