🔍 Read the full analysis: When Does AI Coding Pay For Itself? on ThorstenMeyerAI.com
Get the little things that make your day 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 survey of engineering leaders and senior engineers at 27 workplaces found AI coding budgets ranging from $150 to about $3,000 per developer per month, with no consistent answer on return. Available data shows gains in some tasks and teams, but higher output does not automatically mean more delivered value; companies still need measures such as accepted changes, rework and failures.
A survey of 27 workplaces found monthly AI coding budgets ranging from $150 to about $3,000 per developer, while companies reported sharply different changes in output and few could explain what the spending had delivered. The findings, published October 20 by The Pragmatic Engineer’s Gergely Orosz, come as companies prepare 2027 budgets and some discounted subscriptions near expiration.
The reported budgets do not track neatly with company size. One publicly traded company with about 1,500 engineers caps spending at $150 per developer per month. Ix Technologies, a two-person European .NET agency, spends about $3,000 per developer. Other companies reported caps from $1,000 to $4,000, with typical use around $1,000. These are examples from the surveyed workplaces, not a representative industry-wide average.
Productivity figures also vary, a topic explored in AI coding tools. CodeRabbit reported that developers on its platform produced 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. Microsoft cautioned that a merged pull request does not by itself show the value delivered. In Orosz’s workplace sample, reported outcomes ranged from flat pull-request frequency to increases of five to ten times; at one company, pull requests became three times larger without an increase in frequency.
Other evidence on AI coding performance is less conclusive. METR’s early-2025 study found experienced open-source developers took 19% longer with AI, while a February follow-up estimated an 18% speedup for returning developers and 4% for new developers. METR said both estimates had confidence intervals including zero and described the data as unreliable, in part because some participants withheld 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
Why Output Does Not Equal Return
The gap between more code and more useful work is central to whether these tools pay for themselves. Pull-request totals can rise while review work grows, changes become larger, or defects reach production. Orosz’s report points to triage, code archaeology, prototyping and backlog cleanup as areas where teams described clear time savings, especially on simple, consistent technology stacks. One staff engineer at a SaaS company said investigating customer issues could shrink from much of a day to minutes of agent work plus up to an hour of human checking.
Costs and risks also extend beyond subscription fees. The source cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability. That result is a test finding, not a measure of the vulnerability rate in every company’s code. Still, it underscores why organizations need to track security review and rework alongside speed. A tool can save drafting time yet fail to produce a net gain if people spend that time checking, correcting or securing its output.
For readers responsible for engineering budgets, the practical implication is to compare spending with outcomes rather than token use or raw code volume. The report recommends measures such as cost per accepted change, rework, change failure rate and whether teams close more user stories than arrive. Those measures can show where a tool is helping a workflow and where extra activity is not translating into delivered work.
As an affiliate, we earn on qualifying purchases.
Budgets Tighten Before 2027
Orosz’s October 20 report drew on conversations with engineering leaders and senior engineers at 27 workplaces, plus aggregate data from AI code-review vendors CodeRabbit and Gitar. The article’s freely readable material is the basis for the findings summarized here; it does not establish a controlled comparison across all companies or tools.
According to Orosz, spending on API plans peaked in June and July and has since roughly halved at some companies, which he attributes to a move to open models and the arrival of Opus 5.5. Several companies also told him their discounted AI subscriptions are due to end, potentially raising bills five- to tenfold if usage and pricing remain unchanged. Those are reported company expectations, not confirmed future increases for every customer.
Billing arrangements are changing as well. GitHub said it 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. The company explained: “Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.” That shift makes actual usage and cost exposure more relevant to budget planning.
““a question that few companies seem able to answer – at least for now.””
— Gergely Orosz, The Pragmatic Engineer
How Much Value Reaches Production
The evidence does not establish a single productivity gain that applies across engineering teams. Vendor platform figures, company reports and studies measure different populations and outcomes, so they cannot be treated as directly comparable. METR’s estimates, in particular, carry substantial uncertainty, while self-reported gains may overstate effects: the source says METR found participants in a 2025 survey overestimated AI’s effect by 40 percentage points.
It also remains unclear how many reported time savings become faster releases, reduced costs or improved product quality. The source gives examples of security findings and review concerns but does not provide a common measure of total review time, defect rates or long-term maintenance costs across the surveyed companies. The spending range likewise reflects selected workplaces rather than a market-wide census.
What Teams Should Measure Next
Companies setting 2027 budgets will need to test costs against work completed, particularly as discounts end and usage-based billing makes consumption more visible. The report’s suggested measures include cost per accepted change, rework, change failure rates and the balance between user stories completed and incoming work. Tracking those figures by task and team may help distinguish useful gains in triage or backlog work from increases in code volume that add review burden.
No common industry benchmark or definitive return calculation is established in the available findings. The next useful evidence will come from organizations publishing comparable results that include quality, security and maintenance—not just output counts. Until then, companies should treat productivity claims as context-specific and make spending decisions against their own measured outcomes.
Key Questions
How much are companies spending on AI coding tools?
The 27 workplaces in The Pragmatic Engineer’s report described budgets ranging from $150 to about $3,000 per developer per month. The sample is not a representative industry-wide survey.
Do the findings show that AI coding increases productivity?
Some sources reported higher output, including more pull requests at CodeRabbit and Microsoft. Other companies reported flat pull-request frequency, and METR’s speed estimates were uncertain. Results depend on the team, task and measure used.
Which engineering tasks showed clearer gains?
Companies described benefits in customer-issue triage, understanding existing code, prototyping and backlog cleanup, particularly on simple, uniform technology stacks. These are reported experiences, not a guaranteed result for every team.
How should companies measure whether the tools pay off?
The report recommends tracking cost per accepted change, rework, change failure rates and whether teams close more user stories than they receive. Pull-request counts or token use alone do not establish business value.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
