🔍 Read the full analysis: Where The AI Coding Budget Goes—and How To Judge The Return on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A report published by The Pragmatic Engineer on October 20 gathered AI coding budget and output data from 27 workplaces, finding monthly spending ranges from $150 to about $3,000 per developer in the examples cited. Productivity results vary, and there is no shared evidence that higher tool usage reliably produces more delivered value. Engineering teams setting 2027 budgets are advised to track accepted changes, rework and failures rather than token use alone.
A survey of 27 workplaces found that companies’ AI coding budgets range from about $150 to $3,000 per developer per month in the examples reported, while measured output varies sharply between teams. The findings, published October 20 by The Pragmatic Engineer, arrive as companies set 2027 budgets and face questions about whether higher AI spending is producing more valuable software.
The survey, by newsletter author Gergely Orosz, draws on conversations with engineering leaders and senior engineers at 27 workplaces, alongside aggregate data from AI code-review vendors CodeRabbit and Gitar. The source article says its cited material comes from the freely readable portion of Orosz’s report. Spending examples range from a $150 monthly cap at a publicly traded company with 1,500 engineers to around $3,000 per developer at Ix Technologies, a two-person European .NET agency. The report says company size did not predict where spending fell.
Productivity figures do not point to one consistent result. CodeRabbit reported that developers using 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 over four months than they otherwise would have, according to the source article. Microsoft’s researchers cautioned that a merged pull request does not by itself show the value delivered.
Results among the surveyed companies ranged from no change in pull-request frequency to increases of five to ten times, according to the report. In one company, pull-request frequency stayed flat even as the requests became three times larger. The article also cites a Veracode test in which 44% of AI coding tasks produced a known vulnerability. That figure describes the tasks in Veracode’s test; it is not a measure of vulnerability rates across all AI-written code.
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
Budgeting Beyond Tool Usage
The findings matter because higher usage is not the same as higher returns. More pull requests may indicate faster production, but they can also create larger review burdens or require additional security work. Teams that budget based only on tokens consumed, seats purchased or code produced could miss the cost of rework and failures.
The report recommends tracking outcomes such as cost per accepted change, rework, change failure rate and whether completed stories outpace incoming work. These measures can connect spending to delivery and quality, though the source does not establish a single standard for calculating AI coding return on investment. The practical implication is that companies need to define their own baseline and compare results over time before treating higher output as savings or business value.
As an affiliate, we earn on qualifying purchases.
What the Workplace Data Shows
The report says spending on API plans peaked in June and July and had roughly halved by the time of publication for some companies, a change Orosz attributed to Opus 5.5 and a move to open models. Those observations describe the companies discussed, not the whole industry. The article also says spending caps are common and can be easy to work around, while some company discounts are due to expire. One organization reportedly expected its bill to rise five- to tenfold if it did not optimize; that is a reported forecast, not a general price prediction.
Pricing changes add pressure to measure costs. The source says GitHub moved every Copilot plan 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.”
Controlled research offers a more cautious picture than some workplace or vendor figures. METR initially found experienced open-source developers took 19% longer with AI in early 2025. Its February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but the confidence intervals included zero. METR also described its data as unreliable, citing task selection and other limitations. These studies use different settings and methods, so their results should not be treated as directly comparable.
As an affiliate, we earn on qualifying purchases.
How Much Output Becomes Value
The available figures do not establish a universal productivity gain or a reliable return for AI coding tools. Vendor platform data, company reports and controlled experiments measure different groups and outcomes. The source article also provides no common baseline across the 27 workplaces, making their reported changes difficult to compare directly.
It remains unclear how much additional code is accepted, maintained and useful after review, or how frequently it creates security and reliability costs. The Veracode finding is specific to its test, while the reported workplace examples are not a representative industry survey. The source also does not give full details on each company’s methods for attributing changes to AI rather than other factors.
As an affiliate, we earn on qualifying purchases.
The 2027 Budget Test
Companies setting 2027 budgets will need to decide whether to renew discounts, set usage limits or shift models as billing changes take effect. The source report does not provide a single spending target that applies to all engineering teams. Instead, its evidence points toward measuring results within each company, with comparisons tied to the same team, work type and time period.
Teams can start by recording a baseline for accepted changes, review time, rework and change failures, then compare those measures with AI tool costs. They should also distinguish routine work, such as triage or backlog cleanup on a consistent technology stack, from complex legacy work, where the reported gains are less clear. Whether that approach produces a defensible return will depend on data companies have not yet consistently published.
AI vulnerability detection software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much do companies spend on AI coding tools?
The workplaces cited in The Pragmatic Engineer report range from a $150 monthly cap to about $3,000 per developer. These are examples from a 27-workplace report, not an industry-wide price range.
Does the report show that AI coding improves productivity?
It shows mixed results. CodeRabbit and Microsoft reported increases in pull-request output among the groups they studied, while some companies reported flat counts. METR’s research estimates varied and included uncertainty, so the evidence does not establish one general productivity effect.
What should engineering teams measure to judge return?
The source article recommends looking beyond tool consumption to cost per accepted change, rework, change failure rates and whether teams close more work than arrives. Teams should define a baseline and track the same measures over time.
Does the 44% vulnerability figure apply to all AI-generated code?
No. The source cites a Veracode test in which 44% of AI coding tasks produced a known vulnerability. That is a result from that test, not an estimate of the share of all AI-generated code with vulnerabilities.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
