🔍 Read the full analysis: A Closer Look At AI Coding Costs, Benefits, And ROI on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
A survey of 27 workplaces found AI coding budgets ranging from $150 to about $3,000 per developer per month, with no consistent link between company size and spending. Output measures vary widely, and the source material says most companies cannot yet quantify whether the added cost produces business value.
Engineering teams are spending from $150 to about $3,000 per developer per month on AI coding tools, but a report drawing on interviews at 27 workplaces finds no common way to show what those budgets deliver. The figures come as companies set 2027 budgets and some discounted subscriptions are due to end, putting pressure on leaders to connect tool costs with accepted work and business results.
The October 20 report by Gergely Orosz of The Pragmatic Engineer combines conversations with engineering leaders and senior engineers at 27 workplaces with aggregate figures from AI code-review vendors CodeRabbit and Gitar. The source article says its discussion relies on the report’s freely readable material. Spending varied sharply: one publicly traded company with 1,500 engineers limits monthly spending to $150 per developer, while a two-person European .NET agency reported spending about $3,000 per developer. Other companies described caps between $1,000 and $4,000, with typical use around $1,000.
The evidence on output is mixed. CodeRabbit reported that developers on its platform created 113% more pull requests per week than a year earlier. In a separate Microsoft study covering tens of thousands of engineers during an early-2026 rollout of Claude Code and GitHub Copilot CLI, adopters merged about 24% more pull requests over four months than they otherwise would have, according to the study’s authors. Those measures do not establish that the added work generated equivalent value. The source also reports companies where PR counts stayed flat, and one where PRs became three times larger.
Controlled research cited in the source is less conclusive. METR’s February follow-up estimated an 18% speedup for returning developers and 4% for new developers, but confidence intervals included zero; METR also described the data as unreliable. Its earlier 2025 study found experienced open-source developers took 19% longer with AI. Separately, Veracode’s latest test found 44% of AI coding tasks produced a known vulnerability, a result that underlines the need to account for security and review rather than counting generated code alone.
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 gap between spending and measurable value matters as AI tools shift toward heavier agent use and companies prepare budgets for the coming year. The source reports that spending peaked in June and July and had roughly halved since for companies on API plans, which Orosz attributes to Opus 5.5 and a move to open models. It also says some firms expect discounted subscriptions to expire; one company warned its costs could rise five- to tenfold if it does not optimize. Those are reported expectations, not a general forecast for every buyer.
Higher PR counts can indicate faster work, but they do not by themselves show that customers receive better software or that maintenance costs fall. The report’s examples suggest that benefits may be clearer in triage, code archaeology, prototyping and backlog cleanup on relatively uniform stacks than in complex legacy systems. Teams also need to count review time, rework, failures and vulnerabilities. The source recommends tracking cost per accepted change, rework, change failure rate, and stories closed compared with stories arriving.
As an affiliate, we earn on qualifying purchases.
How Costs and Output Diverge
Reported workplace results range from no change in PR frequency to increases of five to ten times, according to the source article. At one SaaS company with about 600 engineers, merged-PR velocity barely shifted, yet a staff engineer described customer-issue triage falling from as much as a day to minutes of agent work plus up to an hour of human checking. The same engineer said tracing why code had been written a certain way could take minutes instead of a day or two, and a complex product prototype could take hours rather than one to two weeks. These are company-reported examples, not controlled measurements.
A different case involved Ix Technologies, which builds in C# and Blazor. Its co-founder, Rogier van der Hee, said the two-person agency ships about 60 pull requests per developer per week and is working through a backlog said to date back roughly ten years. The source also describes a multinational team closing three times as many user stories per month, for the first time exceeding new incoming stories. These examples illustrate possible gains in specific settings; they do not establish a typical return across organizations.
Billing changes add to the budget question. 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 explained the change by saying, “Agentic usage is becoming the default” and brings higher compute and inference demands. The source also describes spending at one quant trading firm rising from $500,000 a month in January to $8 million in September, with traders rather than engineers among its largest spenders. That firm’s desk-level profit and loss offers a direct measure the source says many other companies lack.
““a question that few companies seem able to answer – at least for now.””
— Gergely Orosz, The Pragmatic Engineer
Returns Remain Hard to Compare
There is no single, comparable productivity result across the material. Vendor figures describe users on particular platforms, company examples use different work and measurement methods, and controlled-study findings are uncertain. METR said its estimates had confidence intervals that included zero and flagged study limitations, including participants holding back tasks they did not want to do without AI. METR also reported that surveyed workers’ self-assessments overstated AI’s effect in 2025 by 40 percentage points.
The source does not provide a full accounting of software costs, staff time spent reviewing AI output, security incidents, or the long-term maintenance burden. It also does not establish whether the reported improvements persist as teams change tools, models or workflows. The 44% vulnerability result is from Veracode’s test of AI coding tasks; the source does not specify enough detail here to treat that rate as representative of all AI-assisted software or teams.
Track Results Before Renewals
Companies setting 2027 budgets will need to compare tool bills with outcomes their teams can verify. The source recommends measuring cost per accepted change, rework, change failure rate, and stories closed versus stories arriving, rather than relying on token consumption or raw PR totals. Teams should also account for review and security work, particularly where usage-based billing or the end of discounts could raise costs.
The cited material does not announce a new industry-wide standard or provide a date for one. The next useful evidence will come from company-level tracking that follows changes through review, release and maintenance, alongside further independent studies that clarify where productivity gains hold. Until then, the report’s examples show possible benefits in specific workflows, while broad claims about return on investment remain unproven.
Key Questions
How much are companies spending on AI coding tools?
The workplaces covered by the report described budgets from $150 to about $3,000 per developer per month. These are reported examples, not a representative average for all companies.
Does AI coding increase developer productivity?
Some measures show increases, including CodeRabbit’s report of 113% more weekly pull requests on its platform and a Microsoft study reporting about 24% more merged PRs among adopters. Other companies reported flat PR counts, and the available measures do not prove equivalent gains in delivered value.
Where did teams report the clearest benefits?
Examples in the source point to customer-issue triage, understanding existing code, prototypes and backlog work on relatively uniform technology stacks. These are reported cases and may not apply to complex legacy systems or other teams.
What should companies measure to judge ROI?
The source recommends tracking cost per accepted change, rework, change failure rate, and stories closed compared with stories arriving. It also points to review effort and security risks as costs that raw output counts can miss.
Is there a proven average return on AI coding tools?
No average return is established in the cited material. The report describes wide differences among companies, while some controlled-study estimates remain uncertain and many organizations lack a direct way to tie spending to business outcomes.
Source: ThorstenMeyerAI.com
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
