🔍 Read the full analysis: How To Weigh AI Coding’s Costs Against Its Benefits on ThorstenMeyerAI.com
Get business pricing on tech for your team
- 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, based on interviews at 27 workplaces and vendor data, found AI coding budgets ranging from $150 to about $3,000 per developer per month. Reported output ranges from flat to sharp increases, but the evidence does not establish a consistent return; teams need to measure accepted work, rework and delivery outcomes alongside cost.
Engineering teams are spending between $150 and about $3,000 per developer per month on AI coding tools, but many cannot show what the spending delivered, according to a report published October 20 by The Pragmatic Engineer. Drawing on interviews at 27 workplaces and aggregate vendor data, the report finds mixed productivity results as companies set budgets for 2027 and face changes to subscriptions and billing.
The spending figures varied sharply and did not track neatly with company size. One publicly traded company with 1,500 engineers caps monthly spending at $150 per developer; a two-person European software agency reported spending about $3,000 per developer. Other companies described caps between $1,000 and $4,000, with typical use around $1,000. A quant trading firm’s monthly AI bill reportedly rose from $500,000 in January to $8 million in September, with traders among its largest users.
Output measures were just as varied. CodeRabbit reported that developers on its platform created 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 than they otherwise would have over four months. The Microsoft researchers cautioned that a merged pull request does not necessarily represent value delivered. At some workplaces in the report, PR counts were flat; at another, they doubled from about four to eight per developer per week.
The report describes practical gains in tasks such as customer-issue triage, understanding unfamiliar code, prototyping and clearing backlogs on consistent technology stacks. It also cites evidence of risks and uneven results: Veracode’s latest test found 44% of AI coding tasks produced a known vulnerability. A METR 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 developers held back some tasks they did not want to do 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 AI Coding Budgets Need Outcome Measures
The report arrives as companies decide how much to spend in 2027 and whether current subscriptions are delivering enough value to keep. Some companies interviewed said discounted subscriptions are due to expire; they expect to optimize spending or risk bills rising five- to tenfold. Those projections are reports from the companies, not a universal forecast.
Counting tool activity or pull requests alone can mislead. More code changes may bring more review work, rework or security exposure without improving what customers receive. The report recommends tracking measures such as cost per accepted change, rework, change failure rate and whether teams close more user stories than arrive. Those measures connect spending to work that is accepted and delivered, rather than to tokens consumed or code produced.
Returns also depend on the work and the codebase. The examples point to clearer benefits in repetitive triage and backlog tasks on uniform stacks, while complex legacy systems are harder to assess. Leaders need to compare like-for-like tasks and account for review and maintenance costs before treating higher output as a business gain.
Products Worth Considering
Jackery Explorer 1000 v2 Portable Power Station
EcoFlow DELTA 3 Max Portable Power Station, 2048Wh LFP Battery
Bose QuietComfort Ultra Earbuds (2nd Gen), Wireless Noise-Cancelling Earbuds with Immersive Audio and Wireless Charging Case, Black
As an affiliate, we earn on qualifying purchases.
Billing Changes and Mixed Productivity Evidence
The Pragmatic Engineer report, by Gergely Orosz, combined interviews with engineering leaders and senior engineers at 27 workplaces with aggregate data from AI code-review companies CodeRabbit and Gitar. The source article says its cited material comes from the report’s freely readable portion. Vendor figures describe activity on their platforms and should not be treated as representative of all engineering teams.
Billing is changing alongside usage. 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.” The article also reports that API-plan spending peaked in June and July and later roughly halved for some companies, which Orosz attributed to a model change and a shift toward open models.
Controlled evidence does not settle the productivity question. METR’s early-2025 study found experienced open-source developers took 19% longer with AI. Its later estimate suggested possible speedups, but the uncertainty was substantial. Self-reported gains were higher: in a survey of 349 technical workers, the median reported a 1.4-to-2-times change in the value of their work. METR said participants had overestimated AI’s effect by 40 percentage points in its 2025 study.
““A question that few companies seem able to answer – at least for now.””
— Gergely Orosz, summarizing the report
How Much Output Becomes Delivered Value
The report does not establish a single productivity gain that applies across companies. Results differ by task, stack, team and measurement method, while vendor platform figures may reflect users who have already adopted AI tools. The Microsoft comparison measures merged pull requests, not their customer or business impact.
It is also unclear how much extra review, rework and security work offsets faster initial coding. The source cites a vulnerability finding from Veracode but does not provide the test’s task count or further methodological detail in the supplied material. The reported five- to tenfold bill increase after discounts end is a concern raised by companies, not a verified outcome for all subscribers. The source’s final discussion of where AI pays is cut off, so no additional conclusion from that section can be confirmed here.
Teams Prepare 2027 Spending Decisions
Companies are setting budgets for 2027 and will need to decide whether to renew discounted plans, adopt spending caps, change models or keep usage-based access. The report does not name a common deadline for those decisions. Teams can use the period to establish a baseline for delivery, review and failure rates, then compare results after introducing a tool or changing a plan.
The next useful evidence will come from workplace-level measurements that connect tool costs to accepted work and account for quality and rework. Until those measures are available, reported increases in pull requests or self-assessed productivity should be treated as partial indicators—not proof that a team’s AI coding spend has paid off.
Key Questions
How much are companies spending on AI coding tools?
The report found budgets ranging from $150 to about $3,000 per developer per month among the workplaces discussed. Spending varies widely, and the report says company size alone does not predict where a company falls.
Does the report show that AI coding reliably increases productivity?
No single result applies across the evidence cited. Some studies and companies report more pull requests or faster work, while other teams report flat counts. METR’s estimates included substantial uncertainty, and a merged pull request does not by itself show business value.
What should engineering teams measure to judge returns?
The report recommends looking at cost per accepted change, rework, change failure rate and whether teams close more user stories than they receive. These measures can show whether output becomes useful, reliable delivery.
Where does AI coding appear most useful in the report?
Examples include triaging customer issues, researching why code was written a certain way, building prototypes and clearing backlogs on relatively simple, uniform technology stacks. The source does not establish that the same gains will apply to every team or legacy codebase.
What is still unknown about the cost of AI coding?
It remains unclear how consistently productivity gains translate into delivered value, and how much review, rework and security work offsets faster coding. The report also does not establish what will happen to every company’s bill when discounts or plans change.
Source: ThorstenMeyerAI.com
