🔍 Read the full analysis: Does AI Coding Deliver Enough Value To Justify Its Cost? on ThorstenMeyerAI.com
Get smart everyday buys 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, while results on productivity varied widely. Available evidence points to gains in tasks such as issue triage and prototyping, but higher PR counts do not establish business value, and many companies still lack a clear return-on-investment measure.
Engineering teams are spending from $150 to about $3,000 per developer per month on AI coding tools, but a survey of 27 workplaces found no consistent relationship between spending and measurable business results. The findings, summarized by Thorsten Meyer on October 23 from reporting by The Pragmatic Engineer and data from software companies, arrive as organizations set 2027 budgets and face uncertainty about which AI-assisted work produces value.
The Pragmatic Engineer’s Gergely Orosz spoke with engineering leaders and senior developers at 27 workplaces and included aggregate figures from AI code-review providers CodeRabbit and Gitar. The available findings show budgets ranging from a $150 monthly cap at a publicly traded company with 1,500 engineers to roughly $3,000 per developer at Ix Technologies, a two-person European .NET agency. Company size alone did not explain the difference.
Reported output also varied. CodeRabbit says developers on its platform generated 113% more pull requests per week than a year earlier. A Microsoft study of tens of thousands of engineers found adopters merged about 24% more pull requests over four months during an early-2026 rollout of Claude Code and GitHub Copilot CLI. Those measures count activity, not the quality or business impact of the work. Several surveyed companies reported flat PR counts, while one saw PRs become three times larger.
Other evidence is mixed. 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 the confidence intervals included zero; METR also described the data as unreliable. The source summary cites Veracode testing in which 44% of AI coding tasks produced a known vulnerability, underscoring that faster production can bring review and security costs.
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
Where AI Gains Reach Customers
The budget question is not simply whether AI tools help developers write code faster. Companies need to know whether that speed translates into accepted changes, completed customer work, and reliable releases after review, rework and security checks. Without those measures, rising usage or more PRs can make a program look productive even when the team’s delivery outcomes have not improved.
The strongest examples in the report concern bounded tasks. At a SaaS company with about 600 engineers, overall merged-PR velocity barely changed, but a staff engineer said customer-issue triage that could take a day was reduced to minutes of agent work followed by up to an hour of human checking. Code archaeology and prototypes also took less time. Ix Technologies, which works on a uniform C# and Blazor stack, reported using AI to clear a long-standing backlog. These cases suggest results depend on task, codebase and workflow, not just tool choice or subscription price.
Costs may also shift quickly. The report says API-plan spending peaked in June and July and had roughly halved by the time of Orosz’s reporting for some companies, which he attributed to a newer model and moves to open models. Meanwhile, discounted subscriptions are due to expire at some companies. If usage rises or subsidies end, finance and engineering leaders may face materially higher bills without evidence that the extra spend produces more delivered value.
As an affiliate, we earn on qualifying purchases.
Budgets Meet Changing Billing
The October 20 Pragmatic Engineer report comes as organizations begin setting 2027 engineering budgets. Orosz’s account says spending caps are common, but developers can often get around them. It also describes a quant trading firm whose monthly AI spending rose from $500,000 in January to $8 million in September. The firm reportedly evaluates the expense against trading-desk profit and loss, a direct measure that most software teams do not have.
Billing changes add pressure to review the economics. GitHub changed Copilot plans on June 1 to usage-based billing, pricing tokens at each model’s API rate and removing the fallback to a cheaper model. GitHub said, “Agentic usage is becoming the default” and brings higher compute and inference demands. That vendor explanation describes the billing rationale; it does not establish that agent use delivers equivalent productivity gains.
Researchers have cautioned against treating activity metrics as value. Microsoft’s study authors said “a merged PR is not the same as the value it delivers.” METR’s reported speed estimates also carry limitations: its follow-up confidence intervals included zero, and researchers said some participants held back tasks they did not want to attempt without AI.
“Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.”
— GitHub, explaining its Copilot billing change
As an affiliate, we earn on qualifying purchases.
ROI Evidence Remains Uneven
The available figures do not establish a single productivity effect that applies across engineering teams. Vendor data reflects developers already using a code-review product, while company reports use different measures and time periods. PR volume, story completion and time saved are not directly interchangeable, and some reported increases lack a clear baseline or independent verification.
The source material also ends during its discussion of DORA’s ROI report and does not provide the full cited result or its conditions. It is not clear how widely the task-specific gains reported by individual companies apply, how much human review offsets time saved, or how often AI-generated vulnerabilities lead to costly rework. The 44% Veracode figure refers to tasks producing a known vulnerability in its test; it should not be read as a rate for all AI-written code in production.
There is no shared accounting method in the reported material for tool subscriptions, model usage, review time, security work and downstream failures. Those omissions leave companies with no reliable cross-company basis for declaring a particular spend level worthwhile.
As an affiliate, we earn on qualifying purchases.
Track Outcomes Before Renewals
As 2027 budgets are set and discounted subscriptions expire, engineering and finance teams will need to compare AI costs with delivery outcomes. The report points to measures such as cost per accepted change, rework, change-failure rates, and stories closed versus stories arriving. These indicators can show whether a tool is improving a team’s work rather than simply increasing usage or code volume.
Companies will also need to examine results by task and codebase. A controlled comparison between similar work—with review and security effort included—could help distinguish real time savings from larger PRs or shifted work. The source material does not report a single forthcoming industry benchmark or a common measurement standard, so whether companies adopt comparable methods remains unclear.
As an affiliate, we earn on qualifying purchases.
Key Questions
How much are companies spending on AI coding tools?
The 27-workplace survey described budgets ranging from $150 to about $3,000 per developer per month. Spending varies substantially, and the report found that company size alone does not predict where a company falls.
Do the reported increases in pull requests prove AI coding is worth the cost?
No. CodeRabbit reported a 113% year-over-year increase in weekly PRs on its platform, and a Microsoft study found adopters merged about 24% more PRs over four months. Neither measure by itself establishes code quality, customer value or net savings after review and rework.
Where did teams report the clearest benefits?
The examples in the report point to issue triage, code archaeology, prototyping and backlog cleanup, particularly on simple, uniform technology stacks. The reported gains are company examples, not a guarantee that other teams will see the same results.
What should a company measure to judge AI coding spend?
The report suggests tracking cost per accepted change, rework, change-failure rates, and stories completed compared with stories arriving. Companies should include review and security work rather than treating usage or PR counts as a full measure of value.
Source: ThorstenMeyerAI.com
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
