Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Short answer: A Wall Street Journal report published in October 2023 said GitHub Copilot charged individuals $10 a month while Microsoft’s reported average cost to serve them exceeded $20 per user per month. Some heavy users reportedly cost as much as $80 monthly. Those figures came from unnamed sources and were not published as an audited Microsoft unit-economics statement. GitHub’s pricing in 2026 is substantially more segmented, with premium tiers, AI Credits and usage controls, so the 2023 headline should not be treated as a current loss figure.
What the October 2023 report actually said
The Wall Street Journal’s report, summarized by Thurrott on October 9, 2023 and covered by Ars Technica, described a severe mismatch between Copilot’s individual subscription price and its reported serving cost.
| Item | Figure in the 2023 coverage |
|---|---|
| Individual subscription | $10 per month |
| Reported average cost to Microsoft | More than $20 per user per month |
| Reported cost for some heavy users | As much as $80 per user per month |
| Reported adoption | More than 1.5 million people had used Copilot |
| Reported productivity signal | Copilot helped generate nearly half of some users’ code |
These are attributed figures from reporting based on unnamed sources. Microsoft did not publish a complete per-user cost breakdown in the cited material. “Used” also should not be read as meaning 1.5 million current paying subscribers.
The headline’s $20 arithmetic is not as precise as it sounds
“Loses $20 per user per month” is a widely repeated shorthand, but the public summaries do not make the accounting definition fully clear. The easiest figures to state precisely are that users paid $10 and the reported average cost to Microsoft was more than $20.
#1 Best Overall
If the cost was just over $20, the implied shortfall against a $10 subscription would be more than roughly $10 before other expenses. A $20 loss would require a reported cost of more than $30, or a separate estimate in which “$20” referred directly to the deficit. The available summaries do not resolve that distinction.
The reported $80 figure is a cost for some users, not automatically an $80 loss. Revenue, inference expense, infrastructure overhead and fully loaded product costs are different measures:
- Revenue per user: the subscription amount collected.
- Serving cost: model inference and related infrastructure consumed by requests.
- Gross margin: revenue minus directly attributable costs.
- Fully loaded profitability: gross margin after engineering, support, sales, administration and other expenses.
Why an AI coding assistant can be expensive to serve
Every interaction requires inference
Inline completions, chat answers, code explanations, edits, pull-request reviews and agent tasks all require a model to process a request. A coding request can include open files, surrounding code, repository context, instructions and conversation history, rather than a short standalone question.
Usage is uneven
A casual autocomplete user may generate little traffic. An engineer running repository-wide edits or agent tasks can generate many more requests and larger contexts. A flat or “unlimited” subscription therefore exposes the provider to a wide spread of consumption, with a relatively small group of power users capable of driving a disproportionate share of cost.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteModel and infrastructure choices matter
More capable models generally require more compute than lightweight models. The bill also includes accelerator capacity, model serving, networking, storage, monitoring, reliability engineering and support. GitHub’s current documentation says consumption varies with the feature, model and complexity of the interaction, which is why modern plans distinguish ordinary completions from more intensive workloads.
Why Microsoft might have accepted losses
The reported figures do not establish Microsoft’s motive. Several strategic explanations are plausible:
- Market acquisition: a low, simple price can accelerate adoption before competitors become entrenched.
- Developer ecosystem strategy: making Copilot a routine part of GitHub and supported editors can strengthen the platform’s position with software teams.
- Cross-selling: widespread usage may support higher-value GitHub, Microsoft and Azure relationships, even if the Copilot subscription itself has weak margins.
- Product learning: real-world workflows can inform model, feature and capacity decisions, subject to applicable privacy and product policies.
- Expected cost declines: Microsoft might tolerate an early loss while models, hardware utilization and serving systems improve.
- Price discovery: usage data can reveal which customers value premium models and agent workflows enough to pay more.
These are business analyses, not statements that Microsoft confirmed as the reason for the reported economics.
Copilot pricing: the 2023 flat plan versus the 2026 structure
GitHub’s pricing page, checked in August 2026, lists a broader individual lineup:
The Tool Desk
Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →| Plan | Listed price | Positioning |
|---|---|---|
| Free | $0 per month | Limited completions and chat |
| Pro | $10 per user per month | Everyday coding, agents and unlimited code completion |
| Pro+ | $39 per user per month | More complex development and premium models |
| Max | $100 per user per month | Sustained, high-volume agent workflows |
For organizations, GitHub lists Business at $19 per user per month and Enterprise at $39 per user per month, with administrative controls and organization-level billing. Details are in GitHub’s organization and enterprise billing documentation.
The current structure is consistent with an attempt to align price with intensity: provide a free or lower-cost entry point, charge more for sustained premium-model and agent use, and give organizations ways to manage shared consumption. It does not prove that the 2023 report caused the changes.
How AI Credits and metered usage change the economics
GitHub now uses monthly AI Credit allowances for paid plans. GitHub defines one AI Credit as equivalent to $0.01; the number consumed depends on the feature, selected model and interaction complexity. Code completions and next-edit suggestions remain unlimited on paid plans according to the current comparison, but agent, chat, code-review, CLI and premium-model workloads can draw on included allowances.
After included credits are used, additional paid usage may be enabled subject to a configured budget. GitHub’s usage-based billing documentation describes allowances, budgets and organizational controls. This design reduces the exposure created by selling every expensive workload for one flat fee, while preserving an easy-to-understand subscription for ordinary completion use.
What the report does not prove
- It does not establish that Copilot is still losing $20 per user today.
- It does not provide audited, public unit economics for Microsoft or GitHub.
- It does not show Copilot’s customer lifetime value, enterprise upsell, attributable Azure consumption or strategic subsidy.
- It does not show whether later model and infrastructure efficiencies eliminated the reported shortfall.
- It does not establish Microsoft-wide or Azure-wide AI profitability.
- It does not make all AI coding assistants, or the entire AI industry, unprofitable.
The strongest supported conclusion is narrower: if the Journal’s figures were accurately summarized, the original flat-price Copilot product had a serious unit-economics problem, especially for heavy users.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What individual buyers should examine
Usage intensity
Occasional autocomplete users may find Free or Pro sufficient. Developers who run frequent agents, repository-wide edits or premium models should compare Pro+, Max or metered usage rather than assuming the lowest seat price will be cheapest.
Model and feature needs
Premium models can help with difficult tasks but consume credits faster. Check which models and features each plan includes instead of treating “unlimited” as unlimited access to every Copilot capability.
Bill predictability
A flat subscription is easier to forecast. If paid usage is enabled, set a budget or disable overages when variable charges are unacceptable.
Recommended Free Tools
Privacy and workflow fit
Copilot’s main advantage is integration with GitHub and supported editors. Users seeking only general-purpose chat may not need a coding-specific subscription. GitHub’s pricing page says interactions on certain individual plans may be used to train or improve models unless the user opts out; check the applicable account settings and current privacy terms before subscribing.
What organizations should check
- Compare per-seat fees with actual active usage, not assigned seats alone.
- Review which agent, chat, code-review and CLI actions consume AI Credits.
- Set budgets, policies and approval controls for overage spending.
- Evaluate SSO, audit logs, policy management, intellectual-property protections and deployment environment.
- Remember that GitHub can enable Copilot code review for pull requests from users without assigned Copilot licenses; when the relevant policies are enabled, that usage can still create organization-level AI Credit charges.
- Distinguish GitHub.com, GitHub Enterprise Cloud and other environments before comparing plan capabilities.
Why the story still matters for AI subscriptions
Copilot illustrates a broader tension without proving an industry-wide rule. Flat subscriptions are attractive to buyers, but inference costs vary with request volume, context size, model capability and agent behavior. Heavy-user concentration can make an average look healthy for light users while damaging margins for the provider. Premium tiers, allowances and budgets are ways to segment that risk.
For readers evaluating an AI coding assistant, the meaningful comparison is not just the advertised monthly price. Check included usage, model access, agent limits, overage rates, privacy settings and administrative controls for the workload you actually expect.
Conclusion
The October 2023 report was an early warning about the economics of unlimited or lightly metered AI subscriptions: Copilot reportedly charged $10 while average serving cost exceeded $20, with some users far more expensive. GitHub’s 2026 lineup—Free, multiple paid tiers, AI Credits and budget controls—shows a move toward charging differently for different workloads. It may improve unit economics, but public sources do not establish whether Copilot is profitable today.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
