Microsoft study estimates monthly enterprise AI costs
Microsoft says research it commissioned and funded with GigaOm estimates monthly costs for Copilot, Claude and ChatGPT under two modeled enterprise-user profiles. Microsoft reports Copilot had the lowest estimated cost for both profiles, with broadly comparable quality across the products on the tested workload.

Key takeaways · 4
- 01
For the modeled everyday user, Microsoft estimated monthly costs of $30 for Copilot, $205 for Claude and $81 for ChatGPT.
- 02
For the modeled advanced user, the estimates were $96 for Copilot, $299 for Claude and $149 for ChatGPT.
- 03
The comparison included usage charges for competitors, while everyday Copilot chat and Microsoft 365 app use carried no extra usage charges in the study.
- 04
Treat the estimates as directional: Microsoft says actual costs depend on workload, usage, settings and pricing, and excludes negotiated discounts.
What the comparison covered
Microsoft says its research with GigaOm examined Copilot, Claude and ChatGPT using two modeled user types: everyday users and advanced users.[1] The evaluation covered 37 unique prompts across Copilot Chat and Cowork and Copilot in Word, Excel and PowerPoint.[1] The modeled month included 100 tasks for an everyday user and 140 for an advanced user, including 20 agentic workflows.[1] Microsoft says the products were tested against the same fictional company and underlying data, with prompts repeated to account for variation in responses.[1]
Estimated monthly costs
For everyday users, Microsoft estimates monthly costs of $30 for Copilot, $205 for Claude and $81 for ChatGPT, adding usage charges to the competitors’ subscriptions.[1] For advanced users, its estimates are $96 for Copilot, $299 for Claude and $149 for ChatGPT.[1] Microsoft says everyday chat and Microsoft 365 app tasks are included in the Copilot user subscription, so they did not add usage charges in the study; Copilot agentic workflows use usage-based billing through Copilot Credits.[1] The figures use October 2026 list or assumed prices and exclude negotiated discounts.[1]
Quality and evaluation
Microsoft says the three products delivered broadly comparable quality across the tested tasks, but that overall finding does not mean results were equivalent on every task.[1] Human evaluators made a checklist for each prompt covering source accuracy, people, facts and figures, completeness, and unauthorized actions.[1] A task failed the quality evaluation if it was only partially correct.[1] Microsoft says its cost analysis and testing were conducted by Microsoft, while GigaOm helped design the research and validated its methodology, results and cost analysis.[1]
Limits on applying the figures
Microsoft cautions that actual prices may differ from the estimates and describes the reported results as directional.[1] Its stated reasons include differences in task selection and frequency, product settings, list or assumed prices, organizational workloads and negotiated prices.[1] Microsoft also says competitor subscription prices may be $10 below Copilot’s, but modeled usage charges erased that difference within the first week.[1] For a purchasing decision, use the figures as a comparison of this modeled workload, then check costs against the tasks and pricing terms relevant to your organization.
The study offers professionals a concrete cost comparison, but its modeled workload and pricing assumptions matter as much as its headline totals. Use it to frame a pilot or procurement review, not as a substitute for estimating your own usage and checking task-level quality.
Why it matters
AI training for your team
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
How this developed
20 November 2026
Microsoft study estimates monthly enterprise AI costs