
What happened
Canadian Imperial Bank of Commerce found that its most powerful models, including Anthropic's Opus, can cost $2,130 per employee a month — a bill that could add up quickly across thousands of workers.
Why it matters
An unpredictable bill of that size puts technology budgets under pressure, and the computing power each agentic AI task consumes is hard to budget for.
WHO IT HITSBank technology and finance teams that must budget for AI tools now face per-employee costs that can run into the thousands of dollars a month, with no reliable way to predict spending per task.
Summaries like this, in your inbox every morning.
The sticker shock at Canadian Imperial Bank of Commerce is the concrete case that opens a broader budgeting problem in banking technology. Technology budgets there are already under pressure from competing demands such as legacy IT modernization and cybersecurity, so a per-employee AI cost that can reach $2,130 a month lands on an already stretched line item rather than on a blank check.
The difficulty is structural. Agentic AI — systems that tackle complex tasks autonomously — is where some of the greatest hopes for productivity improvements sit, yet the computing power each task consumes can be unpredictable. As Bhavi Mehta of Bain & Co. put it, nobody in the organization has a good handle on what a given task will cost, and the tokens required can vary significantly as AI moves toward more sophisticated reasoning and agentic workflows. That makes understanding cost alongside business value increasingly important as banks scale AI, according to Mehta.
The early rush to embrace AI, when technology teams were effectively handed a blank check, has given way to an imperative to control costs after one-too-many surprise bills. How banks reconcile that imperative with the productivity hopes pinned on agentic AI is likely to shape their technology spending choices.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Tesla reported third-quarter capital expenditures of $4.73 billion, short of the $7.95 billion analysts expect…

Calif used GPT-5.6 Sol to uncover the critical PHP vulnerability PHP2Shell (CVE-2026-109618/CVE-2026-109619)…

Ming Lu argues AI-related capital spending has surged to about 2% of U.S

The author details a setup where Claude Code reads an entry document called CLAUDE.md, writes results into rol…

Google opened Gemini 4 Argon to the public on October 13, 2026, after a September 30 announcement limited at f…

Using atopile, KiCad, Freerouting and Claude Code, the developer automated circuit description, placement, rou…
