Claude Code Enterprise Usage Continues to Surge as Competitors Accelerate Catch-Up Efforts
Claire Weston
Claude Code's median per-user cost jumped from $69 to $219 in six months, yet enterprise active users tripled over the same period — engineers are voting with their feet while finance teams watch the bills pile up, and that tension is reshaping how companies buy AI coding tools.
Cost tripled — why are companies still paying?
Data from AI analytics provider Weave shows Claude Code's median per-user cost rose from $69 to $219 between January and June this year — more than a threefold increase.
Over the same period, active Claude Code users across Weave's enterprise clients tripled. This means → companies aren't blind to the cost; they've done the math and concluded that the productivity gain still outweighs the bill.
By contrast, Cursor's median per-user cost rose modestly from $38 to $54, yet its active user count fell 17%. In plain terms = cheaper doesn't mean preferred — engineers are scoring tools by output, not price.
How are companies fighting the bill?
IT services firm Workato has roughly 1,300 employees, half of whom use coding tools. 80% of that usage is Claude Code, with engineers spending about $1,000 per month each. CIO Carter Busse says he is "very concerned" about cost but will keep providing access.
Automation software maker Automation Anywhere saw its Claude Code bill double to about $40,000 a month over the past half-year. The team built an internal model router — a system that automatically routes simpler tasks to lower-tier Anthropic models to control spend.
This reflects a shared playbook: companies aren't cutting tools — they're cutting waste, using routing to strip out the tasks that don't need the top-tier model.
Who is switching, and where are they going?
Data-security firm Druva had used Cursor for three years at roughly $1 million a year. When Cursor tried to quadruple the price at renewal, Druva shifted 70% of its AI coding tasks to Claude Code and expects combined spending of about $2 million this year.
Canadian 55-person startup Punchcard Systems switched from GitHub Copilot to Claude Code three months ago after GitHub moved to usage-based billing, erasing its price advantage. This means → Claude Code is currently the "value endpoint" — not the cheapest, but the tool engineers rate highest on output.
Punchcard managing partner Sam Jenkins put it bluntly: "What's best today won't necessarily be best tomorrow — we can't lock ourselves into a single vendor."
How strong is the competition's hand?
OpenAI's Codex is becoming a more credible challenger; Cursor is also accelerating multi-model support, moving beyond Anthropic-only access.
Open-source models from Chinese AI labs are gaining ground fast — Automation Anywhere's chief AI officer Adi Kuruganti says the team is evaluating Kimi, Qwen, and Nvidia's Nemotron, but adds: "Claude is incredibly good right now — my engineers are all using it."
In plain terms = Claude Code's moat today is "engineer preference + frontier model capability," but rivals are closing in on both price and open-source fronts. The next few quarters are the proving ground.
Content is for reference only, not financial advice.