The Pulse: token spend breaks budgets – what next?

2026-04-30 · Source: The Pragmatic Engineer · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, long

Summary

Token spend for AI tools has surged by approximately 10x in the last six months across companies of all sizes, prompting leadership concerns about sustainability. A survey of developers at 15 businesses reveals two main strategies for managing this increase: "let it rip and start measuring" or "curb spending." Large companies are observing costs "off the charts," with some developers spending up to $500 daily on tools like Claude Code, leading to doubled employee costs and bottlenecks in human code reviews. Mid-sized firms are implementing model routing to cheaper defaults, considering pooled spend models, or factoring AI costs into overall engineering expenses. Smaller companies are exploring options like increasing budgets while measuring ROI, optimizing token consumption, integrating more AI providers, or pivoting to local models. Discounts from vendors like Cursor are available for spending above a few million dollars, but Anthropic currently offers no discounts even for $5M+ annual spend.

Key takeaway

For CTOs and VPs of Engineering grappling with soaring AI token costs, your strategy should balance immediate productivity gains with long-term financial sustainability. Consider implementing a "let it rip and measure" approach initially to capture momentum and quantify ROI, while simultaneously exploring model routing to cheaper defaults for less demanding tasks. Be prepared to negotiate custom discounts with vendors once your spend reaches significant thresholds, and continuously monitor usage to prevent "stupid overspend" without stifling innovation.

Key insights

Rapidly escalating AI token costs are forcing companies to re-evaluate usage strategies and measure productivity impacts.

Principles

• Prioritize measuring AI impact over immediate cost-cutting.
• Default to cheaper models for routine tasks.
• High AI spend can significantly increase engineering output.

Method

Companies are either allowing high AI spend while implementing robust measurement of impact, or actively curbing costs through model selection, usage limits, and exploring local models for long-term control.

In practice

• Implement model routing to cheaper defaults.
• Negotiate custom discounts with AI vendors.
• Track AI adoption and its impact on productivity.

Topics

• AI Token Spend
• Large Language Model Costs
• Developer Productivity
• AI Cost Management
• Claude Sonnet

Best for: CTO, Entrepreneur, VP of Engineering/Data, Software Engineer, AI Engineer, Director of AI/ML

Related on AIssential

• The Pulse: a trend of trying to cut back on AI spend within eng departments? — Companies struggle to link high AI spending to direct productivity gains or user-facing feature delivery, prompting cost optimization efforts.
• How AI is changing Software Engineering: A Conversation with Gergely Orosz, @pragmaticengineer — Measuring AI token output can lead to counterproductive "token maxing" behavior among engineers.
• Executive Briefing: Uber Burned Its Entire AI Budget Early. The Bill Was Trying to Tell Them Something. — High token demand, not just cost, signals useful AI adoption, challenging simplistic ROI metrics for agentic work.
• Let's spend 250K$ on tokens just for sake of spending — The provided content critically analyzes a "flawed logic," attributed to Jensen, which suggests that engineering value is now measured by high AI token consumption rather than efficient, optimized solutions.
• 📈 Data to start your week: The cost of tokenmaxxing — Exponential token consumption is rapidly increasing AI operational costs, challenging traditional budgeting and cost management strategies.
• Meta shifts from "tokenmaxxing" to token managing as internal AI costs reportedly hit billions — Unmanaged internal AI usage can lead to billions in costs and artificial consumption without productivity gains.

Counsel's verdict on this

AIssential's Counsel cites this article in its editorial verdict on the decision it informs:

• Roll AI coding tools across the whole engineering org — and how do we measure it? — Layering multiple AI tools accelerates delivery by 45-50% over single-tool standardization, but unrestricted agentic AI usage scales token costs to $3,000 per developer monthly. Leaders must navigate this tension to avoid unsustainable cost spikes or productivity plateaus.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Pragmatic Engineer.