Skip to content
Find a job
Contact us

The AI Tax Is Real: How Token Costs Are Blowing Up Enterprise Budgets

Every time someone at your organization uses an AI tool, a meter is running. Most employees using those tools have no idea. Most organizations paying for them are only beginning to understand what they've signed up for — and by the time the invoice arrives, it's too late to trace where the costs came from.

AI token pricing is now a serious budget line item. In 2025, organizations spent an average of $1.2 million on AI-native applications — more than double the prior year, according to Zylo's 2026 SaaS Management Index. Nearly 8 in 10 IT leaders report being hit with unexpected charges tied to consumption-based AI pricing. And experts are warning that the cost won't stay in the engineering department for long.

Here's what HR and people leaders need to understand about how AI gets priced — and why it's becoming everyone's problem.

What Is an AI Token?

A token is the basic unit an AI model uses to process input and generate output. Tokens can be whole words, partial words, or punctuation. For example, the sentence "Hello, how are you?" breaks into six tokens.

Unlike a human reader who might skim a long document for the relevant parts, an AI model reads everything you send it — every word, every comma, every extra space. That thoroughness is part of what makes AI useful. It's also part of what makes it expensive.

There are two primary token types, and both cost money:

Input tokens are what you send to the model: your question, your instructions, any documents you paste in.

Output tokens are what comes back. Output tokens typically run 3–4x more expensive than input tokens — which means a bloated prompt costs you twice: once for the unnecessary input, and again for the lengthy response it tends to trigger.

A third type — cached tokens — applies when a model stores a representation of content it's already processed. On subsequent requests containing the same material, the model pulls from that stored state rather than reprocessing it from scratch. Cached tokens cost significantly less than fresh input tokens, making prompt caching one of the most effective cost controls available.

AI Token Pricing Is Not Fixed — And the Range Is Wide

All major AI providers, including OpenAI and Anthropic, charge per token on both sides of the conversation. The rate isn't fixed, and the price range is significant.

Simple tasks on basic models can run a few cents per million tokens. Premium models handling complex work can cost anywhere from $20 to over $100 per million. Running a query through OpenAI's reasoning model o1 costs six times more than running the same query through GPT-4o. Newer "reasoning" models that think through problems step by step before answering can require more than 100 times the compute of a standard response.

Pricing models have also multiplied. Organizations today may be managing subscriptions, usage-based billing, performance-based contracts, hybrid structures, freemium tiers, and agentic seat pricing — sometimes across multiple tools simultaneously. Microsoft Copilot appears to cost $30 per user per month, but that price assumes an existing Microsoft 365 subscription, making the true all-in cost substantially higher.

Some vendors are moving away from token pricing entirely. Adobe is shifting its new enterprise AI suite to outcome-based pricing — charging based on what the AI actually accomplishes, such as how many ad campaigns it completes, rather than how many tokens it processes.

Meanwhile, AI has pushed software prices up by 20–37% across vendor categories, regardless of whether customers are getting meaningfully more functionality. Some researchers are calling it an "AI tax."

"The promise of enterprise AI has always been clear: automate more, serve customers better, reduce cost. What was never advertised in the fine print was the bill." — Tarell Harrison, AI Thought Leader, via LinkedIn

Read more via LinkedIn

Token Bloat: The Hidden Cost Driver Most Organizations Don't See Coming

Token bloat is what happens when an AI system processes far more information than it actually needs. It's one of the most common — and least visible — sources of runaway AI costs.

It often starts with anxiety. AI users worry that if they don't give the model everything, it will get things wrong. So they paste in entire documents, database files, and error logs — most of which the model doesn't need. A prompt that is 20% less efficient than it should be can produce costs 200% higher than expected, as the model hallucinates, self-corrects, and burns through tokens trying to make sense of the noise.

Most organizations don't see token bloat until the monthly invoice arrives — at which point there's no way to trace which tool, feature, or prompt was responsible. At enterprise scale, what seems like a small inefficiency becomes a serious financial liability.

Read more via Cloudatler

What Organizations Are Actually Spending

Token costs have moved from line item to board-level concern — with real-world examples making the stakes concrete.

Venture capitalist Chamath Palihapitiya reports that his software startup's AI spending has more than tripled since late 2025, heading toward $10 million annually. The problem, he says, isn't the spending itself — it's that costs are growing 3x every three months while revenue isn't keeping pace.

Uber burned through its entire planned AI budget for 2026 within the first few months of the year, after encouraging engineers to use AI coding tools aggressively and tracking usage on internal leaderboards. The company's R&D spend rose 9% to $3.4 billion in 2025 and is expected to keep climbing.

For now, finance leaders are more permissive than usual — willing to absorb costs they can't fully justify on the bet that efficiency gains will eventually catch up. But that window isn't indefinite.

"Token pricing looks manageable in a demo. A few cents per thousand tokens. A flat monthly subscription that seems reasonable per seat. But the math at enterprise scale tells a different story." — Tarell Harrison, via LinkedIn

Read more via Yahoo Finance

Is High Token Spend a Badge of Honor?

At some tech companies, it's becoming one. Databricks CEO Ali Ghodsi publicly praised an engineer who ran up more than $7,000 in token charges over two weeks — and the engineering department applauded. Customer support startup Sendbird has built a formal ranking system around token consumption, topping out at "AI God" status for engineers burning through at least 100 million tokens per day, with perks including extra vacation time.

Experts push back on this logic. Token volume signals AI-generated work — but what matters is how much of that AI-generated work actually makes it into a finished product. CFOs are starting to notice: some engineers are now generating an extra $2,000 a month in AI charges on top of their salaries.

The cultural dynamic is worth watching for people leaders. Organizations that reward raw usage without measuring output quality are building a cost structure that will be difficult to unwind.

What Organizations Can Do to Manage Token Costs

Token costs are manageable — but only with intention. Most organizations aren't there yet.

  • Send models only what they need. The core principle of token cost management is precision. Pull in relevant information dynamically rather than dumping entire documents into the context. Ask for specific output formats rather than open-ended responses. Use lighter models for tasks that don't require heavy reasoning.

  • Give the model parameters. Simply telling the model how to format its response and how long to make it can cut token usage by 60–80%.

  • Use prompt caching. When a system prompt or document stays the same across many requests, major AI providers can store it so it doesn't have to be reprocessed each time. Prompt caching can cut costs by 50–90%, depending on the model.

  • Track usage before the invoice arrives. The organizations getting ahead of AI costs are building monitoring into their workflows now — not waiting until they've blown through a budget they didn't know they had.

Sources & Further Reading

Get the full interactive edition → View in digital reader

Stay ahead of the market → Subscribe to the Need to Know Briefing to get workforce intelligence delivered to your inbox every week.

About the Need to Know Briefing

The Need to Know Briefing is published weekly by Kelly, curating the most important workforce and hiring insights for HR leaders and hiring managers.
View Related: Article Need to Know Briefing

Let’s solve your workforce challenges today.

We create limitless opportunities by successfully connecting you to the people and solutions you need. Let’s talk about how we can help your business thrive.

Let's Talk!