Nvidia CEO Jensen Huang has a simple test for whether an engineer is worth keeping. Speaking on the All-In Podcast at GTC 2026, he said that if an engineer’s annual AI token consumption came in under half their salary, he would be “deeply alarmed.” Nvidia itself is working toward a $2 billion yearly token bill for its engineering force.
That might sound like a problem only for Silicon Valley giants. But the same dynamic is quietly affecting businesses of every size. The companies selling AI tools are spending nearly $700 billion on AI infrastructure this year — almost double last year. That money has to come from somewhere, and for many businesses, it is coming from budgets you might recognise: payroll, software subscriptions, and marketing spend.
The layoff trap
Data from outplacement firm Challenger, Gray and Christmas shows AI was the most-cited reason for US job cuts for a record fourth consecutive month. Meta cut 8,000 roles in May, describing the move as offsetting AI investments — despite revenue growing 33% that quarter. These are not survival measures. They are financing decisions.
Here is the problem: the financing is not buying what it promises. Gartner surveyed 350 executives at companies deploying AI agents and found that roughly 80% had cut headcount with no correlation to improved returns. Analyst Helen Poitevin put it bluntly: “Workforce reductions may create budget room, but they do not create return.”
Uber learned this the expensive way. The company gave 5,000 engineers AI coding tools in December and exhausted its entire 2026 AI budget by April. Chief Operating Officer Andrew Macdonald admitted that despite 70% of committed code being AI-generated, the connection to anything customers notice is “not there yet.”
Where the real savings live
The good news is that token spending is more flexible than most business owners realise. The cheapest fix is also the least glamorous: stop paying to process the same text repeatedly. Prompt caching — now standard across major AI providers — cuts the cost of repeated inputs by up to 90% because static content like instructions and reference documents get processed once and then reread at a fraction of the rate.
Security firm ProjectDiscovery documented raising its cache hit rate from 7% to 84% by restructuring its prompts. That single engineering exercise cut their total AI spend by 59 to 70% while serving nearly 10 billion tokens from cache. They recovered more budget than most AI-attributed layoff rounds save.
Other levers include routing simple tasks to cheaper, smaller models rather than always using the most expensive option, and batching work that does not need an instant response. Together, these techniques can cut AI costs by half or more — without letting a single person go.
What Klarna learned the hard way
The other half of the fix is human. Klarna ran the experiment everyone was watching: it replaced roughly 700 customer service roles with an AI assistant. Customer satisfaction fell. Chief Executive Sebastian Siemiatkowski admitted publicly: “The result was lower quality, and that is not sustainable.” Klarna now runs a blended model, with AI handling routine volume and humans handling everything requiring judgement. Gartner predicts that by 2027, half the companies that cut customer service staff for AI will end up rehiring them.
What Irish businesses should do
You do not need a $700 billion budget to manage AI costs wisely. Start by tracking what you are actually spending on AI tools per month — you may be surprised. Many small businesses sign up for premium AI subscriptions and use only a fraction of what they pay for.
Audit whether you are paying for premium AI features that simpler, cheaper tools could handle. A free or low-cost AI tool can often handle basic tasks like drafting emails, summarising documents, or generating social media posts just as well as an expensive enterprise platform.
And before you cut staff to fund AI, ask whether the efficiency gains are real or theoretical. The companies that come out ahead will not be the ones that spent the most on tokens or cut the most people to afford them. They will be the ones that treated the token budget as something to engineer, not something to endure.