This past June, the CEO of hiring platform Mercor revealed that the company's AI token costs had, for the first time, exceeded its total employee payroll. In April, Nvidia executives said their own internal calculations showed the same thing: AI spending outpacing salary costs. What sounded like isolated remarks got outside confirmation in June. According to the AI Index published by Ramp, which aggregates corporate card spending data, the top 1% of US companies most aggressively adopting AI are now spending more than $7,500 per employee per month on AI-related costs — roughly 10 million Korean won. The question this piece asks is what that number should mean for solo founders and working managers in Korea.
What $11.38 and $7,500 Say About Two Different Worlds
The Ramp AI Index breaks the numbers into three tiers: $7,500 per employee per month for the top 1% of companies, $611 for the top 10%, and just $11.38 for the overall median.
To put that median in concrete terms: a single ChatGPT Plus subscription costs $20 a month, so the typical company is spending roughly half of that. That's about what you'd expect from occasional use of a free plan, or a single Copilot license auto-bundled into a work email account. These companies are using AI, but it isn't woven deeply into how work actually gets done.
The $7,500 tier represents a different kind of spending altogether. This is where automation pipelines built on direct API calls, multimodal processing, and operations that benchmark multiple models against each other to optimize cost and performance actually run. Ramp notes that these companies aren't locked into a single vendor — they manage costs by blending open-source models with commercial APIs. Under the same label of "AI-adopting company," wildly different realities coexist.
The rate of change matters too. AI spending among the top-tier companies is growing 14.1% month over month. Compounded monthly, that would put per-employee spending at those same companies at roughly $36,000 a year from now. The average US software engineer currently earns about $16,000 a month. Today's $7,500 is still less than half an engineer's monthly pay, but Ramp attaches a caveat to that fact: "yet." The point isn't where the number sits now, but where it's headed.
Bigger Spending Doesn't Mean Proportionally Bigger Value
Treating these figures as a benchmark to chase calls for caution. Criticism that the link between AI spending and actual performance remains unclear is coming from multiple directions.
In a 2025 McKinsey survey, more than half of companies that had adopted AI tools said they couldn't quantitatively measure any productivity gain. Gartner has projected that roughly 30% of enterprise AI projects will be scaled back or shut down after failing to deliver expected results. The fact that spending is rising and the fact that it's producing real output are, for now, two separate stories.
There's also the fact that subscription numbers don't equal actual usage. It's hard to deny that the signal of "we're using AI" alone creates a positive impression, whether aimed at employees or at investors and customers. That's why some organizations subscribe to more tools than they actually need, or treat adoption itself as the goal. When spending functions as a signal rather than a strategy, chasing the number ends up as wasted money.
The more the top 1%'s spending figures get repeated in the press, the greater the risk that more organizations rush into adoption without adequate scrutiny. A big number becomes pressure in itself.
When AI Costs Line Up Next to Labor Costs
Even so, there's a separate reason this data matters for solo entrepreneurs and small teams in Korea: AI spending is starting to shift categories, from "software subscription fee" to "cost of replacing labor."
Until now, AI tools were budgeted as an add-on that made existing work more convenient — a SaaS fee running a few tens of thousands of won a month, booked as a line item entirely separate from payroll. But as API-based automation spreads, the nature of that spending changes. When AI is handling recurring research, drafting, data cleanup, or customer responses, that cost isn't a software expense — it's the cost of replacing a specific role's labor.
Long-range analyses tracking how work is changing through 2030 argue that human-AI collaboration is moving past simple tool use into a phase that reshapes role structures themselves. The core question in organizational design becomes not which tasks AI handles, but which judgment calls and responsibilities people retain. Ramp's numbers show that shift is already starting to show up as a line item on the books.
I'd argue this math cuts even sharper in the Korean market. Few people have actually compared the cost of hiring an experienced employee, freelance contractor rates, or the internal labor burned on repetitive tasks against today's AI operating costs. $7,500 is still a substantial sum for a small or mid-sized Korean business, but the cost of experimenting — starting from that $11 median — to find out which tasks can be automated, and how fast, is far lower.
There are two things worth checking. First, identify whether any of the tasks you're currently outsourcing repeatedly or burning internal staff time on could be handled by an AI workflow — and if so, pin down what fraction of the outsourcing rate that processing would cost. Without that calculation, there's no way to judge whether AI spending is expensive or cheap. Second, track whether the quality of actual output or processing speed rises in proportion as AI spending scales up. More subscriptions and better results are not the same thing.
There's no reason to chase the $7,500 figure, and no reason to avoid it either. What the number says is that certain companies have started treating AI as part of their workforce structure rather than a software line item. Whether that judgment is correct is still being tested, but the fact that it's already showing up as a cost is visible in the numbers. The people who can actually put that figure to use are the ones who first decide where and how that judgment applies inside their own organization.
