Generative artificial intelligence (AI) is becoming a core productivity tool for companies, but costs that are rising faster than expected are emerging as a new concern. At some companies, spending has surged to the point that annual AI budgets are exhausted within months, creating situations in which firms are forced to choose between hiring and AI investment.

On May 29 local time, CNBC reported that major U.S. companies are paying more attention to cost control and efficient model operations than to expanding AI adoption. AI use is continuing to grow, but how to raise returns on investment is emerging as a new task.

Arvind Jain (아르빈드 제인), chief executive of enterprise AI platform company Glean, cited surging AI costs as the issue corporate customers are most concerned about. "At many companies, annual AI budgets are bottoming out in a month or two," he said. "AI use is increasing, but price sensitivity is also rising faster than expected," he added.

The industry expected AI costs to fall quickly as technology advanced and competition intensified, but reality has been somewhat different. Instead, the latest large language models (LLMs) are arriving with higher performance and higher price tags than previous generations.

Jain said that leading-edge AI models often see per-token costs rise to about double those of earlier models. He said the current structure is difficult to view as sustainable in the long term.

What stands out is that AI costs have begun to compete directly with corporate labor costs. In the past, technology investment was classified as part of operating expenses, but companies now face situations in which they must make real decisions on whether to hire more people or allocate more budget to AI.

"Recently, we are increasingly seeing cases where higher AI budgets are executed in a way that replaces expanded hiring," he said. "Companies increase AI investment in expectation of productivity gains, but that reduces the room to hire more people," he explained.

Factory AI, which provides an AI software development platform, is also feeling similar changes. "Management is now 고민ing whether to optimize headcount or optimize AI costs per employee," said CEO Matan Grinberg. "AI is no longer a symbolic innovation investment. It is turning into a full-scale budget allocation issue," he said.

Companies' AI use strategies are also evolving quickly. Grinberg said that early on, boards pressured management to adopt AI. He said this was followed by a period in which companies used as much AI as possible without paying much attention to costs.

But he said the market has now entered a third stage. Companies are moving into a cost-optimization phase, examining whether every task really needs the most expensive latest model and whether some tasks can be replaced with cheaper models.

The problem is that while AI is producing clear results, it has not yet proved profitability sufficient to offset the pace of cost increases. "AI is a powerful tool but still inefficient," Jain said. "In many cases, the value companies get does not keep up with the pace of rising spending," he added.

He said about 95% of enterprise AI usage is still concentrated on the most expensive top-tier models. But he said many tasks can be handled with much cheaper models even if performance is somewhat lower.

Some analysis says AI costs can be sharply reduced simply through proper task classification and model placement. Jain said using different models by task difficulty could deliver cost savings of up to 10 times.

As a result, the axis of competition in the enterprise AI market is changing. Until now, the key question was who could release a more powerful model, but going forward, a more important competitive factor could be which models are assigned to which tasks to raise cost efficiency.

The industry sees the AI market moving from a performance race to a race on economics. As companies begin to scrutinize returns on investment more strictly than performance improvements in top-tier models, the shift is expected to have a significant impact on the pace of growth in the AI industry and on corporate valuations.