The cost of accessing artificial intelligence (AI) services has been falling rapidly, even as the expense of producing the underlying technology continues to rise, creating tension within the market over who ultimately benefits from AI’s expansion.

In recent weeks, a free AI model called Ox Alpha was released online, delivering performance close to the cutting edge without any disclosed developer. Concurrently, OpenAI has reduced prices on its leading AI model three times in about a month. Despite these price cuts, metrics tracking AI token usage and expenditure, such as the Silicon Data LLM Token Expenditure Index, continue to decline, signaling a decrease in what buyers are willing to pay for AI access.

This trend—falling prices for AI output—is common as new technologies become widely adopted. Typically, advances in chip manufacturing reduce the cost of generating AI tokens more quickly than providers lower their prices, allowing profit margins to expand. However, the current environment shows the opposite pattern: AI service prices are dropping faster than the costs associated with building and running the technology, raising concerns over profitability.

The semiconductor industry is facing a tightening supply environment. Foundries and memory manufacturers are limiting output amid strong demand, pushing prices higher. Nvidia has informed some customers that server prices incorporating its AI chips will rise more than 15% for systems shipping early next year. The company remains optimistic, projecting roughly 70% revenue growth by fiscal 2028. Samsung Electronics has increased prices by up to 15% on advanced chip manufacturing contracts and is locking much of its memory capacity into long-term deals with data-center buyers. Meanwhile, SK Group has warned of a worsening memory shortage in 2027 as SK Hynix explores joint ventures to finance new fabrication plants. This supply-driven price increase in semiconductor inputs contrasts sharply with the declining prices for AI services.

Proponents argue that lower AI service prices drive higher volume and adoption. The share of U.S. businesses paying for AI approaches 60%, with spending tripling across various user segments. Major cloud providers collectively reported approximately $106 billion in revenue last quarter, a 40% year-over-year increase. Data suggests a highly variable spending pattern among businesses, from about $12 per employee monthly at the median to $7,400 at the highest end. This indicates that price reductions may be part of the business model rather than a threat to it.

Despite these trends, quantifying AI’s financial return remains challenging. A recent study analyzing 919 earnings calls from 60 large U.S.-listed financial firms over three years found widespread discussion of AI, but only one company explicitly reported a measurable dollar benefit amounting to around $19 million. Claims of AI-driven gains touted in investor presentations and media often do not translate into concrete earnings statements. “Three years into the AI buildout, the firms buying the technology still cannot put a dollar figure on the payoff,” noted Milos Maricic, founder of AI advisory firm Maximand.

Credit markets have grown wary about financing AI infrastructure expansion. Broadcom is reportedly seeking over $60 billion in debt to fund AI chip development. However, the cost to insure this debt has increased sharply, reflecting lender skepticism. Market analysts suggest that as financing risks rise, companies may face pressure to raise equity or reduce capital expenditures, potentially limiting growth.

Nvidia’s recent earnings offered some reassurance, though the company cautioned that rising memory prices could compress profit margins. Investors continue to grapple with balancing bullish views on growing AI adoption against concerns over escalating production costs and uncertain returns.

Overall, the market currently lacks a clear consensus on who will ultimately reap the financial rewards as AI becomes more pervasive. The price of AI services is falling, semiconductor inputs are growing more expensive, and companies are investing heavily despite unclear immediate profits. While widespread adoption may eventually yield substantial returns, a complex mix of factors must align to sustain growth and deliver tangible financial benefits. The prevailing market sentiment reflects prudence amid uncertainty rather than confidence or alarm.