The cost of artificial intelligence is rapidly increasing, impacting everything from corporate debt markets to the price of AI servers. Corporations have already offered almost $1.5 trillion in debt this year, a 36% increase from last year, partly due to tech hyperscalers like Microsoft, Amazon, Alphabet, Meta, and Oracle borrowing heavily for AI-related projects. These companies, along with Nvidia, have committed up to $4.2 trillion through debt and contractual obligations to AI, raising concerns about the sustainability of such massive investments.
Nvidia-based AI servers are set to become significantly more expensive, with some data center customers being notified of price hikes exceeding 15%. This increase is primarily attributed to soaring memory costs, particularly for systems containing Nvidia's Blackwell and next-generation Vera Rubin platforms. Analysts note that given the substantial increase in memory costs and the amount required for AI platforms, a 15% price hike for servers doesn't seem excessive. Nvidia, despite generating roughly $1 billion in free cash flow every two days, is passing these increased costs onto customers, which has led to investor concerns about peak earnings and the overall financing of the AI trade.
In the broader AI market, companies are also adjusting their pricing. Chinese AI developer DeepSeek is raising prices for its flagship V4 models by more than four times, bringing its rates closer to competitors. This comes ahead of a potential initial public offering. Similarly, OpenAI and Anthropic are urging customers to re-evaluate the true cost of using different AI models. Alibaba has also made a significant move, raising $10.2 billion in Hong Kong's largest follow-on offering to fund its AI initiatives, committing $56.5 billion over the next three years to build out its AI infrastructure.
The massive spending on AI infrastructure by tech giants is also having an unexpected effect on financial markets. Some investors suggest that the "crowding out" theory, where government borrowing limits corporate funds, is being reversed. Instead, the intense borrowing by tech hyperscalers for AI is absorbing so much capital that it's contributing to higher US Treasury yields, making it more expensive for the government to borrow as well.
