Why AI Spending Is Soaring
As major AI model vendors shift from per-seat licenses to usage-based billing, corporate treasuries face soaring technology costs and changing infrastructure demands. The post Why AI Spending Is Soaring appeared first on Global Finance Magazine .
As artificial intelligence technology advances, corporate treasuries must prepare for soaring costs, according to this Global Finance Magazine article. Traditional per-seat licensing models have given way to more expensive usage-based billing, with AI model vendors now charging by token usage, making it one of the most expensive software licenses per employee.
Ron Karp, co-founder and CEO of cloud management vendor Wiv, notes that AI costs are growing faster than other IT expenses such as cloud computing. While AI providers have lowered their token prices following a market peak, the reduction in price has not translated into lower spending. Instead, AI use continues to increase rapidly.
Jevons' Paradox, as illustrated by 19th-century British economist William Stanley Jevons, suggests that as technology becomes more efficient, its adoption increases. Elon Musk mentioned during SpaceX Corp’s Q3 earnings call that AI or superintelligence will require 1,000 times more bandwidth than today due to the growth of agentic AI.
AI workloads have expanded from simple input-output queries to more complex tasks, such as untangling legacy code bases, accelerating scientific research, robotics, and domain-specific tasks requiring in-depth reasoning.
Companies like Google and Microsoft have seen significant increases in AI usage. Google's AI Mode searches involve voice, images, or documents and have grown to 20% of global AI Mode searches with over 50 billion visual searches per month. Their first-party model APIs are processing more than 32 billion tokens per minute, a 40% increase from the second quarter and more than four times what the search engine processed a year ago. Microsoft's Copilot workloads grew 50% quarter over quarter.
However, the supply of AI compute capacity may face long-term shortages due to energy constraints. Elon Musk highlighted this concern during SpaceX's Q3 earnings call, stating that the market should expect increasing values of compute per watt rather than a decrease. The constraint is now power, not chips. While companies are still early in managing their AI spending, developing key performance indicators and implementing model routing and memory management strategies are essential steps towards better AI management.
Written by urgent.news from Global Finance's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.