Enterprise adoption of AI is facing headwinds primarily due to challenges in predicting ongoing inferencing expenses, coupled with concerns over potentially hefty cloud service bills.

Analyst firm Canalys reports an increase in spending on cloud infrastructure and platform-as-a-service to $90.9 billion globally in Q1, citing AI development as a key driver.

While the transition to AI deployment progresses, enterprises are grappling with recurring costs tied to inferencing, which unlike the one-time training expenses, weigh heavily on operational budgets.

Rachel Brindley, Senior Director at Canalys, notes that as AI scales, companies are scrutinizing the cost-efficiency of their AI infrastructure more closely, exploring different models and platforms to optimize expenses.

The frequent fluctuations in pricing models, often based on usage metrics like API calls, make financial forecasting challenging for companies scaling AI operations, explains Canalys researcher Yi Zhang. This uncertainty leads organizations to apply restrictions or limit deployments, stunting broader AI integration.

Concerns are not unfounded, as some organizations face surprisingly high cloud bills due to unforeseen usage spikes. Instances like 37signals’ costly cloud experiences highlight the risks involved.

Canalys suggests that while public cloud providers enhance AI efficiency, costs could remain unsustainable at scale. Some enterprises might find alternative options in colocation or niche hosting services to manage costs better.

The report confirms that AWS, Azure, and Google Cloud dominate the market with a 65 percent share, though Microsoft’s and Google’s growth outpaces AWS, largely due to their higher year-on-year growth rates.