In companies' race to adopt AI, the key variables are shifting from securing GPUs to token costs, data sovereignty and model selection. [Photo: ChatGPT]

Companies’ race to adopt AI is shifting from securing GPUs to cutting operating costs, controlling data and choosing models. An analysis says overall system efficiency and profitability are becoming more important than hardware performance as AI pilot projects are applied to real work.

SiliconANGLE reported on Nov. 17, local time, that Dell Technologies explained the shift in enterprise AI infrastructure strategy at the Supercomputing Conference SC26 in Chicago.

Brad Maltz (브래드 말츠), senior director of Dell AI Solutions, stressed that companies need a “full-stack AI factory” that integrates software, data, models and security, as well as servers and GPUs, to move AI pilots into production.

He cited token costs from rising AI usage as a major burden. Maltz said token spending by cutting-edge AI model developers and large cloud providers is larger than expected. He said they did not anticipate costs as high as $10,000 to $20,000 per user per month, or about 13,335,000 to 26,674,000 won.

Beyond costs, data sovereignty, intellectual property leakage, security and governance were cited as obstacles to the spread of enterprise AI. Companies are therefore showing growing interest in open-weight models that can run on in-house servers while reducing reliance on external clouds. Nvidia’s Nemotron was mentioned as a leading option.

Dell is strengthening its AI factory business that integrates software and hardware in response to such demand. Through its AI ecosystem program, it verifies partner solutions for compatibility, security and stability, and about 60 companies have applied to join or are undergoing evaluation.

According to the latest materials Dell released the previous day, on Nov. 16, the number of customers for Nvidia-based AI factories has surpassed 6,500. Its products and services spanning data, computing and networks also expanded to more than 420.

Its newly unveiled agentic data gateway supports the safe use of internal company data by AI agents. Dell said its own tests cut token consumption by up to 76 percent through reuse of data context and raised answer accuracy from 60 percent to 94 percent.

Competition in the enterprise AI market appears to be reshaping beyond simply securing computing power, focusing instead on cost efficiency, data control and real business results.

Keyword

#Dell Technologies #SC26 #Chicago #Nvidia #Nemotron
Copyright © DigitalToday. All rights reserved. Unauthorized reproduction and redistribution are prohibited.