NVIDIA’s soaring GPU prices and continued memory shortages are increasing AI infrastructure costs worldwide, forcing cloud providers and enterprises to rethink how they secure affordable computing capacity.
For businesses and startups using artificial intelligence, the impact could eventually be higher costs to run AI systems. As companies already struggle with the growing computing costs of processing large volumes of AI tokens, expensive hardware could add further pressure.
NVIDIA’s RTX PRO 6000 Blackwell GPU is now listed at $16,000 on its US Marketplace, compared with $13,250 earlier this year and about $8,565 when it launched in March 2025. The rise in the cost of powerful AI hardware comes as the world faces a shortage of memory chips, another key component needed to run AI systems and data centres.
The problem is being driven by the rapid expansion of AI. Technology companies are building more data centres and buying large quantities of GPUs and memory chips to train and run AI models. This has increased demand faster than suppliers can expand production.
Research firm TrendForce expects memory supplies to remain tight through 2027. It has said companies making memory chips are increasingly focusing their production on specialised memory used in AI systems and servers, leaving less supply available for other customers.
The shortage matters because memory chips are needed alongside powerful processors. When the cost of these components rises, cloud providers and companies running AI infrastructure face higher costs.
Large technology companies are in a stronger position because they can buy equipment and secure supplies through long-term agreements. Smaller cloud providers, startups and businesses may have less bargaining power and could end up paying more.
The pressure is also reaching India.
“NVIDIA's pricing is driven largely by demand and component costs, especially memory, which remains tight and is projected to rise through 2027. Blended across our existing fleet and volume purchase commitments, we anticipate GPU cloud pricing in India to rise by 15–20% over the next 12–18 months, unless memory supply eases or alternative architectures gain traction,” said Ishan Talathi, Founder & CEO, CloudPe.
CloudPe said it may be partly protected from the immediate impact because it bought much of its existing GPU capacity before prices rose.
“We buy GPUs at scale and we bought early. A significant share of our fleet was secured at pre-surge prices, and that advantage flows directly to our customers,” Talathi said.
For companies using AI, the rising cost is also changing the decision between buying expensive hardware and renting computing power from cloud providers.
Buying GPUs can make sense for companies that use them constantly, but it requires a large upfront investment and brings electricity, maintenance and upgrade costs. Renting avoids those expenses but leaves companies exposed to changes in cloud prices.