Dogukan Gurel
February 12, 2026•Update: February 12, 2026
ANKARA
Prices for random access memory (RAM), a type of high-speed storage component whose price has reached record highs due to artificial intelligence (AI) giants training chatbots, have triggered a hardware bottleneck worldwide, and the Turkish tech sector could benefit from this development, according to an industry representative.
Cihan Sari, general secretary at Istanbul-based nonprofit Artificial Intelligence and Technology Association, told Anadolu that the rising prices of computer hardware and the consequent hardware bottleneck may not be a “simple disruption in the supply chain, but a tipping point where AI investments have hit the limits of the physical world.”
Sari described the current situation as a “crisis that has turned into a memory war.”
“The massive appetite for generative AI is sort of devouring the memory supply, and this is a problem not only for tech giants but also represents a macroeconomic risk that drives inflation and affects the price of even the smartphones in our pockets,” he said.
He noted that major manufacturers, such as Samsung, SK Hynix, and Micron, are shifting their production capacity from standard consumer RAM modules to high-bandwidth memory (HBM) components used by data centers — one gigabyte of HBM requires the same factory capacity as three gigabytes of standard RAM, effectively slashing consumer supply.
“These companies are sort of saying, ‘I won’t bother selling RAM to the at-home gamer anymore and will allocate my capacities to the server side' with this shift, because AI firms like OpenAI, Google, and Meta are sort of waiting at the door with cash in hand,” he said, noting that this has resulted in a serious supply gap in the market with prices surging by over 100% already.
Sari urged Turkish entrepreneurs to pivot away from hardware-intensive model training to take advantage of this global scarcity as a unique opportunity, calling on Turkish developers to focus on optimized solutions that can run on existing models with fewer resources.
“Instead of more general AI models that seemingly know everything, developing small but domain-specific models that only know law, health care or production, for instance, is the only way out for the Turkish tech sector.”
“The tech world is facing physical limits, and the winners in the coming period won’t be those with the biggest computers but those who use their resources most wisely — AI training won’t stop, but it’s shifting, and as Türkiye, we need to focus on the new efficiency path we can take,” Sari said.
*Writing by Emir Yildirim in Istanbul
