"Memory shortage" sweeps the globe, becoming a critical bottleneck in AI competitions? Another Silicon Valley bigwig speaks out

Currently, artificial intelligence (AI) companies are engaging in intense competition to acquire more memory chips. The entire industry is facing a serious supply shortage dilemma: costs are rising, product deliveries are being obstructed, and some companies - especially those in the consumer electronics sector - have started to raise prices. Google DeepMind CEO Demis Hassabis recently discussed the global "memory shortage" in an interview. He stated that the entire supply chain of memory chips is constrained, and hardware challenges are "limiting the deployment of a large amount of (AI)". The market demand for Google's Gemini and other AI models far exceeds the current supply capability of the company. "Furthermore, this also limits research work to a certain extent," Hassabis said. "To test new ideas on a large scale, verify their feasibility, a large number of chips are needed to support this." Whether they are researchers working at Google, Meta, OpenAI, or other tech giants, they all have an urgent need for chips - and memory is a key component. Meta CEO Mark Zuckerberg once stated that one of the most important things for AI researchers, apart from funding, is to obtain as many chips as possible. Hassabis pointed out that as long as there is capacity limitation, there will be a "bottleneck". "The entire supply chain is under tension," Hassabis said. "We are somewhat fortunate because we have our own tensor processing unit (TPU) and have the capability of chip design." For a long time, Google has been independently developing tensor processing units (TPU) for internal use. The company also rents TPUs to external customers through its cloud services - which has also put pressure on Nvidia. However, even with its self-developed TPU, Google cannot avoid the fiercely competitive memory market. Hassabis said, "Ultimately, the number of suppliers for core components is still very limited." Currently, three companies dominate the global memory chip production: Samsung, Micron, and SK Hynix. These companies are working hard to meet the chip demands of AI mega-enterprises while also maintaining their long-standing electronic product customers, which poses challenges. To make matters worse, the types of memory chips needed by AI companies are different from those of personal computer (PC) manufacturers. Large-scale language model manufacturers require HBM (High Bandwidth Memory) chips. Intel CEO Pat Gelsinger recently warned that the bottleneck in AI development has shifted from "computing power" to "memory" and more broadly to infrastructure systems. He bluntly stated that the memory shortage issue will not be alleviated before 2028. This article is reproduced from "Cailian News"; Edited by GMTEight: Yan Wencai.