NVIDIA's Rubin platform treats memory architecture as a primary design constraint alongside compute performance.

At CES 2026 in Las Vegas, Nevada, Nvidia addressed the industry's most pressing challenge: the booming demand for AI computing and the accompanying shortage of memory supply. The company officially launched the Rubin platform, comprised of six chips that combine into one AI supercomputer, which Nvidia claims is more efficient than its Blackwell models and offers increases in compute and memory bandwidth. "Rubin arrives at exactly the right moment, as AI computing demand for both training and inference is going through the roof," Nvidia CEO Jensen Huang said in a press release.

Rubin-based products will be available from Nvidia partners in the second half of 2026, with adoption expected from AWS, Anthropic, Google, Meta, Microsoft, OpenAI, Oracle, and xAI. Anthropic CEO Dario Amodei praised the announcement, stating that "The efficiency gains in the NVIDIA Rubin platform represent the kind of infrastructure progress that enables longer memory, better reasoning, and more reliable outputs." According to company executives, Rubin delivers up to ten times reduction in inference token costs and four times reduction in the number of GPUs used to train models that rely on mixture of experts (MoE) architecture, such as DeepSeek.

The severe memory shortage driving these product announcements has reached critical levels. According to a recent report from Tom's Hardware, gigantic data center projects required roughly 40% of the global DRAM chip output. This scarcity is causing price hikes in consumer electronics and rumored GPU price increases as well. According to a report from South Korean news agency Newsis, chipmaker AMD is expected to raise the prices of some of its GPU offerings later this month, and Nvidia will allegedly follow suit in February. Nvidia has been aggressively responding to this bottleneck, making its largest purchase ever last month with the acquisition of Groq, a chipmaker specializing in inference.

Beyond Rubin, Nvidia is addressing the shifting infrastructure demands of agentic AI by unveiling a new class of AI-native storage infrastructure called the Inference Context Memory Storage Platform. Agentic AI systems, which have become central to the industry over the past year, require increased memory capacity as they must remember information from earlier interactions to autonomously carry out complex tasks. Nvidia's senior director of HPC and AI hyperscale infrastructure solutions, Dion Harris, explained the core challenge: "The bottleneck is shifting from compute to context management. To scale, storage can no longer be an afterthought." He further noted that "As inference scales to giga-scale, context becomes a first-class data type, and the new Nvidia inference context memory storage platform is ideally positioned to support it."

While these innovations may help address the chip shortage and memory bottlenecks that have strained the industry's growth, other challenges remain. Even if the memory problem is resolved, the AI industry will continue to face other bottlenecks in its unprecedented growth, most notably the immense strain that data centers put on the U.S. power grid.