OpenAI CEO meets Samsung to secure memory chips for AI demand

Securing the hardware supply

OpenAI CEO Sam Altman recently traveled to Yongin, south of Seoul, to meet with roughly 50 senior executives from Samsung. The primary goal of this meeting was to address the ongoing shortage of memory chips, which are essential components for running large-scale artificial intelligence systems. These discussions focused on how AI will influence major industries like biotech and finance over the next decade.

The meeting highlights a growing urgency for AI companies to secure physical hardware. As demand for computing power accelerates, the ability to guarantee a steady flow of components becomes a critical business priority. By engaging directly with a major manufacturer like Samsung, OpenAI is attempting to stabilize its supply chain against a backdrop of global scarcity.

This partnership follows a letter of intent signed in October 2025 between OpenAI and two major chipmakers, Samsung Electronics and SK Hynix. While specific timelines for the delivery of these chips have not been disclosed, the agreement signals a deep commitment to scaling up infrastructure. The collaboration aims to ensure that the massive computing requirements of future models do not outpace the available manufacturing capacity.

The limits of local inference

During a recent podcast appearance, Sam Altman addressed a common misconception regarding how AI models will be run in the future. Many people believe that the rise of open-source models will allow individuals to run AI locally on their own desktop computers. Altman argued that this perspective overlooks the reality of the current global supply of DRAM, which is the type of memory used in both computers and data centers.

Altman expressed concern that distributing limited memory resources to individual home computers is inefficient from a global allocation standpoint. He noted that if everyone keeps a large supply of memory in their personal devices, it might only be used a small fraction of the time. This approach could hinder the ability to maximize the utility of the hardware in high-demand environments like the cloud.

The CEO emphasized that the current environment is very supply-constrained, making it necessary to prioritize maximum utilization. He suggested that the intuition of many people is the opposite of what is needed for the industry to scale effectively. By focusing on cloud-based infrastructure, companies can ensure that the limited memory chips available are being used as much as possible to power AI services.

A shift in global production

The memory chip industry is currently facing a significant shift in where its products are going. Reports from TrendForce indicate that by 2026, data centers could consume as much as 70 percent of the memory produced worldwide. This massive pivot toward AI infrastructure means that the traditional balance of the market is being fundamentally reorganized to favor large-scale computing projects.

Memory companies have issued warnings that these shortages are likely to persist well into 2027 and beyond. This trend suggests that the demand for AI is not a temporary spike but a long-term structural change in how hardware is manufactured. Manufacturers are increasingly prioritizing the high-volume needs of data centers over the requirements of smaller, individual projects.

As a result of this shift, consumer electronics are the most likely category to suffer from ongoing shortages. When a significant majority of the global output is earmarked for AI, the availability of components for phones, laptops, and other gadgets may decrease. This creates a situation where the growth of AI technology directly impacts the accessibility and cost of everyday technology for the general public.

Expanding the collaboration

While the specific details of the new agreements remain private, the relationship between OpenAI and Samsung is clearly deepening. Following comments from OpenAI Korea's general manager regarding the development of next-generation chips, a spokesperson for the company clarified their stance. They stated that the partners are simply planning to expand their existing relationship in the context of their current work together.

This expansion includes the ongoing adoption of ChatGPT by Samsung Electronics and the previously announced letter of intent. The partnership serves as a blueprint for how AI software companies are integrating with hardware giants to build out the necessary physical foundations. It moves the relationship beyond just using a product to co-developing the underlying technology required for the next generation of computing.

The scale of this demand is underscored by the projected needs for OpenAI's Stargate project. The fact that the company is looking for 900,000 DRAM wafers a month shows the sheer magnitude of the hardware required to sustain their ambitions. As these companies work together, they are attempting to build a stable ecosystem that can support the massive growth of AI over the coming decade.