|
Getting your Trinity Audio player ready...
|
OpenAI Says Working With Samsung on Next-Generation Chips, Deepening Cooperation
OpenAI is expanding its relationship with Samsung Electronics well beyond memory supply deals and enterprise software licensing, confirming that the two companies are now collaborating directly on next-generation chip development. The disclosure came from Harrison Kim, general manager of OpenAI Korea, during a press conference in Seoul on Wednesday, and it adds another layer to a partnership that has been building steadily since OpenAI’s Stargate infrastructure push began pulling in major Korean suppliers.
Kim told reporters that one of the areas where OpenAI has made the most progress with Samsung Electronics involves joint production and research on the next generation of chips the two companies are developing together. He didn’t go into specifics about which chips are involved or what exact role Samsung is playing in the collaboration, but the framing itself is notable. This isn’t OpenAI simply buying memory from Samsung, it’s the two companies working together on the underlying silicon that could eventually power OpenAI’s AI systems.
The timing lines up with OpenAI’s broader effort to build out its own chip capabilities rather than relying entirely on established GPU makers. Earlier this year, OpenAI introduced Jalapeno, its first custom AI inference processor, developed in partnership with Broadcom. OpenAI handled the architecture design tailored to the demands of large language model workloads, while Broadcom contributed the silicon and networking expertise needed to bring the chip to production. That project signaled OpenAI’s intent to reduce its dependence on dominant GPU suppliers and gain more direct control over both performance and cost as its compute needs keep growing. The deepened Samsung collaboration looks like a continuation of that same strategy, potentially giving OpenAI another manufacturing and research partner as it works to diversify its hardware supply chain.
Samsung’s involvement with OpenAI has been building for a while through the Stargate initiative, OpenAI’s ambitious plan to massively scale up AI infrastructure. Samsung Electronics and SK hynix both joined that effort, agreeing to ramp up production of advanced memory chips as part of a plan to reach roughly 900,000 DRAM wafer starts per month at an accelerated pace, a scale of output critical to powering OpenAI’s most demanding models. That agreement followed a meeting in Seoul between South Korean President Lee Jae-myung, Samsung Electronics Executive Chairman Jay Y. Lee, SK Chairman Chey Tae-won, and OpenAI CEO Sam Altman, underscoring how closely this partnership is tied to South Korea’s own ambitions to become one of the top three AI nations globally.
Kim explained that demand for memory chips is only going to keep climbing as more advanced processors are needed to handle increasingly complex and faster computing tasks, which is why deepening ties with Korean semiconductor suppliers has become a central priority for OpenAI’s local office. That demand isn’t limited to training massive models either. Inference, the process of actually running a trained model to generate responses for millions of users, requires enormous and continuous compute resources, and memory bandwidth has become one of the biggest bottlenecks in scaling that up efficiently. Samsung’s position as one of the world’s largest memory chip producers makes it a natural partner for OpenAI as it tries to secure supply for years of anticipated growth.
The chip collaboration isn’t happening in isolation from Samsung’s broader embrace of generative AI either. The company said back in June that employees in its Device eXperience division, the unit responsible for mobile phones, televisions, and home appliances, would be permitted to use ChatGPT, Google’s Gemini, and Anthropic’s Claude for their work. Kim went a step further at Wednesday’s press conference, describing Samsung Electronics as one of the largest-scale deployments of ChatGPT anywhere in the world, with employees across Korea and internationally using the tool for research and development, marketing, and sales tasks.
That enterprise adoption appears to be accelerating quickly across South Korea more broadly. OpenAI said the number of ChatGPT Enterprise users at South Korean companies and institutions has surged roughly 28-fold as of the end of August compared to a year earlier. Businesses in the country are moving beyond individual employees using AI for isolated tasks and toward company-wide deployment across functions including legal, sales, finance, marketing, and operations. That kind of growth curve suggests South Korean firms are treating generative AI less as an experimental tool and more as core infrastructure for how they run day-to-day business, a shift that mirrors what’s happening at large enterprises in the US and Europe but is playing out at a notably faster pace.
Samsung, for its part, has not confirmed details of the chip collaboration publicly. The company said it could not comment on matters related to client relationships when asked, which is fairly standard practice for a supplier discussing work tied to a major customer’s unreleased technology. That silence leaves several open questions about exactly what shape this chip partnership will take, whether it involves Samsung’s foundry business manufacturing designs OpenAI has developed in-house, joint research into new chip architectures, or some combination of both.
What is clear is that OpenAI’s hardware strategy has been evolving quickly this year, moving from a company almost entirely reliant on Nvidia GPUs toward one actively building relationships across multiple points in the chip supply chain, from custom inference silicon with Broadcom to memory supply agreements and now joint chip research with Samsung. For a company burning through enormous amounts of compute to train and run its models, that kind of diversification isn’t just a nice-to-have, it’s increasingly viewed as essential to avoiding supply bottlenecks and maintaining leverage over both cost and technical roadmap decisions as competition in the AI infrastructure race intensifies.
For Samsung, deeper integration with OpenAI offers a different kind of upside. As custom AI silicon becomes a bigger part of the semiconductor industry’s growth story, positioning itself as a manufacturing and research partner to one of the biggest names in AI gives Samsung another foothold in a market it’s been aggressively trying to expand into, alongside its existing memory chip dominance. It also reinforces South Korea’s broader strategic push to cement itself as critical infrastructure in the global AI supply chain, not just as a memory supplier but as a genuine technology partner shaping how future AI hardware gets built.
Neither company has laid out a timeline for when products of this collaboration might become public, and Kim’s comments Wednesday were notably light on specifics. But the disclosure itself marks a meaningful shift in how OpenAI is talking about its relationship with Samsung, from a customer buying components to a collaborator shaping what comes next.
