OpenAI teams up with Samsung to build its own AI chips
OpenAI is deepening its push into custom silicon, announcing a chip co-development partnership with Samsung Electronics at a Seoul press conference on September 9. Harrison Kim, OpenAI's Korea general manager, confirmed the two companies are collaborating on joint production and research for next-generation AI chips. No technical specifications or timelines were disclosed — and Samsung declined to comment — but the announcement marks a clear escalation in OpenAI's ambition to own its hardware stack.
Why chips, why now
NVIDIA controls roughly 80% of the AI accelerator market, per Silicon Analysts (Q1–Q2 2026), and that dominance comes with supply bottlenecks and margins above 85%. Every major AI lab has felt the pressure. OpenAI's answer has been to build its own silicon: in June 2026, it unveiled the Jalapeno inference chip, co-designed with Broadcom and manufactured by TSMC. The chip was purpose-built for LLM inference — the compute-heavy process of actually answering user queries — and reportedly designed in just nine months.
Samsung is the next move. Where Jalapeno proved the concept, a Samsung partnership brings manufacturing scale. Samsung is one of the world's largest chip foundries and memory producers, and adding its fabrication capacity could help OpenAI undercut NVIDIA on per-unit cost if designs are shared broadly.
The broader trend supports the bet. Custom ASIC shipments grew 44.6% in 2026 versus just 16.1% for merchant GPUs, according to CommandLinux. Inference accounts for roughly two-thirds of all AI compute spending — and that's exactly the segment OpenAI is targeting with its own silicon.
Samsung's side of the deal
The partnership isn't purely about chips. Samsung has already deployed ChatGPT internally at scale, with employees across R&D;, marketing, and sales using the service. That makes Samsung one of the largest enterprise ChatGPT deployments in the world — which gives OpenAI both a paying customer and, now, a manufacturing ally.
Reuters (Seoul press conference, Sept 9) reported Kim's direct statement: the area of "greatest progress and recognition" from Samsung has been joint production and research in next-generation chip design. Whether that means Samsung fabs will eventually produce OpenAI-designed chips — or something co-engineered from scratch — remains unconfirmed.
What comes next
Microsoft is expected to purchase around 40% of initial Jalapeno production. A Samsung-backed successor could widen that supply significantly. The inference market is shifting fast: analysts project it will fragment away from NVIDIA-dominated hardware by 2029. OpenAI appears to be positioning itself to supply that market rather than simply buy from it.