ChatGPT might not be using the model you think — and it’s also hiding others in settings
OpenAI just picked Samsung to be the only supplier of HBM4 for their first homegrown AI chip, codenamed Titan. It’s not a huge volume win yet, but this ChatGPT Samsung HBM4 supply deal matters more for what it signals.
Samsung’s been behind SK Hynix in the HBM race for two years straight. Their HBM3E stumbled through NVIDIA’s quality tests, while rivals continued to secure orders. But now they’ve clawed back a win that actually matters.
OpenAI plans to launch a self-developed AI chip, codenamed “Titan,” by the end of 2026, using TSMC’s N3 process. Samsung is getting a chunk of over 5.5 billion gigabits of HBM4 production, making it the third largest deal after NVIDIA and AMD.
Lee Jae-yong met Sam Altman in October and signed a letter of intent. OpenAI gets guaranteed access to next-gen memory, whereas Samsung gets validation that its redesigned tech works.
Titan’s being built for inference workloads, not training. OpenAI is at a scale that almost requires it to co-design its own hardware to better match its own software and also reduce its costs.
The OpenAI deal gives Samsung a customer that doesn’t care about NVIDIA’s approval stamp. It validates that their redesign worked. If Titan Gen 2 and Gen 3 stick with Samsung memory, this becomes a multi-year revenue stream.
The post ChatGPT is getting a Samsung upgrade thanks to this HBM4 deal appeared first on Sammy Fans.