Microsoft and Broadcom Are Quietly Teaming Up on New Custom AI Chips

With the boom of the AI era, demands for AI chips have been at an all-time high. Microsoft, which is among the prominent names in the ongoing AI race, has revamped some of its AI chip strategy, at least according to a report by The Information.

Microsoft is reportedly in talks with Broadcom to co-design custom AI chips. And honestly, this push feels inevitable. As mentioned earlier, the demand for AI hardware keeps exploding, and everyone wants more control over the silicon running their models.

Microsoft had already been working with Marvell on chip design. But with competition for high-end semiconductors heating up, the company now seems ready to expand its partnership. Broadcom, which is already a partner of OpenAI, appeared to be the next logical step for designing customized chips for massive generative AI workloads.

And it’s not just Microsoft; the entire industry is moving in the same direction. Google continues to double down on its in-house TPU platform and is preparing to sell the chips more widely. AWS recently announced Trainium3, its most powerful AI accelerator yet. Meta is also working on custom silicon with Marvell, aiming for a 2027 launch window.

While NVIDIA remains the undisputed leader of the AI chip market, other companies are diversifying their data-center chip suppliers. And their custom AI chips now pose a real threat to NVIDIA’s grip on the AI chip industry. That being said, only time will tell if any of them will be able to surpass it.

Do you think these in-house chip efforts will genuinely reshape the AI hardware market, or will NVIDIA stay untouchable for years to come?