Trillion-dollar tech company emerges as key partner to help Google, Meta and other hyperscalers build an Nvidia-free AI future waynewilliams@onmail.com (Wayne Williams) | usagoldmines.com

• Broadcom reportedly secures large HBM order from SK Hynix
• These chips will likely be used in AI hardware for at least one major CSP
• This is part of a broader trend to reduce reliance on Nvidia

Hyperscalers such as Google, Meta, and TikTok owner ByteDance are increasingly seeking to diversify their AI hardware supply chains, reducing their reliance on Nvidia, which has long dominated the sector.

Broadcom is playing a growing role in this shift, and in its recent earnings conference call, CEO Hock Tan told investors the company has three hyperscale customers who are each planning to deploy one million XPU clusters by 2027, and that it has been approached by two additional hyperscalers who are also in advanced development for their own AI XPUs.

Although Broadcom hasn’t named names, it is widely believed to be working with Google, Meta, ByteDance, and OpenAI on custom AI chips. The firm is also thought to be collaborating with Apple to develop the iPhone maker’s first artificial intelligence server chip, codenamed “Baltra,” providing the advanced networking technologies essential for AI processing.

Bad news for Nvidia

According to TheElec, Broadcom has now approached South Korean memory giant SK Hynix to provide the HBM it plans to use in custom AI chips for a “major” (but predictably unnamed) tech company.

TheElec’s sources say that Broadcom actively pursued SK Hynix to supply its verified HBM solutions and has secured a large order of the in-demand memory, with shipments expected to begin in the second half of 2025.

SK Hynix, Samsung’s biggest memory rival, is a key supplier of HBM to Nvidia, which likely won’t be thrilled with this latest news.

To meet the increased demand from Broadcom, SK Hynix is reportedly adjusting its production capacity. TheElec says the company will boost its 1b DRAM wafer output (used as the core die in its HBM) from 140,000–150,000 units to 160,000–170,000 units in 2025. There is a knock on with this expansion however, which is that it could delay the rollout of SK Hynix’s next-generation 1c DRAM as the company prioritizes its immediate production needs.

You might also like

• Broadcom is making some major changes to stop VMware exodus
• OpenAI is talking to Broadcom to produce AI chip capable of rivaling Nvidia
• Broadcom says three hyperscalers want to deploy 1,000,000 GPUs by 2027