- OpenAI (probably) joins Google, Meta and ByteDance in Broadcom’s custom ASIC partnership
- Broadcom secures $10 billion AI rack orders as Nvidia faces new rivals
- Nvidia’s largest customers pursue in-house chips with Broadcom guiding the transition
As we’ve reported more than a new times in the past, the AI hardware market is changing, with some of Nvidia’s biggest customers looking for ways to cut costs and gain more control over their systems.
Rather than relying solely on Nvidia’s costly GPUs, companies are beginning to design their own ASICs tailored to their workloads.
Broadcom is one of the bigger players in this space, offering the expertise needed to turn those custom designs into production-ready chips and systems.
Reporting on Broadcom’s financial results for its third quarter, The Next Platform says the silicon supplier has now secured a fourth customer for its custom XPU program, adding to partnerships with Google, Meta, and ByteDance.
Industry reports and timing, suggest this newest client is OpenAI, which is developing its own inference processor known as Titan under the leadership of Richard Ho, a former Google TPU engineer.
Nvidia still dominates the market of course – by some way – with its Blackwell GB300 NVL72, but deploying such rackscale systems is expensive, and firms with massive AI models want hardware designed to better match their needs.
Custom ASICs are seen as a way to rein in costs while offering greater flexibility than an off-the-shelf GPU and Broadcom is well positioned to guide complex accelerator projects through design, production, and packaging.
On a call with Wall Street analysts, chief executive Hock Tan expanded on Broadcom’s unnamed fourth client (cough, OpenAI, cough), saying, “Now further to these three customers, as we had previously mentioned, we have been working with other prospects on their own AI accelerators.”
“Last quarter, one of these prospects released production orders to Broadcom, and we have accordingly characterized them as a qualified customer for XPUs and, in fact, have secured over $10 billion of orders of AI racks based on our XPUs,” Tan continued.
“And reflecting this, we now expect the outlook for our fiscal 2026 AI revenue to improve significantly from what we had indicated last quarter.”
That $10 billion figure refers to complete AI rack systems, not Broadcom’s share for the underlying chip design.
Revenue from those orders is scheduled to begin in the third quarter of fiscal 2026.
It’s clear that Nvidia’s biggest buyers are no longer content to depend solely on GPUs, and by investing in ASICs they are betting that custom hardware will bring efficiency and control. With its expertise, Broadcom is positioning itself as the company that can make those designs a reality.
