- OpenAI (probably) joins Google, Meta and Bytedance at the Broadcom Custom Asic Association
- Broadcom ensures AI rack orders of $ 10 billion since Nvidia faces new rivals
- Nvidia’s largest customers pursue internal chips with Broadcom by guiding the transition
As we have reported more than a new time in the past, the AI hardware market is changing, with some of the largest Nvidia customers who are looking for ways to reduce costs and obtain more control over their systems.
Instead of trusting only the expensive NVIDIA GPU, companies are beginning to design their own ASIC adapted to their workloads.
Broadcom is one of the largest players in this space, which offers the necessary experience to convert those custom designs and production ready for production.
Informing about Broadcom’s financial results for your third quarter, The next platform He says that the silicon supplier has now secured a fourth client for his personalized XPU program, which adds to Google, Meta and Bytedance associations.
Industry reports and time, suggest that this new 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, in some way, with its Blackwell GB300 NVL72, but the implementation of such rack scale systems is expensive, and companies with massive models want hardware designed to better meet their needs.
Custom ASICs are seen as a way to control costs while offering greater flexibility than a GPU and Broadcom is well positioned to guide complex accelerator projects through design, production and packaging.
In a call with Wall Street analysts, executive director Hock so expanded in the fourth client without Broadcom’s name (cough, Opadai, cough), saying: “Now more to these three clients, as we had mentioned above, we have been working with other prospects in their own AI accelerators.”
“The last quarter, one of these prospects published production orders for Broadcom, and consequently we have characterized them as a qualified customer for Xpus and, in fact, we have obtained more than $ 10 billion orders for the racks based on our XPU,” so.
“And reflecting this, now we hope that the perspective of our fiscal income from AI 2026 will improve significantly what we had indicated the last quarter.”
That $ 10 billion figure refers to completing Rack AI systems, not Broadcom’s participation for the underlying chips design.
The income of these orders is scheduled to begin in the third quarter of fiscal year 2026.
It is clear that the largest buyers of NVIDIA are no longer happy to depend solely on the GPUs, and when investing in ASIC is betting that personalized hardware will bring efficiency and control. With his experience, Broadcom is positioning himself as the company that can make those designs come true.
