- OPENAI (probably) joins Google, Meta and Bytedance in the Personalized Asic Partnership of Broadcom
- Broadcom obtains $ 10 billion in Rack AI commands while Nvidia faces new rivals
- The biggest NVIDIA customers are pursuing internal fleas with Broadcom guiding the transition
As we have reported more than a new time in the past, the IA equipment market changes, some of Nvidia’s largest customers looking for means to reduce costs and take more control over their systems.
Rather than relying solely on NVIDIA costly GPUs, companies are starting to design their own ASICS adapted to their workloads.
Broadcom is one of the largest players in this space, offering the necessary expertise to transform these personalized conceptions into chips and systems ready for production.
Broadcom’s financial results report for its third quarter, The next platform Said that the silicon supplier has now obtained a fourth customer for its personalized XPU program, adding to partnerships with Google, Meta and Bytedance.
The reports and the industry calendar suggest that this new customer is OPENAI, who develops his own inference processor known as Titan under the direction of Richard Ho, a former engineer from Google TPU.
Nvidia always dominates the market of course – in a way – with its Blackwell GB300 NVL72, but the deployment of these rackscale systems is expensive, and companies with massive AI models want equipment designed to better meet their needs.
Personalized ASICs are considered a way to curb costs while providing greater flexibility than a standard GPU and Broadcom is well positioned to guide complex accelerator projects by design, production and packaging.
During a call with Wall Street analysts, the Director General Hock Tan extended to the fourth client of Broadcom (cough,, OPENAI,, cough), saying: “Now, more to these three customers, as we mentioned above, we worked with other prospects on their own AI ACC.”
“In the last quarter, one of these prospects published production orders on Broadcom, and we therefore characterized them as a qualified customer for XPU and, in fact, have obtained more than $ 10 billion in IA rack commands according to our XPU,” said Tan.
“And reflecting this, we are now expecting the prospects for our IA income for the 2026 fiscal year to improve considerably compared to what we had indicated in the last quarter.”
This $ 10 billion figure refers to the complete Rack AI systems, not to Broadcom’s share for the design of underlying fleas.
The income of these orders should start in the third quarter of the fiscal 2026.
It is clear that the biggest buyers in Nvidia are no longer content to depend only on GPUs, and by investing in Asics, they bet that personalized equipment will bring efficiency and control. With its expertise, Broadcom is positioning itself as the company which can make these conceptions a reality.
