Last year, AMD and Samsung announced that they’ve co-developed an RDNA2-based GPU for Exynos chipsets. And the first chipset with such a GPU is the Exynos 2200, which debuted earlier this year with the Galaxy S22 series. However, the AMD-Samsung collaboration might be in danger, and we might not see AMD GPUs inside Exynos chipsets in the future.
According to a report from SemiAnalysis, ARM is changing its entire business model after its ongoing legal spat with Qualcomm. It all started with Qualcomm’s acquisition of Nuvia, which has been developing high-performance chips for laptops and servers. Nuvia was started by star chip designers from AMD, Apple, Broadcom, and Google. It was widely reported that Nuvia’s designs could be on par with Apple’s and that it could be the brand that truly challenges Apple in efficiency and performance.
ARM is said to be changing its business model of chip design licensing
After this dispute started, ARM claimed that Qualcomm doesn’t have the license to offer products (CPU, GPU, NPU, and ISP) based on ARM technologies. ARM is reportedly changing its entire business model and will no longer license its CPU designs to semiconductor companies like Qualcomm under its TLA (Technology License Agreements) starting in 2024. Instead, ARM will license its technologies directly to OEMs (like Apple, OPPO, Samsung MX, Vivo, and Xiaomi).
However, according to Qualcomm’s counterclaim, the company has the license to use ARM designs well beyond 2025. Moreover, this license reportedly gives Qualcomm the right to extend the contract for several more years past the initially agreed period. Qualcomm claims that ARM is falsely stating to multiple Qualcomm clients that the only way to use ARM technologies in their devices is to “accept a new direct license from ARM on which they pay royalties based on the sales of the OEM’s products.”
ARM could force smartphone OEMs to use its GPU, NPU, and ISP if they want to use ARM’s CPU designs
With its new strategy to license its chip designs directly to OEMs (smartphone, tablet, and smartwatch brands), it cuts out chip firms like Qualcomm and Samsung, who have custom designs (CPUs from ARM but GPUs and ISPs developed in-house or by other brands). ARM is planning to bundle all its technologies in a take-it-or-leave-it model. This means that brands must use ARM’s ISP, NPU, and GPU to use their CPU designs.
If this is correct, Qualcomm can’t use its own Spectra ISP, NPU, and Adreno GPU inside Snapdragon chips. Samsung’s System LSI arm won’t be able to use RDNA2-based GPUs in its Exynos chipsets, and Google won’t be able to use its TPU inside Tensor chips. They all will have to use ARM’s GPU, ISP, and NPU if they decide to develop a chipset with ARM’s CPUs.
And that could force Samsung (System LSI) to abandon its collaboration with AMD for mobile GPUs
If true, this appears like anti-competitive behavior from ARM. Qualcomm argues that ARM can’t change its existing license terms for several more years and has no right to ask for additional royalties from Qualcomm’s customers (smartphone brands that use its Snapdragon chips). This information comes from Qualcomm’s counterclaim documents, and ARM hasn’t publicly announced its plans to change its business model and licensing terms.
ARM claims that Qualcomm’s counterclaims are riddled with inaccuracies and that it is in the process of releasing an official response. So, there’s still a chance that ARM might not go ahead with its plans or bring some additional terms that OEMs can use to negotiate the use of custom components. But if it doesn’t, Samsung’s partnership with AMD for mobile GPU development could effectively end, and it might be forced to use ARM’s Mali GPUs.