SoftBank and Nvidia are negotiating an agreement for the graphics company to buy ARM in a cash and equity exchange for $40 billion, as per a leak. SoftBank purchased Arm in 2016 for about $31 billion and has been searching in recent years to sell the chip design and IP business in an attempt to reduce losses. The most active party throughout recent years is Nvidia.
While the terms of the agreement are still not announced, a formal statement may be released in the coming weeks. The contract will certainly be reviewed and regulatory consent will be given. Arm’s co-founder confirmed earlier that a purchase by Nvidia was a failure and Arm is about to be returned to the UK.
One big issue arising from an acquisition by Nvidia is the retention of the neutrality of ARM to approve the production of the chip and IP. SoftBank is not a chipmaker, and as such is not competitive, to name but a handful – all of them are ARM licensees of other suppliers, including Samsung, Apple, Intel, Broadcom, and Qualcomm, for instance. Nvidia regulates access to such a vast and valuable market that the industry can be seen as a matter of interest.
After weeks of uncertainty about ARM’s fate, an arrangement between SoftBank and Nvidia seems to be being signed, which enables the GPU manufacturer to purchase ARM as one of the most expensive inventions in history. Any tech giant offer is likely to be reached by the regulatory authority, but Nvidia would presumably bet it will be able to support policymakers and regulators.
Nvidia has shown that its strategic development and advancement in high-performance computing, AI, data center architecture, and supercomputers can be accelerated. In order to expand its dominance for not just its GPUs, but also its technology and networking, they made multiple acquisitions. Mellanox Technology and Cumulus Networks were recently purchased by Nvidia and incorporated under the Nvidia Network name. The end goal is to clean up the networking hardware and software of Nvidia’s HPC platforms.
The same goes for GPUs and CPUs — because it does not have CPUs of its own. The ARM will help them to bid for profitable exascale modern computer contracts enjoyed by Intel and AMD. Recently, ARM-based supercomputers have been more sprung, with the Summit of IBM becoming the most efficient supercomputer in the world by Japan’s Fugaku supercomputer. Nvidia has its Tegra sockets and others which have traditionally used the ARM and Cortex-Axx series commands, but it doesn’t have a custom-designed CPU and GPU under the roof.
With its own in-house CPU architecture, Nvidia would gain greater integration of ARM CPUs with its GPU accelerators, making it much easier to compete with AMD and Intel. It will introduce Cuda to ARM, and make Nvidia’s complete stack of HPC software inside the ARM environment since Nvidia has already confirmed it. The next move in relation to Nvidia could be custom ARM CPUs. We will hopefully not have to wait a lot for this to happen, because if so, it will be one of the best technology acquisitions in the world and will have a huge influence on the semiconductor industry.