Nvidia is reportedly pondering snapping up Arm, going as far as approaching the Softbank-owned microprocessor designer to talk about a potential takeover.
Arm licenses chip blueprints, which range from individual CPU cores to graphics and machine-learning accelerators to entire instruction set architectures, to system-on-chip manufacturers, who place billions of the components in cellphones, tablets, Internet-of-Things gear, gadgets, appliances, sensors, embedded electronics, and other equipment.
Founded in Cambridge, UK, in 1990 as a joint venture between Britain's Acorn Computers and Silicon Valley's Apple and VLSI Technology, it was acquired by Softbank for $32bn in 2016, making it the most valuable Brit tech company to date.
Whispers that Softbank CEO and Arm Chairman Masayoshi Son was considering selling off Arm began swirling earlier this month after the Japanese giant reportedly hired Goldman Sachs to explore IPO and sale options for the chip design biz.
Arm hopes AWS, FOSS, cloud-native combo will wrestle you onto its architectureREAD MORE
And it emerged Softbank had been putting Arm under considerable pressure to hike its licensing fees for customers to ramp up its revenues and profits.
Today, the rumor-mill is buzzing again. Nvidia is potentially interested in acquiring Arm, Bloomberg first reported, citing sources, and the graphics processor titan is said to be in discussions with Softbank. Nvidia and Arm declined to answer El Reg's questions, while Softbank was not immediately available for comment.
Nvidia is most well-known for its GPUs, which are optimized for running AI algorithms and rendering 3D scenes in real-time by cranking through vector math calculations at high speed. It unveiled its latest graphics-processing architecture Ampere in May, and its share price skyrocketed in July – it was more valuable than Intel at one point.
Nv's GPUs still need general-purpose CPUs to run application software that sets up and instructs the vector engines, and depending on the machine, those CPU cores will most likely be x86-64 ones from Intel or AMD, Power chips from IBM, or Arm-compatible ones licensed to Nvidia.
By buying Arm – including its Neoverse server processor family – Nvidia can bring all that technology, know-how, and staff in house, and closely couple it with its Tesla GPUs and Mellanox networking, and produce an all-Nvidia server hardware platform to take on Intel and AMD.
Presumably Nvidia thinks that may be a better deal than continuing to license CPU designs from Arm, which has a somewhat uncertain future right now, plus it gets its hands on Arm's revenues from the industry.
Intel and AMD each have their own CPU and GPU core technology, whereas Nvidia just has GPUs and licenses in the Arm side (plus Nvidia sprinkles in some RISC-V as glue). With Arm in its nest, Nvidia may feel more in control of its destiny – and that of all those component suppliers licensing Arm's technology for their own silicon.
Apple, we're told, was not interested in buying Arm: it has an architectural license to produce its own Arm chips, primarily for its iThings and soon its Macs, without any, or very little, input from Arm. ®