Nvidia Seeks to Strengthen its Lead in AI Chips with a $40 billion Bid for ARM

Nvidia has made an offer of up to $40 billion to buy ARM, the British technology company who’s IP is found in most of the world’s smartphones from Japan’s Softbank. The deal follows the closing of Nvidia’s acquisition of Mellanox the Smart NIC (Smart Network Interface Card) manufacturer in April 2020 over a year after the deal was initially announced. The rationale behind this latest acquisition appears to lie with Nvidia’s ambitions to dominate the market for AI chips in datacenters and edge computing applications. They aim to branch out from their core GPU technology to closely related areas in order to build even faster and more efficient solutions to datacenter needs. Nvidia revealed a long-term road map of how it wants to play this area earlier this month with the announcement of its Bluefield-2 DPU’s (Data Center Processing Units) which combines their Ampere GPUs’, Mellanox Smart NIC technology and ARM processors into a single unit. This combination leads to what it essentially calls an ‘infrastructure-on-a-chip’ card that can easily networked together to form a highly flexible AI processing system. ARM CPU chips are seen as being ideal for this solution due to their high power efficiency, a critical cost consideration for today’s datacenter’s. Given ARM’s long-list of very large customers which include both Apple and Microsoft as well as the current trade tensions between the US and China this is by no means a done deal with various competition authorities around the world set to have their say on the matter, like the Mellanox deal it is likely to take well over a year before it will be completed if at all. Regardless, Nvidia is pressing on with their vision and are set to release at least two Bluefield variants next year even if the bid for ARM is unsuccessful.