Imagination Technologies recently announced its financial results for the first half of 2021, with the company generating revenues of $76 million, a 55% increase compared to $49 million for the same period in 2020, and with no external third party debt.
Underpinned by this year’s good results, Imagination has also formally announced its next step in expanding its reach: seeking new business growth by developing and launching a RISC-V CPU family that will address both the discrete CPU market and the heterogeneous computing sector.
Imagination Technologies is a UK-based semiconductor company founded in 1985. Its primary business is the design of PowerVR-based GPUs for mobile devices.
Most notably, until 2017, the GPUs used in Apple’s iPhone/iPad were PowerVR GPUs designed by Imagination Technologies, for which Apple paid Imagination a license and usage fee. The two companies announced the end of their partnership in 2017, but renewed it in 2020. In addition to Apple, the PowerVR GPUs developed by Imagination are also integrated into other smartphone chips.
Now by entering the RISC-V CPU space and integrating its GPUs together seems like a smart and logical way forward. According to the report, Imagination expects to invest $150 million over the next two years in order to enter the processor design market for the RISC-V architecture. Interestingly, the company already launched a RISC-V training course back in September 2020, which is perhaps one of the earliest expressions of interest in the architecture.
The decision to enter the RISC-V space is a recent one for Imagination, and it’s not clear when the product will actually be available.
RISC-V is an open instruction set architecture (ISA) managed by the RISC-V Foundation, a not-for-profit organisation founded in 2015 that started with 29 members and has grown to a large organisation of over 2,000. This has been a key reason for its overwhelming growth since its launch. Don’t assume that its open source and licence-free nature means that it is a substandard, cost-cutting architecture, which is now very widely used in the supercomputing and machine learning markets, most recently in Esperanto’s ET-SoC-1 supercomputer processor.