Japan's SoftBank Group has purchased British chip designer Graphcore; the group still maintains substantial control over its previous semiconductor venture, Arm Holdings.

The Japanese investment group SoftBank Group, which owns a number of properties, including CPU maker Arm Holdings, has acquired Graphcore, a semiconductor business focused on artificial intelligence, in what is a noteworthy second semiconductor acquisition.

According to the terms of the agreement, Graphcore, a company established in Bristol, United Kingdom, will become a fully owned subsidiary of SoftBank, retain its identity, and carry on producing highly skilled jobs across a variety of disciplines. Since Graphcore is a privately held company, the deal's financial details were kept a secret.

Graphcore, which was founded in 2016, is one of several firms hoping to make money in the extraordinarily lucrative AI sector. Numerous investors have contributed money to it, including Dell and Microsoft.

The Intelligence Processing Units (IPUs) of the Bow family of products from Graphcore are designed especially for AI processing. The business claimed that their IPUs were competitive with Nvidia products and spent a few years marketing them. But the business has had difficulties in the past; there have been rumors of layoffs and the closure of the China office.

SoftBank, which attempted to exit the CPU market by selling Arm to Nvidia for $40 billion, may have made a significant turnaround with this sale. However, the acquisition encountered strong opposition in the United Kingdom. In September of last year, SoftBank went public with the firm despite holding the majority of Arm's shares, following the collapse of the Nvidia-Arm agreement. That has paid off; since the IPO, Arm's stock price has increased by 185%.

It appears that Softbank is attempting to enter the chip industry, and Kevin Krewell, principal analyst at Tirias Research, believes that this could be successful. According to him, Graphcore was among the first AI startups and introduced a distinct strategy for AI accelerators. Sadly, they were unable to stay up, and their design did not provide enough of an advantage over Nvidia.