Graphcore Brand Guidelines: Bristol’s IPU Chip Designer, Now Part of SoftBank

Two former Icera directors and a new kind of processor.

Graphcore is a British chip designer based in Bristol. It was formally founded in 2016 by Nigel Toon and Simon Knowles, according to English Wikipedia and The Guardian, which reported in December 2020 that plans for the company were first drawn up in late 2013 by the pair, who had both been directors at Icera, a maker of chips for 3G mobile networks that was sold to Nvidia in 2011. In a post on the company’s blog in July 2026, Toon dated the idea to a lunch in Bath in January 2012 and the company itself to July 2016; he wrote that the two spent the years from 2012 to 2014 meeting researchers in artificial intelligence, many of whom became early investors, and the next two years building a team.

The name describes the idea. Toon wrote that the company was built around a native graph-execution approach to parallel programming, “which ultimately is why we called the company Graphcore.” Machine-learning models are described as computational graphs, made of vertices joined by edges, and the company calls its chip a graph processor rather than a scalar CPU or a vector GPU.

The first outside money came in the autumn of 2016, in a round led by Robert Bosch Venture Capital with backers including Samsung and Amadeus Capital Partners, English Wikipedia says. Atomico led the next round in July 2017, and Sequoia Capital invested later that year. By December 2018 Graphcore had passed a valuation of one billion dollars, and its investors included Microsoft, Samsung and Dell Technologies. The Guardian reported a further round in December 2020, led by the Ontario Teachers’ Pension Plan, and noted that in 2019 the company had staff in Norway, the United States, China and Taiwan as well as Britain.

Colossus, Poplar and the IPU.

Graphcore calls its processor the Intelligence Processing Unit, or IPU. English Wikipedia describes it as a massively parallel chip that keeps the whole machine-learning model inside the processor, relying on memory spread across its many cores rather than on a conventional cache hierarchy.

The product names come from British computing history. The company named its first IPU Colossus after the codebreaking computer built by Tommy Flowers and his team and delivered to Bletchley Park in January 1944, and named its software stack Poplar after the district of east London where Flowers was born, Toon explained on the company blog in June 2021. The Guardian also noted the link to the wartime machine in 2020.

The first chip, the Colossus GC2, was announced in July 2017 as a 16-nanometre processor and became available in 2018, packaged two to a PCI Express card, English Wikipedia says; Toon wrote that the first chips shipped to the company’s first customer, Microsoft, in 2018, and Microsoft Azure offered the IPU in preview from November 2019. In July 2020 Graphcore introduced its second generation, the Colossus Mk2 GC200, made on TSMC’s 7-nanometre process, along with the IPU-Machine M2000, a 1U server blade holding four of the chips, and the rack-scale IPU-POD systems. In 2022 Graphcore and TSMC presented the Bow IPU, a GC200 die bonded face to face with a second die that delivers power, English Wikipedia says.

Poplar, the software, translates models written in frameworks such as TensorFlow and PyTorch into programs for the IPU. Its graph compiler can also draw the computational graph it builds, and in 2017 Graphcore published a set of these images, which it said looked like a scan of a human brain once the connections were revealed.

Design and documentation.

AI chip companies sell something most of their customers will never see. The processor sits inside a server, the server sits in a rack, and the rack sits in a data hall far from the people who use the models running on it. So the industry’s visual language has grown up around making that hidden hardware visible: a whole silicon wafer, as wide as a dinner plate, held up on a keynote stage; a single chip pinched between finger and thumb for the cameras; a die shot recoloured in neon false colour so that its blocks of memory and compute read like a city map at night.

The second set of images is the data centre. Press photographs tend to show long rows of racks in dark halls, lit by the glow of status lights and cold aisles, with cabling arranged as neatly as a diagram. These pictures stand in for scale, which is what buyers of AI compute are being asked to imagine, and they appear again and again in launch announcements, partner news and investor material.

Keynote slides do much of the remaining work. Model names, token counts and benchmark charts appear beside the chip’s own product name, and the companies’ identities are usually kept spare to let those numbers and images lead: black or white grounds, a single accent colour, a sans-serif wordmark and very little ornament. Several firms have adopted coined product categories, names for their processors that set them apart from the graphics processor.

The shop window is the technical conference. Hot Chips, NeurIPS and the AI Infra Summit are among the events where architectures are presented, demonstrations run and booths compete for engineers’ attention. This guide looks at how each company in the group presents itself, and keeps to what can be checked about who founded it, who owns it and what it makes.

A SoftBank company under its own name.

Graphcore’s independence ended in 2024. English Wikipedia records that Meta Platforms took on Graphcore’s AI networking team in early 2023, and Toon wrote that Graphcore entered a close partnership with SoftBank Corporation the same year. On 12 July 2024 the trade publication Verdict reported that SoftBank Group had acquired Graphcore for an undisclosed amount and that it had become a wholly owned subsidiary; the company’s own announcement said it would continue to operate under the Graphcore name, with its headquarters in Bristol and offices in Cambridge, London, Gdańsk and Hsinchu. SoftBank already owned Arm, the Cambridge chip designer, which The Guardian described as SoftBank-owned in 2020.

Under SoftBank the company has grown. In October 2025 Graphcore announced an AI engineering campus in Bengaluru, India, with 500 new jobs planned over a decade, and said its British headcount would double to about 750. In June 2026 it opened an office and engineering lab in Taipei’s Nangang district, alongside its site in Hsinchu, the company says, to work more closely with its manufacturing and supply-chain partners in Taiwan. In July 2026 Toon stepped down as executive chair; his farewell post said the company was approaching 1,000 employees, had opened development centres in Austin, Texas, and Bengaluru, and planned to move into a purpose-built headquarters in Bristol in September 2026. The company’s about page now describes it as a wholly owned subsidiary of SoftBank Group Corp.

Graphcore designs its chips and relies on a foundry to make them: English Wikipedia names TSMC as the manufacturer of the GC200 and the partner on the Bow IPU. The company has not published a colour standard or typeface for its identity.

For a design guide, Graphcore shows how a chip company can borrow its product names from national computing history and keep its own name after a takeover. Compare Arm, the other British chip designer owned by SoftBank, and Nvidia, which bought Icera, the company its founders came from. For this guide we use a deep plum, a pale lilac and a coral with a sans, an unofficial interpretation; no product or logo is reproduced.

Keep Graphcore’s identity on hand — save it to a board in the Made Good inspiration library.