Samsung Strategy and Innovation Center

Generating, collecting, and analyzing enterprise data will never be the same.

Edge Computing

Graphcore provides a microprocessor for AI and ML applications (dubbed intelligence processing units [IPU]) that allows developers to build, train, and deploy AI/ML models in the cloud using its Bow Pod hardware systems, which are designed for production deployment in enterprise data centers, private, or public cloud. The Bow IP‍U‍ is also the world’s first AI processor to use Wafer-on-Wafer (WoW) 3D stacking technology, delivering 350 teraFLOPS of AI compute.In addition, the company’s Poplar graph framework software is designed with the IPU and integrates with ML frameworks to port existing models and run AI applications. Graphcore also provides a set of open-source code graph libraries to support ML models and an engine that executes models and streams data through models running on IPUs.Graphcore was <a href="https://sp-edge.com/updates/31430">acquired by Softbank</a> in July 2024 to bolster its AI capabilities and tap into the growing demand for AI hardware. Graphcore will operate as a wholly-owned subsidiary of SoftBank, retaining its name, headquarters in Bristol, and additional offices in the UK, Poland, and Taiwan.Key customers and partnershipsThe company serves customers in the finance, biotech, research, and consumer internet sectors. Notable customers include Citadel, University of Oxford, Imperial College London, Carmot Capital, and Stanford Medicine.It partnered with Spell, a company specializing in operationalizing AI, to develop IPU-based infrastructure that allows developers to run AI applications quickly and easily at scale. In addition, it collaborated with Aleph Alpha to enable bulk semantic embedding on the latter’s Luminous-Explore model supporting semantic search.Funding and financialsIn December 2020, the company raised USD 222 million in a Series E funding round led by Ontario Teachers' Pension Plan Board at a valuation of USD 2.77 billion. The funds were used to accelerate future IPU silicon, systems, and software development and drive further global expansion.In 2022, the company reported revenue of GBP 2.7 million (USD 3.3 million), 46% lower than the previous year, and its net losses worsened 18.4% YoY to GBP 198.8 million (USD 245.9 million).

The toolkit to build machine learning applications at scale

Machine Learning Infrastructure

Graphcore provides a microprocessor for AI and ML applications (dubbed intelligence processing units [IPU]) that allows developers to build, train, and deploy AI/ML models in the cloud using its Bow Pod hardware systems, which are designed for production deployment in enterprise data centers, private, or public cloud. The Bow IP‍U‍ is also the world’s first AI processor to use Wafer-on-Wafer (WoW) 3D stacking technology, delivering 350 teraFLOPS of AI compute.In addition, the company’s Poplar graph framework software is designed with the IPU and integrates with ML frameworks to port existing models and run AI applications. Graphcore also provides a set of open-source code graph libraries to support ML models and an engine that executes models and streams data through models running on IPUs.As of January 2024, Graphcore operated in the UK, US, Germany, Poland, and China. In February 2024, the company announced that it was exploring a <a href="https://sp-edge.com/updates/26706">potential sale</a> to foreign buyers after struggling to capitalize on the AI boom and dealing with heavy losses.Key customers and partnershipsThe company serves customers in the finance, biotech, research, and consumer internet sectors. Notable customers include Citadel, University of Oxford, Imperial College London, Carmot Capital, and Stanford Medicine. It partnered with Spell, a company specializing in operationalizing AI, to develop IPU-based infrastructure that allows developers to run AI applications quickly and easily at scale. In addition, it collaborated with Aleph Alpha to enable bulk semantic embedding on the latter’s Luminous-Explore model supporting semantic search.Funding and financialsIn December 2020, the company raised USD 222 million in a Series E funding round led by Ontario Teachers' Pension Plan Board at a valuation of USD 2.77 billion. The funds were utilized to accelerate future IPU silicon, systems, and software development, and drive further global expansion. In 2022, the company reported revenue of GBP 2.7 million (USD 3.3 million), 46% lower than the previous year, and its net losses worsened by 18.4% YoY to GBP 198.8 million (USD 245.9 million).

Empowering Tomorrow's Innovations with Generative AI Infrastructure

Generative AI Infrastructure

<div>
 
 UK-based Graphcore specializes in developing proprietary processors to accelerate the development of AI applications. The company is best known for its Intelligence Processing Unit (IPU), a specialized chip designed to accelerate the performance of large language models (LLMs).&nbsp;
 
 
 Graphcore's key technological offerings include 1) Bow IPU Processors, 2) Poplar Software, and 3) Bow-2000. Through these technologies, Graphcore offers cloud IPUs and data center IPUs to enable developers to construct, train, and implement AI/ML models in private and public cloud environments.
 
 
 Bow-2000 and the Bow IPU processors make up the Graphcore Bow Pod hardware system, which offers different scales of AI compute needs from exploration to large-scale deployment. The system aids in accelerating and optimizing the training and inference processes for extremely large and resource-intensive machine-learning models.
 
 
 The Bow IPU processors use Wafer-on-Wafer (WoW) 3D stacking technology, delivering up to 350 teraFLOPS of AI compute. This technology improves performance by 40% and power efficiency by up to 16% compared to previous generations.&nbsp;&nbsp;
 
 
 Furthermore, the Poplar SDK is a software stack designed specifically for its IPU, facilitating machine intelligence development. It integrates with popular machine-learning frameworks, allowing straightforward model porting and development.&nbsp;
 
 
 Notably, Graphcore claims that its IPUs offer precise, efficient computing capabilities supported by ample locally distributed SRAM. Users can also transition from their current ML setups with minimal code adjustments.
 
 
 Graphcore has encountered difficulties in securing large-scale corporate clients ever since losing Microsoft as a client in 2020. In the last quarter of 2023, Graphcore laid off its workforce in China and announced its decision to cease product sales in the country. This move was prompted by US export restrictions targeting the sale of advanced AI chips, which aimed to limit China's access to high-end semiconductor technology with components from the US. Hence, as of February 2024, Graphcore was reportedly exploring a sale to foreign buyers.&nbsp;
 
 
 Key customers and partnerships
 
 
 Graphcore's products are tailored for enterprises, serving various industries such as finance, healthcare, scientific research, consumer internet, and telecom. Its key partnerships include collaborations with UbiOps (August 2023) to improve ML and AI workloads and with G-Core Labs (June 2022) to provide Graphcore's IPU technology to G-Core Labs to scale AI integration.&nbsp;
 
 
 Funding and financials
 
 
 In December 2020, Graphcore raised USD 222 million in Series E funding in a round led by the Ontario Teachers&rsquo; Pension Plan, bringing total funding to USD 710 million. The company planned to utilize the funding to expand its global presence and accelerate the development of IPU silicon, systems, and software.
 
 
 In FY2022, Graphcore's revenues declined by 46% to USD 2.7 million from USD 5 million in FY2021, attributing the decrease to challenges in the broader macroeconomic landscape and delays in hardware purchases by important strategic clients, including a major buyer based in China. Pre-tax losses also increased by 11% to USD 204.6 million.
 
 
 
</div>

Graphcore is a startup that develops a microprocessor designed for AI and machine learning applications. It has created a new processor, the Intelligence Processing Unit (IPU) for artificial intelligence. The IPU’s unique architecture means developers can run current machine learning models orders of magnitude faster.  The company was founded in 2016 and is headquartered in Bristol, United Kingdom.

Created by mobile and hardware veterans from Google, Apple, Nest, Amazon, Twitter, and Netflix, Afero is a comprehensive and secure platform for cloud-connected devices in any industry. The Afero turnkey architecture incorporates a secure Bluetooth® Smart module, scalable cloud services, and a range of development tools that enable companies and developers to quickly prototype and build connected devices, then gather meaningful analytics. Afero connects your smart devices and enables them to work together. Afero empowers entrepreneurs and companies to improve the way people communicate with each other and everything around them.

Sentiance is the most complete provider of motion insights. It is an intelligence-driven data science and behavior change company. We turn motion data into contextual insights and use behavioral change techniques to personalize engagement for safer and sustainable mobility experiences. Sentiance empowers the world's leading companies to create personalized engagement services and products through intelligent contextual recommenders, delivering technology for a new human-centric economy where the user is in control of data.

Keyssa reinvented the connector with Kiss Connectivity. It’s a tiny solid-state system that makes device-to-device data flow simple, secure and instantaneous - freeing product designers and consumers from metal connectors, cables and wireless networks.  Kiss Connectivity is a new way to transmit huge amounts of data and video rapidly between devices in close proximity, with virtually no battery drain. Keyssa’s coffee bean-sized connector combines breakthroughs in mechanical design, electromagnetics, materials science, and semiconductor technologies. Because this tiny system uses extremely high frequency (EHF) signals to transmit information securely using standard protocols, it avoids the manufacturing and consumer challenges of today’s other wired or wireless-network options for connecting devices.

Ineda Systems is a startup company that was created by industry veterans from the US and India with the ultimate goal of becoming a leader in developing low power SOC’s for use in both consumer and enterprise applications. The advisory and management team has world-class experience of working in both blue-chip companies as well as fast-paced technology start-ups. Ineda's expertise is in the area of low-power SOC/IP development, architecture, and software that is necessary to design silicon and systems that would be of use in future applications. Their products will address the wearable platforms from a ground-up manner including vertical engineering of the wearable device: silicon, system, software, application ecosystem. In addition, they will enable the ecosystem with API’s and the application development framework. In combination with their leading-edge technology Ineda silicon will be the first in the world with power usage in the envelope required by wearable devices.

Samsung Strategy and Innovation Center (SSIC) is focused on identifying and nurturing new technologies in promising new areas such as mobile health and wellness, data center/cloud computing, Internet of Things, human-computer interface technologies, privacy/security, and more. Headquartered in Silicon Valley, with offices in Korea, Israel and the U.K., SSIC’s mission is to develop and accelerate groundbreaking technologies through open innovation working in collaboration with entrepreneurs and strategic partners to enhance people’s lives and open new possibilities.

Samsung Strategy and Innovation Center

Portfolio

About EDGE

Contact Us

Resources

Area of Focus

Technology Research Made Effortless