Graphcore artificial intelligence (AI) processors allegedly outperform current hardware in an industry that may snag a $12bn market share by 2025.

As an indication of market confidence, Graphcore has secured a $50 million Series C funding round by Sequoia Capital as the machine intelligence it prepares to ship its first Intelligence Processing Unit (IPU) products to early access customers at the start of 2018. The funds enable Graphcore to accelerate growth to meet the expected global demand for its machine intelligence processor and will be dedicated to scaling up production and building a community of developers around the Poplar software platform.

 

Benchmarks for the IPU demonstrate that it can improve performance of machine intelligence training and inference workloads by 10x to 100x compared with current hardware. One test showed that a developer could use eight IPU PCIe cards to run a training model in the same amount of time as 128 GPU cards.

 

Such advances in performance enable developers to run existing machine intelligence models in orders of magnitude faster, using orders of magnitude less power. More importantly, the IPU's unique architecture and efficiencies will enable researchers to undertake entirely new types of work, which are simply not possible with current hardware. In turn, these will drive the next great breakthroughs in general machine intelligence.

 

Graphcore maintains close partnerships with the AI community - across academia, institutions and corporations. Graphcore has already attracted investments from major names in machine intelligence including Demis Hassabis, co-founder of DeepMind, Zoubin Ghahramani of Cambridge University and Chief Scientist at Uber and Pieter Abbeel, Greg Brockman, Scott Grey and Ilya Sutskever, from OpenAI. The Series C round has full support from existing investors: Amadeus Capital Partners, Atomico, Robert Bosch Venture Capital, C4 Ventures, Dell Technologies Capital, Draper Esprit, Foundation Capital, Pitango Venture Capital and Samsung Catalyst Fund. For more info, visit Graphcore.