The speed, the complexity and the quantity of data to analyze is increasing so much that a new generation of processor is needed to face this explosion, both on Cloud and at the Edge.
Computing at the edge (“Edge Computing”) is therefore booming and is complementary to computing in the Cloud. Edge Computing is required when local processing becomes critical for low latency issues (e.g. Augmented Reality /Virtual Reality), for privacy of data (e.g. manufacturing enterprises), for conservation
of bandwidth (e.g. smart factory IoT) or for high performance localized compute (e.g. automotive).
Other strong evidence of the surge of the amount of data to analyze is the growth in artificial intelligence semiconductor solutions.
According to McKinsey, it will be multiplied by 4 from now to 2025. Today’s semiconductor market for AI is 99% in the Cloud ($ 5.5 billion) and 1% Edge. By 2025 the market will evolve into a market of $ 5.5 Billion in the Edge and will reach $14B on the Cloud ! Both Cloud and Edge demands are exploding.
A few examples: 5 billion videos watched every day on YouTube, more than 1 gigabyte per second created in an autonomous car, more than 200 million gigabytes generated per day in one smart city and more than 5 million gigabytes per day in an intelligent factory.
According to the Cisco Global Cloud Index* 2018-2021, only 25% of data generated today reaches a centralized data center on the Cloud! More and more, data is transient and neither recorded nor stored. This data must therefore be processed on the spot, where it is created, at the Edge of the Cloud. This is new.
For details, click Kalray_white_paper_MPPA_DPU_Coolidge.
