Intel Corporation (INTC 1.74%) is probably the world's most well-known chipmaker, but it completely missed the transition from desktop computer processors to mobile computing. After a decade of failed or unpopular products, the company shelved its efforts in the smartphone chip market.
The latest trend dominating tech circles is that of artificial intelligence (AI). The ability of these advanced algorithms to recognize patterns and make associations from vast stores of data has led to a number of groundbreaking developments. Speech recognition used by digital assistants and smart speakers, the tagging of friends on social media, and even self-driving cars are all the result of advances in AI.
Intel is determined not to miss this trend. Intel CEO Brian Krzanich revealed in a blog post earlier this week that the company plans to ship its first family of processors designed specifically for AI uses by the end of the year, and the company has gotten some help from an unexpected source.
Intel's Nervana Neural Network Processor is the company's first chip dedicated to AI applications. Image source: Intel.
Get by with a little help from my friends
Intel's entry into the space began with its acquisition of AI start-up Nervana for $400 million last year. The company developed an application specific integrated circuit (ASIC) called the Nervana Engine, eliminating features present in a CPU not necessary for AI applications. It also possessed a re-engineered memory, and Nervana claimed it achieved 10 times the computing power found in current GPUs.
The fruit of that labor is the Nervana Neural Network Processor (NNP) formerly codenamed Lake Crest. Intel revealed that it has been collaborating with Facebook, Inc. (META -2.28%) on the design of the chip. "We are thrilled to have Facebook in close collaboration sharing its technical insights as we bring this new generation of AI hardware to market," Krzanich said. By having Facebook and others provide input early in the process, Intel hopes to avoid any design pitfalls and create a ready market when the processor is released later this year.
Intel will make the chips available to Facebook and a limited number of its partners for evaluation and feedback before embarking on the second version due next year.
There's a lot riding on this
The area of artificial intelligence has been dominated by graphics processing units (GPUs) from NVIDIA Corporation (NVDA -2.48%). The parallel processing capabilities used for rendering graphics in games also made the chips perfect for training AI systems. Intel's desktop CPUs are great at what they do, but can't hold a candle to GPUs for AI applications. As a result, NVIDIA has grown its AI revenue by over 175% year over year in each of the last four quarters and its stock has been on a tear, nearly tripling in the last 12 months.
Intel and other tech companies have been racing to build a better solution and grab a piece of the nascent market. Research firm Markets and Markets estimates that AI chipset sales will achieve a compound annual growth rate of 62.9% over the next five years, and will be worth over $16 billion by 2022.
Google's TPU is designed for AI applications. Image source: Google.
Earlier this year Google, a division of Alphabet Inc. (GOOGL -1.82%) (GOOG -1.80%), announced the latest version of its Tensor Processing Unit could handle both AI training and inference -- the job a system does once it's trained -- though Google is using the chips for its cloud computing service and they aren't currently for sale.
Better late than never?
Some believe that Intel is late to the AI revolution that began nearly five years ago, but the industry is still in its infancy. Intel's CEO has stated that the company was on track to achieve "100 times greater AI performance by 2020." The company has yet to share any benchmarks, though it plans to provide performance metrics later this year.
If Intel can engineer technology that can outperform current offerings, it may have arrived just in the nick of time.
