Intel Pumps Funds Into Data Processing In All Shapes And Sizes
October 20, 2017 Jeffrey Burt
Intel’s multi-year effort to expand its reach beyond its PC and server processor roots has taken the chip maker down multiple paths, some of which have ended in dead ends.
The most memorable of those was the billion-plus-dollar attempt to challenge ARM Holdings and its various partners – such as Qualcomm and Samsung – in making chips for mobile devices. Under current CEO Brian Krzanich, Intel has retrenched, dropping its mobile device efforts and pulling back from wearables, and instead is pushing to provide the foundational technologies that will underpin the trends that will continue to shape the industry, from cloud computing to its telemetry extension, the internet of things, or IoT, to artificial intelligence in its many guises, to robotics, and finally to data analytics.
The vendor’s efforts were on display this week, with the announcement of the Nervana Neural Network Processor (NNP), a chip that Intel officials claim will run neural networks better than current processors and drive the development of new classes of AI applications in a broad range of industries, from healthcare to automotive. In addition, Intel this week rolled out what it calls the Speech Enabling Developer Kit, which is a bit of hardware and software that will extend the voice control capabilities of Amazon’s Alexa to other smart home devices.
They’re part of Intel’s larger strategy to touch all parts of the trends that are driving the industry, with investment dollars, acquisitions and R&D efforts going not only into hardware but also software, developer tools, algorithms, open-source projects and standards. The focus runs from silicon and 5G connectivity to AI, virtual and augmented reality (VR and AR), networking and storage, to self-driving cars.
At the core of what’s driving the industry’s evolution is data and that is where Intel is spending a lot of its investment money, according to Wendell Brooks, senior vice president at Intel and president of Intel Capital, the chip maker’s capital venture arm. Brooks was speaking at an event this week where the company unveiled more than $60 million in investments in 15 startups and small companies that focus in some way on data – including processing and analyzing it, managing it, and securing it.
It is no secret that companies are awash in data, and the amount of data being generated is only going to grow. Brooks used the fast-developing self-driving vehicle industry as an example. By 2020, every autonomous vehicle will create 4TB of data every day, he said. A million such cars will create the same amount of data every day as 3 billion people. Add in such trends as industrial automation, AR and VR, and high-definition video, and the potential for massive data growth is in place.
The average internet user in 2020 will generate 1.5 GB of traffic a day, and a smart factory will create a petabyte to data daily.
“There’s an explosion of data coming,” Brooks said. “We get very excited about this trend continuing. “
In the halls of Intel’s headquarters, Krzanich is fond of saying that the chip maker is now a data company. The trend in Intel’s investments is moving quickly in that direction. According to Intel Capital numbers, only $13 billion of Intel’s $43 billion in revenue in 2010 came from data-centric parts of the business. In 2016, that number was $26 billion of the chip maker’s overall $59 billion in revenue, or about 40 percent. Intel also is getting more aggressive with its investments, being the lead investor in 70 percent of deals and being early stage – or Series A – investor in 60 percent of the deals over the past two years. The company has been consistent in recent years, with aggregate annual investments totaling more than $500 million in two of the past three years.
The startups highlighted during the event varied in many ways, with the one constant being the focus on data, often with a machine learning hook. The companies included:
- TileDB, one of the upstart multi-dimensional array databases, which was co-developed by Intel and MIT and which is a competitor to database pioneer Michael Stonebraker’s SciDB.
- Reniac, which has created what it calls the Distributed Data Engine, a generic data platform that works on public or private infrastructure and that aims to increase throughput and decrease latency in existing applications without having to tweak them
- BigStream, which uses hardware and software accelerators to speed up Apache Spark big data and machine learning workloads, with FPGA acceleration for those who want to employ it
- Amenity Analytics, which has created a platform that takes unstructured textual data and performs sentiment analysis and other predictive analytics on that data using various machine learning techniques
- LeapMind, which is developing technologies and algorithms to improve the accuracy of neural network models and reducing the computational complexity of deep learning workloads
- Synthego, which has a software package with machine learning hooks that works with RNA kits and the CRISPR genome editing technique to do genetic engineering
- Bossa Nova Robotics, which is developing autonomous service robots to help manage inventory in large stores in the retail industry
By investing in such startups, Intel Capital not only gets a return on its investment, which is important to the company’s bottom line. It also gets deep insight into some of the innovators that are on the front end of trends, helping it better understand the larger IT sector better. Intel has been at this for a while, of course, and has made some huge investments, such as the funds it has pumped into Hadoop distributor Cloudera, to name its biggest one. Since 1991, Intel has invested a total of $12.2 billion in more than 1,500 companies, and of these, more than 640 of them have gone public.