By Adrian Bowles
(Aragon Research) – Intel has quietly been building out its competencies in AI. They have been buying AI startups, investing in education and open source initiatives, and now they are consolidating and coordinating their efforts by putting all AI R&D together in a new Artificial Intelligence Products Group (AIPG) and launching an AI lab under Naveen Rao. Rao, a Qualcomm alum who was the CEO of Nervana when it was acquired by Intel last year, will serve as VP and GM of the AIPG, reporting directly to Intel CEO Brian Krzanich.
With all that investment, Intel is still a puzzle in the AI space, where Qualcomm has advantages in neuromorphic hardware and nvidia is seen as the name for the emerging AI GPU-based architecture model. In this blog we take a quick look at how Intel reached this point, and what it means to buyers.
A Quick History Lesson
2015
- Intel acquired Saffron, a firm that was launched in 1999 to build an associative memory platform to support learning applications. This followed an earlier investment by Intel Capital.
2016
- Intel announced an automated driving group to work on technology for autonomous vehicles
- Partnered with Microsoft on project EVo, to bring Cortana into the home market
- Acquired Nervana and launched the Intel Nirvana AI Academy with machine learning/deep learning training & tools.
2017, First 100 days
- Intel acquired Mobileye for $15.3B + for computer vision technology
- Open sourced BIGDL, their Apache Spark-based deep learning library
- Announced collaboration with Preferred Networks (Japan) on an open source deep learning framework, and
- Intel announced a program in India to provide AI developer education (including machine learning) to 15,000 professionals.
Of all these moves, the Nervana acquisition was probably the most significant. Nervana was founded in 2014 by three former Qualcomm employees, and raised approximately $24M in venture funding before being acquired by Intel for a reported $400M+, all before establishing a clear market presence. Nervana started with an aggressive hardware play, then focused on using their hardware to optimize their own cloud-based services before being acquired by Intel.
Market and Mindshare Battles: Intel vs. Qualcomm and Nvidia
Intel still dominates the laptop, desktop and supercomputer markets. However, as enterprise computing markets bifurcate into mobile and cloud, that dominance cannot be taken for granted. Intel has a much more precarious position in AI and the IOT, two of the hottest growth markets.
Qualcomm leads in the commercial neuromorphic handset component space with their Zeroth chips and their Snapdragon 835 processor brings flexible deep neural networks to the small screen. The acquisition of NXP last October gives Qualcomm a strong IOT and automotive position. Nvidia has strong position in the rapidly growing GPU space, which is becoming critical for machine learning workloads. Servers with Nvidia GPUs are already available as options for AI developers in Microsoft Azure, IBM Bluemix, and Amazon Web Services. IBM is also heavily promoting their new Power System AI server, which uses Nvidia’s CPU:GPU NVLink technology.
Bottom Line: Intel vs. Inertia
Intel’s AI moves are encouraging, and could shake up the market. Instead of resting on their conventional CPU laurels, they are clearly positioning the firm for a leadership position in AI infrastructure. They aren’t there yet, but if the AIPG succeeds and drives corporate strategy—rather than getting crushed by the status quo—the next 12 to 24 months could see a shift in perception and market forces.