AI systems need to process massive amounts of data quickly. While the performance of general-purpose chips has improved enough to kick-start a new generation of AI technology, they cannot keep up with the exponential increase in the volume of data that AI systems process. As a result, chip design emphasis has shifted from a race to place more transistors onto a square millimetre of silicon to focus on building microprocessors as systems, made up of multiple components, each of which is designed to perform a specialised task.

As the tech industry’s customer base consolidates, it designs more and more of its own chips and sends them straight to foundries for manufacture. This is partly because merchant market suppliers have disappointed when it comes to delivering sufficiently powerful, power-efficient chips to support AI workloads, and partly because these companies want to gain a proprietary edge with their tech stack.

Google is far from alone in having developed custom AI chips. Apple has designed its own chips for quite some time, with the ARM-based A-series on its 14th generation. Microsoft has re-engineered its servers with FPGA subsystems to achieve the power and flexibility it needs. Amazon, which has long designed router chips and servers and has its own captive chip company Annapurna, introduced its Inferentia AI chip in 2019.

However, not all companies are equal when it comes to their capabilities and investments in the key themes that matter most to their industry. Understanding how companies are positioned and ranked in the most important themes can be a key leading indicator of their future earnings potential and relative competitive position.

According to GlobalData’s thematic research report, Artificial Intelligence, leading adopters include: NVIDIA, Intel, AMD, Google, IBM and Qualcomm.

Insights from top ranked companies

NVIDIA

NVIDIA has dominated the GPU market for years. However, it faces competition from companies such as Google, Intel, and Graphcore. NVIDIA has made efforts to stay ahead. In 2019 it acquired Mellanox, which focuses on high-performance interconnected technology and will enable NVIDIA’s accelerated computing platform for AI workloads. In May 2020, NVIDIA unveiled its next-generation GPU technology called Ampere, which will become the foundation for its AI strategy and product portfolio. Ampere features the third generation of Tensor Core, a chip that’s purpose-built for accelerating AI. The A100 GPU is the first AI accelerator based on the Ampere architecture and will offer unified support for training and inference. On the software front, NVIDIA unveiled Jarvis, a new application framework for building conversational AI services. Like Intel, NVIDIA is also making its mark in the autonomous vehicle sector, adding Mercedes to its already impressive client-roster in mid-2020. In September 2020, NVIDIA announced plans to acquire UK-based chipmaker Arm for $40bn.

IBM

IBM was an early mover in AI with Watson, the ML system which beat two human champions on the American quiz show Jeopardy! in 2011. Since then, IBM has invested in bringing AI to businesses. However, Watson has struggled to live up to expectations. While its industry-specific focus appeals to clients, its implementations are overly reliant on expensive consulting services at a time when competitors are providing plug-and-play APIs, which developers tend to prefer, at a much lower price. Acquisitions may be the best path back to the top (although it needs to do a better job selecting acquisition targets than it has in the past). With the acquisition of Red Hat in 2018 for $34bn, IBM made a significant step towards building a hybrid cloud strategy. In his first earnings call in April 2020, IBM’s CEO Arvind Krishna called hybrid cloud and AI “the two dominant forces driving transformation” and stated that the company would resume its “acquisitive strategy” later in 2020.

Intel

The semiconductor giant has reengineered itself and expanded its AI platform through acquisitions. Notable purchases include AI processor makers Movidius and Nervana, FPGA maker Altera and CV chip maker Mobileye. Intel has built on these acquisitions, bringing on Vertex.ai – a deep learning start-up – in 2018 to join its Movidius unit, Omnitek in 2019 to strengthen its FPGA video and vision offering, and Moovit – an urban mobility start-up – in 2020 to accelerate Mobileye’s MLaaS offering. However, the most notable acquisition is that of AI chip start-up Habana Labs at the end of 2019 for $2bn. The acquisition resulted in Intel killing off its Nervana chips a month later. In mid-2020, Intel announced a slew of new and future developments across its portfolio of processors, emphasising enhanced AI capabilities. This included a new set of Xeon Scalable processors, suitable for data-intensive AI workloads, and a future generation of Xeon processors and Xe data centre GPUs, the latter of which will create more competition for NVIDIA and AMD. It also unveiled a new AI-focused FPGA, a third-generation Movidius vision processing unit (VPU), and the development of AI processors Gaudi and Goya from Habana Labs.

To further understand the key themes and technologies disrupting the technology industry, access GlobalData’s latest thematic research report on Artificial Intelligence.

Data Insights

From

The gold standard of business intelligence.

Blending expert knowledge with cutting-edge technology, GlobalData’s unrivalled proprietary data will enable you to decode what’s happening in your market. You can make better informed decisions and gain a future-proof advantage over your competitors.

Be better informed