Market Insight

Artificial intelligence will be transformative across industry domains and verticals

December 10, 2019

Luca De Ambroggi Luca De Ambroggi Senior Research Director, Artificial Intelligence (AI)

Want to learn more?
Have an expert contact you.

The powerful new technology of artificial intelligence (AI) is here to stay, disruptive in impact and pervasive in presence, its significant computational capabilities and vast potential for deep learning applicable to a sea of use cases in daily life and commerce. And together with 5G connectivity, the internet of things (IoT), and the Cloud, AI will be part of the group of transformative technologies that are now converging and intersecting in new ways, giving rise to fresh opportunities and challenges alike.

IHS Markit estimates that global revenue for AI systems covering both hardware and software components will reach nearly $1 trillion in 2025, with a compound annual growth rate (CAGR) of 28% from $142 billion in 2017, as shown in the graph below. AI adoption will be heaviest in the automotive, manufacturing, healthcare, and defense industries, with datacenter and cloud computing infrastructure forming the principal areas in which AI training will be carried out.

IHS Markit | Technology graphic of Global revenue for AI systems

These findings and many important insights are contained in the new IHS Markit | Technology report, Artificial Intelligence (AI) Technology Adoption & Impact Report – 2019, a comprehensive 78-page examination of AI that tracks its development, disruption, and impact across key industries, including four horizontal domains and six verticals.

AI and deep learning; US vs. China

Artificial intelligence, as defined by IHS Markit, refers to the body of science that studies how machines are enabled to perform independent problem-solving, derive inferences, achieve learning, demonstrate the acquisition of knowledge, and make logical decisions based on applicable inputs continually being processed by the machines and by their own accumulated store of knowledge, all executed in a split-second duration closely resembling the complex calculations made by the human brain in real-life situations and events.

A subset of AI is machine learning (ML), the set of algorithms that provides machines with the ability to automatically find and learn patterns through data fed to them. Under ML are several technologies and techniques, including neural networks and deep learning, which are computational models that try to emulate the structure and workings of the human brain, including process phases like training and inference.

Competitively, the two foremost proponents of AI are the United States and China, although the general consensus is that AI in China is slightly behind that of the United States. For now, China also trails the United States in chip technology know-how where advancements are propelling AI forward, making China very much dependent on the West for its semiconductor needs.

Yet China has numerous advantages in many other areas where AI initiatives are involved, including exceptional government support and committed funding, which could propel the country past the US to become the world’s leading repository of AI expertise in the not-so-distant future. In contrast, Europe, the only other region of significance in the AI market, remains far behind both the United States and China in its AI initiatives.

Among the more pressing issues facing AI is the question of ethics. The implementation of AI and associated machine-learning technologies in unmanned devices and robotics poses concerns on whether—and how—society will be able to maintain control of a potent and formidable technology without losing sight of the human factor and element. The ethics question brings up the need, much discussed in circles today, for a framework of ethical principles that will govern the deployment of AI and autonomous machines in representative situations and operating scenarios, and how data gathered for AI-training purposes can be regulated and made secure.

AI in four industry domains

The AI effect and impact will be most pronounced in the four horizontal industry domains of semiconductors, HMI, datacenters, and autonomous machines.


A major trend in the semiconductor market today is heterogeneous computing, defined as the use of more than one kind of processor or core, or of dissimilar coprocessors. Deployed not only in high-performance computing but also in embedded applications that can take advantage of basic machine learning algorithms, heterogeneous computing is also a solution that enables AI and ML across applications and industry domains.

As a result, new chip solutions are emerging that contain different cores and accelerators, with components possessing ML capabilities optimized for high efficiency, lower energy use, and a reduced build-cost. The existence of varying chip architectures for AI implies that a uniform one-size-fits-all approach will not be suitable for the AI solutions currently being developed. Instead, a diverse array of AI-bearing chip solutions will be available to meet the needs of differing applications.

To this end, there are slices of the AI market open to dedicated and optimized devices, such as application-specific integrated circuits (ASICs) and co-processors that are not integrated into a heterogenous integrated circuit. Instead, these ASICs and co-processors can be found located close to a system-on-chip (SoC) on the same printed circuit board, performing a sort of companion role specifically reserved for AI functions. Increasingly, system designers are turning to ASICs and SoCs as efficient mainstream and turnkey solutions.

The ASIC market for AI is, in fact, flourishing, driven by several startups that aim to offer completely new architectures in processing capability and memory interface to challenge traditional chip architectures found today in central processing units (CPUs), graphics processing units (GPUs), field-programmable gate arrays (FPGAs), and digital signal processors (DSPs).

Worth noting, is that hardware architecture optimized today for specific algorithms may become obsolete very quickly because of the rapid pace of innovation, jeopardizing investments. For these reasons, hardware—even at the silicon level—must be flexible and programmable “on the fly.” Software, meanwhile, will be driving hardware—not the other way around—in AI systems.

Such programmability will require the resolution of some critical issues, including the effective use of working memory. This is because advanced AI technologies, especially in algorithms involving deep learning, require huge amounts of volatile memory to perform properly. To this end, the chip industry is looking at various innovative approaches, including new processor architecture where memory is closer to the computational core to reduce the burden of data movement, enabling high-processing parallelism and dedicated memory cells per processing core.

The attempt at memory innovation illustrates how AI can be an enabler and creator of new services and business models, disrupting current ecosystems and supply chains. But before the full potential of AI can be harnessed, a period of settlement and consolidation of the supply chain must occur first and the value chain simplified, IHS Markit believes. Only after culling and solidification have taken place in AI can the technology be maximally utilized to enhance lives and industries alike. Consolidation will occur also in the integrated circuit (IC) market addressing AI, which always happens when disruptive technologies face and enable new markets. This means that just a few will survive in the next 10 years.


Datacenters are a key part of the industrial revolution and digital transformation being driven today by that all-important resource: data. And within datacenters, new high-performance servers have been developed with AI and ML algorithms in mind given the importance of essential performance parameters like real-time delivery and low latency in new and critical applications. By 2022, servers shipped with specialized co-processors for AI and ML will make up more than 10% of total global server shipments, IHS Markit research shows. Businesses will also be ramping up investments in servers that contain co-processors, with a preference for servers containing general-purpose graphics processing units (GPGPUs) and field-programmable gate arrays (FPGAs).

Overall, a sharp rise can be expected in processing requirements and data analytics for future equipment, whether in cloud-based datacenters or servers at the edge. Of interest is the increasing migration taking place in different industries and domains toward edge network computing. The need for low-latency performance will drive growth in more localized servers with a round-trip time (RTT) below the standard 20 millisecond (ms) range. The AI report contains more information on edge servers, including the latest shipment figures broken down by industry domain.  

Among industries investing strongly in edge networks are healthcare, banking and finance, telecommunications, and manufacturing. In healthcare, for instance, services at the edge can include telepresence, IoT sensors for remote diagnostics, and chirurgical equipment.


HMI, or the human-machine interface, is a basic but extremely powerful application that can be deployed in multiple applications and across several industries and domains, from consumer and mobile electronics to automotive, passing through smart homes and smart appliances or the industrial space. Wherever humans need to interact with machines, an advanced HMI will be able to add value. Although Siri, Alexa, and Google Assistant have been developed for consumer electronics applications, the companies responsible for the assistants—Apple, Amazon, and Google, respectively—are also trying to expand their use and interface into other domains, including the automotive industry.

But while the three players have a strong footprint in Europe and the United States, their presence is extremely limited in the fast-growing Chinese market. In contrast, because of geopolitical reasons and language competence, digital assistants from local Chinese technology companies, such as Baidu, Alibaba, Tencent, Huawei, and Xiaomi, are better-placed to succeed in China than their Western counterparts.

In automotive, legal issues are likely to slow the implementation of self-determining technologies, especially for safety-critical applications. Original equipment manufacturers (OEMs) will be subject to much greater liability if machines are to take over the role of humans, as is anticipated in the self-driving car. Even so, in-vehicle HMI for speech and gesture recognition, with its AI-based algorithms and demonstrable advantages, has been enthusiastically embraced by vehicle manufacturers and car owners alike.

The inclusion of AI capabilities in the human-machine interface can bring about several features and benefits, including added value and differentiation for brands; visibility and predictability in system performance and maintenance, owing to steady algorithms; and ubiquity across industries, given the universal applicability of HMI to a wide array of use cases spanning consumer, enterprise, and industrial concerns.

Autonomous machines and robotics

Machine learning technologies, together with advanced electro-mechanical and sensory science, are enabling the rapid development and deployment of different autonomous entities able to accomplish a multiplicity of tasks with varying levels of complexity, traditionally the domain of humans in the past.

Autonomous machines, based on the categorization carried out by IHS Markit, span a broad range of devices and equipment in various industry domains. Autonomous machines include light vehicles, buses, and trucks in the automotive space; harvesting and crop machinery in agriculture; robots in the industrial and healthcare sectors; toys, drones, and service robots in consumer; the marine and military industries; and the mining and construction industries. Machine vision typically includes lighting, a camera or other imager, a processor, software, and output devices.

The leading technology driver of AI in autonomous machines is the automotive industry, which has the most complex set of requirements for autonomous machines in its quest to create the self-driving car. However, it is the consumer market that will retain the lead in the shipment of autonomous machines, with total worldwide shipments expected to reach 13.3 million units by 2025, up from 4.5 million units in 2018.

AI in key industry verticals

The overall market for AI-based systems indicates that all industries are embracing the AI promise of additional features, higher efficiency, and greater performance. The industries covered by the IHS Markit AI report include automotive, banking and the financial sector, industrial automation and manufacturing, healthcare, video surveillance, and smartphones.

In automotive, the entire automotive supply chain—from semiconductor suppliers to original equipment manufacturers (OEMs)—is seriously investing in AI, where advanced technologies are paving the way toward realization of the self-driving, autonomous vehicle. Already, machine learning is available in the infotainment HMI of various car manufacturers, including speech-recognition technologies that rely on algorithms based on neural networks running in the Cloud. And in Advanced Driver Assistance Systems, or ADAS, deep learning applies to realms like camera-based machine vision systems, radar-based detection units, driver drowsiness, and sensor-fusion electronic control units (ECU).

Beyond infotainment, however, extra requirements in automotive AI will be called for where safety issues are concerned, driven by critical considerations in latency, performance, power consumption, and data acquisition, handling, and storage.

In banking, the business value of AI in 2018 was estimated at $41.1 billion, which included the cost-savings and efficiencies of newly introduced technologies like Blockchain, compared to the use of existing infrastructure and processes. By 2025, the business value of AI in banking will surge to $165.0 billion, IHS Markit estimates show, as AI becomes more mainstream.

Specifically, real-time delivery and low latency will be key performance parameters in the four banking areas involving stock market transactions, fraud detection, authentication, and security. In these areas, high-performance edge computing will be a key trend, marked by the increasing use of on-premises datacenters featuring turnaround time of just a few milliseconds.

In industrial automation and manufacturing, AI will enable new use cases, such as generative design, or the autonomous creation of optimal designs from a set of system requirements. With generative design, users will be able to interactively specify the functional needs and goals of their design, including preferred materials, engineering constraints, and manufacturing processes, with a manufacture-ready design resulting from the process. 

Generative design will democratize the practice of design because it will not require users to possess an engineering background or an extensive understanding of structures, mechanics, and materials; all that will be needed is an understanding of the usage and purpose of the part being designed. Generative design will also optimize the simultaneous design of multiple objectives while providing several novel design alternatives, enabling companies to substantially reduce product-engineering cycles.

In healthcare, Big Data analytics and AI will play an increasingly important role. By providing the tools to assimilate large volumes of disparate, structured, and unstructured data produced by healthcare systems, it will be increasingly possible to analyze such data in their entirety to increase the speed and accuracy of diagnosis, evaluate therapy, and expand the ability of researchers to explore disease pathologies. New business models will emerge, such as Software-as-a-Service (SaaS), featuring an annual subscription model based on the number of estimated cases for a site. The use of AI technologies will also affect elements of the healthcare continuum, such as pricing, dataset validation, equipment use and replacement, diagnostic imaging, virtual health assistants, and return on investment.

Altogether, machine learning and AI will enable the remote monitoring of more than 1 million patients by 2022, up from just 31,000 in 2017, in countries including Canada, China, France, Germany, Japan, the United Kingdom, and the United States, according to IHS Markit data.

In video surveillance, AI will be poised to revolutionize an industry affected by various factors, including the aggressive competition in pricing among video surveillance products and players in the market, the outsize presence and prevalence of China, the rise of safe cities, and the continuing move from standard-definition to high-definition video.

While the Cloud and its large, virtualized processing power is being utilized to run video analytics, some analytics—such as crowd monitoring, counting, and object detection—will be run at the camera to save bandwidth and relieve computing capacity at the backend. This means more powerful centralized analytics can be used to run processor-intensive applications, such as feature extractions of humans or vehicles, as well as object searching. Such an approach has already gained interest in the Chinese market, where AI deployments have been the most abundant, and where projects are large and can benefit most from distributed computing and analytics.

A key differentiator in video surveillance will be software in the form of video management software (VMS), video content analysis (VCA), physical security information management (PSIM), and software used by central monitoring stations (CMS).

In smartphones, where AI already accounts for a big chunk of communications revenue, IHS Markit projects that two out of every three smartphones will have prebuilt AI hardware and features by 2025. Global revenue for AI-bearing smartphones will reach $378 billion by 2025, up from $29 billion in 2017, equivalent to a CAGR of 39% throughout the period.

Leading smartphone manufacturers racing to introduce dedicated AI-capable solutions include Huawei, Apple, and Samsung.

The new IHS Markit | Technology report on AI, Artificial Intelligence (AI) Technology Adoption & Impact Report – 2019,  is offered in two research categories, Transformative Technologies and Semiconductors, and is accompanied by an Excel database with extensive market data and relevant industry statistics. IHS Markit | Technology is now a part of Informa Tech.

Contact us for more details on the report or if you wish to become a subscriber.

Share facebook Twitter Google Plus Linked In Add This Contact Us