Demands for accelerating system processing continues. Domain specific development infrastructures and entirely new computers are appearing. Consequently, the ability to select a high-speed infrastructure that enhances performance is now a requirement for businesses.
The competitive advantage of digital businesses hinges on the strategic selection of IT infrastructure. Previously, the performance of applications determined competitiveness and general-purpose processors were the mainstream. Now things have changed. New applications have emerged based on deep learning that cannot be handled by existing processors. In addition, mobile has become the most important point of customer contact. As a result, businesses require service design, which often places the greatest priority on customer experience.
The ability to use AI will undoubtedly have a powerful influence on the relative merits of modern business. The essential component behind today’s IT infrastructure is the graphics processing unit (GPU), which provides the massive parallel processing capability to support AI and deep learning. Businesses everywhere have implemented these high-speed processors at scale to support innovation, while others have developed unique AI processors to gain a competitive edge.
Smartphones are now the most critical direct point of contact with customers for all kinds of digital businesses. High-speed, low-power consumption mobile processors determine the operability, usable biometrics, high-resolution photo quality and longer battery life of smartphones. The smartphone industry has matured considerably making it possible for firms to assemble standard parts and enter the market relatively easily. Existing companies vying for the largest market share, like those shipping 200 million units annually, or new entrants in emerging markets are focusing on developing proprietary processors with distinct features to offer unique functions.
Mobility is without a doubt the next realm of competition for digital businesses. For example, certain functions are required for situation recognition and decision-making in autonomous driving technology, such as sensor fusion, deep learning and cloud connection. IT infrastructure that can support these functions close to where it is required even under severe conditions and with low power usage, is collectively referred to as edge processing. Businesses are betting their future supremacy on acquiring or developing a processor that underpins an organizational plan. Consequently, an increasing number of businesses consider IT infrastructure as part of core strategy. The need to develop one’s own hardware has arisen from the pursuit of improvements in points of customer contact in a bid for a competitive advantage in service.
Miniaturization of semiconductor manufacturing has helped accelerate the processor, which is the core of IT infrastructure. This technology has also downsized transistors embedded on processors, which has allowed businesses to benefit from shorter processing time and power savings without modifying software assets. Miniaturization is continuing, with gate lengths now down to 7 nm in the latest manufacturing process. Gate lengths of 5 nm are even within reach by using extreme ultraviolet lithography (EUV).
The technology, however, driving continuous miniaturization over many years has grown complex and progress has slowed. In fact, only one firm was able to ship products with gate lengths reduced to 7 nm in 2018. Others fell behind schedule in supplying such products or dropped out of the market. Technological complexity and cost pressures are likely to inhibit further rapid miniaturization. Business strategies premised on ever-faster IT infrastructure will therefore require reevaluation.
As miniaturization becomes a less likely solution for further acceleration, purpose-specific processors have come to the forefront. These processors focus on specific functions to reach higher speeds and efficiency when combined with specialized software. One example is an AI-specialized processor. New processors have emerged embedded with algorithms specializing in AI learning or inference that limit functions depending on the framework. Since offering high-speed, high-efficiency AI will help innovate service and add value to the use of cloud environments, major cloud vendors are presently developing original AI processors.
Edge processors are also being used for various purposes such as autonomous driving and module-connected robots for collaborative sensing and work. Purpose-specific processors offer a solution to the issue of rapid obsolescence in digital businesses. For instance, processors that respond flexibly to changes have been proposed for machine-learning applications. These processors adapt and reconfigure to changes in learning frameworks. Such innovation is expected to continue using recently developed open source processors.
Current computers are still unable to operate near the capacity of the human brain. For example, supercomputers require more than five minutes to simulate one minute of the human brain’s neural network, which is roughly 0.5% of the entire brain’s capacity. Accordingly, computers capable of high-speed processing of massive data, such as for fluid dynamics and drug discovery, continue to be in high demand.
Quantum computers could transcend the performance of current computers as they use quantum mechanics to perform massive, parallel processing functions through quantum gates. Numerous businesses have invested heavily and have launched development efforts in commercializing the quantum gate method, which is expected to have versatile qualities. Meanwhile, other vendors have invested in the quantum annealing1 method, which specializes in optimizing combinations.
Of note is that operation of these new computers will contrast starkly from the way computers are coded today, making it difficult for present-day computer engineers and programmers to use. Moreover, quantum computers will be custom-designed for individual business needs, and cross-section use will likely require massive revisions in formula development. The same limitation applies to purpose-specific processors. The use of such infrastructure will require a high-quality development environment and a specialized software development community. As a result, not all companies will be able to pursue this strategy. In short, today’s business has entered an age requiring an even more strategic selection of IT infrastructure and development partners.
1Quantum annealing is a metaheuristic for finding the global minimum of a given objective function over a given set of candidate solutions (candidate states), by a process using quantum fluctuations.