Arm, the world's largest semiconductor intellectual property (IP) company and the monopolist in the mobile chip design market, is the biggest beneficiary of the generative artificial intelligence (AI) investment boom. As AI ecosystem leaders such as Nvidia and Amazon Web Services (AWS) adopt Arm's "custom" central processing unit (CPU) design platform as mainstream, Arm, which has long used low-power, high-efficiency chip design as its weapon, is expanding its business portfolio into data centers.
Mohamed Awad, Arm senior vice president and head of the Infrastructure Line of Business, said in an interview with ChosunBiz on the 27th at the Westin Josun Hotel in Sogong-dong, Seoul, "Arm's success lies in the corporations DNA that has long made the most power-efficient and flexible customer response a tradition in the semiconductor industry." He noted that the know-how, data, human asset, and software that maximize chip design efficiency honed in the mobile market have also shined in the data center market, which requires high-performance computing.
Arm grew into a global corporations as it monopolized mobile application processor (AP) design alongside the opening of the smartphone market in the 2000s. At the time, many semiconductor giants such as Intel, AMD, and Nvidia jumped into the mobile chip market, but they could not catch up with Arm's reduced instruction set computer (RISC) architecture. Apple, Qualcomm, and Samsung Electronics, among others, responded quickly and flexibly to the low-power, high-efficiency chip trend based on this design, and for nearly 20 years, more than 90% of the mobile market still uses Arm's design as its backbone.
Arm then turned to high-performance computing. As the deep learning boom took off in 2012, sparking an AI data center surge, it expanded its business portfolio. Awad said, "Demand began to emerge to replace heavy and expensive data center CPUs with efficient and lightweight design," adding, "Arm developed a data center-specialized design called Neoverse and leveraged a vast software asset to support chip corporations."
Awad emphasized power efficiency, custom design, and simplicity as the most important keys to success in the semiconductor market for AI data centers. He said, "In data center chips, power is an issue that cannot be overstated. With AI service demand added on, data centers are unable to keep up with demand."
He continued, "In this process, power consumption at data centers is becoming the biggest stumbling block. There is even talk that some corporations are building nuclear power plants to supply electricity to data centers," adding, "This is mainly due to heat management in data centers and because it is becoming difficult to handle the power required on the grid."
He explained, "The reason Arm is in the spotlight is that it can implement CPUs that deliver the same or higher performance than other architectures at low power, which is the long-honed know-how from the mobile market and is also the identity of the corporations," adding, "Arm's customer-friendly, unique business model helps fabless (chip design) corporations design chips simply, easily, and flexibly to meet their own needs."
He drew a line on recent foreign reports that Arm would enter fabless and manufacturing. He added, "Arm directly designing and producing chips is intended (not so much to sell chips directly) to help corporations that use Arm's designs. Arm is the only corporations that can provide not only design licenses but also data on how customers should design and manufacture chips." The following is a Q&A with Senior Vice President Mohamed Awad.
-Arm is synonymous with low power and high efficiency. Its dominant position in the mobile chip market is thanks to that. What allowed it to gain a foothold in the high-performance computing (HPC) market such as data centers?
"Arm has historically been well known for low-power chip design. When the company was first founded 35 years ago, it expanded its presence centered on portable devices and battery-powered equipment. And after the smartphone market opened up, it firmly established its position. What Arm has found in recent years is enormous demand in data centers for efficiency in performance per power comparable to that of mobile devices.
The most differentiating element of Arm is that it provides the most efficient designs in terms of power and performance. Data centers and AI devour enormous computing resources. As corporations build their own systems to provide various AI services, they have had to operate this computing infrastructure more efficiently (than conventional general-purpose servers) to be competitive."
-As demand for AI services surges, there is a frenzy of data center investment. The problem is the power expense after establishment. Do you think Arm has offered a meaningful alternative?
"Right now, we are watching people around the world consume vast amounts of data on mobile devices, and that data is flowing into infrastructures such as data centers. Recently, new demand for AI services has been added, and the processing capacity of existing data centers cannot keep up.
To process vast amounts of data, you need commensurate power, which is nearly impossible with existing methods. Some corporations have even mentioned building nuclear power plants to supply electricity to data centers or attempting to place data centers underwater or in space. Some cities have gone so far as to ban the establishment of data centers within city limits due to heat management issues. The existing power grid cannot withstand the AI era. In this market, Arm's architecture has begun to shine. Arm's low-power, high-performance design is made to run CPUs with the same performance at lower power."
-In the past, big tech companies mainly used general-purpose CPUs specialized for high-performance computing in data centers. What differentiators has Arm put forward?
"Arm has a unique business model. By providing flexible, customer-friendly design, both fabless and IT corporations can use it as they wish. They can use the standard design as is, or optimize it for their own services. The point is that we have an open business model that can collaborate with partners in various ways.
This is a singularity that semiconductor corporations dominating the existing data center market cannot provide. In the early days of the data center business, IT service corporations, including big tech, used existing general-purpose CPUs. They had excellent performance but were heavy, expensive, and had high power expense. Demand has changed now. Various IT service corporations, including Nvidia, want custom CPUs tailored to the characteristics of data centers. Simply put, it is no longer an era of building data centers to fit chips, but of designing chips to fit data centers."
-One of Arm's strengths is speed. It released new designs every year in the mobile chip market. Does this strength apply in the data center field as well?
"Yes. Arm's differentiator is that it has a broader partner ecosystem than anyone else. Not only semiconductor corporations that design and manufacture chips, but also electronic design automation (EDA), IP (intellectual property), and even foundry (semiconductor contract manufacturing) corporations are all tied together in an organic integrated system called Arm Total Design (ATD).
The core of this approach is to simplify every process of chip design and manufacturing. By efficiently integrating the traditionally complex series of steps from design to manufacturing, we made it possible to reach the market the fastest and at minimal expense. Arm's constant goal is time to market."
-Do business portfolios beyond mobile, such as the Internet of Things (IoT), automobiles, and robotics, also work to your advantage?
"One of Arm's greatest strengths is that it participates in every technology market. In fact, Arm is the only corporations that provides design from the smallest sensors to the largest data centers. Arm exists in almost every field, including mobile, laptops, automobiles, and robotics. It is true that this portfolio gives the company a unique perspective."
Right now, we are observing computing demand increase across all markets along with AI. The nature of demand is also gradually changing because of AI. The world will increasingly revolve around AI, and the semiconductor industry must also establish a flexible response system."