While the semiconductor industry is a central component of the global data centre boom, there are other sectors that are also gaining from the explosion of Artificial Intelligence Computing, such as those that provide electricity infrastructure, transformers and advanced cooling systems. Though the AI investment frenzy is dominated by Nvidia, the sheer power consumption of running modern data centres is creating a huge opportunity for players in other key markets such as Asia to compete in the equipment delivery market.
The investment is in the magnitude of substantial. By 2030, global investment in data centre infrastructure could reach as close to $7 trillion, estimates McKinsey. The growth is primarily due to the rising adoption of artificial intelligence, cloud technology, and other data-heavy solutions. The need is spreading across the technology infrastructure supply chain, as companies build new facilities to accommodate increasingly sophisticated AI systems.
But squeezing in data centres is getting to be harder, faster and more expensive. Advancements in processors are just the first of the difficulties developers are encountering. Electric power service, electricity generators, transformers, coolers and such may take much longer than the computers do. The time it takes to connect a new facility to the electricity grid varies widely by emerging market—up to two years in some—and by major developed market—several years in some.
Digital infrastructure service provider BodaData’s CEO, Wing Kin Cheung, said that “outside the industry circle, people are talking about (graphics processing units), but within the circle, people most certainly question you about the lead time for generators and transformers”.
They are especially important in that they convert the electricity from the grid to the level needed by the servers, cooling systems and other equipment within a data centre. As the facilities grew in size and their power consumption, the provision of sufficient transformer capacity has become a growing concern in the development process.
This has put high-voltage electrical equipment manufacturers in a good position. South Korea’s HD Hyundai Electric and China’s Hainan Jinpan Smart Technology are both seeing strong demand growth tied to AI infrastructure projects, notably in North America. The demand is particularly strong, revealing the impact of the AI boom on less consumer-facing but vital industries for computing infrastructure operations.
As key technology firms turn to Europe to increase investments in data centres in Finland, Germany and Britain, among other countries, HD Hyundai Electric has also begun to be in demand. Demand from the Middle East has remained strong as well. The long-term investment cycle of the current infrastructure spending spree was evident, as the company’s order book rose by 23% to $8.5 billion at the end of June from a year ago.
The company has said that its capacity for producing major power equipment is already fully booked. The company said it now has an order backlog of over three years and a large portion of its production capacity has been booked for future orders of major power equipment in the next three years.
It also said that it was making its “order discussions” with some important customers as far as 2030.
A similar boom has taken place in Jinpan. Its new data centre orders for the first half of the year grew over 4 times compared to the same period last year and its backlog of related orders rose nearly threefold. These numbers illustrate the scope and scale of demand for AI infrastructure that is driving growth beyond chip manufacturers and tech firms.
Changes in the power demand are also affecting the demands on data centre equipment. AI servers use much more power than traditional computers, and having dense arrays of high-performance processors in a single rack is driving demand to increase quickly. Based on Nvidia’s technology roadmap, Bank of America predicts power usage for an AI rack may exceed 1.5 megawatts by the end of 2030. It would be a significant upgrade over standard server racks.
Operators are also facing increased pressure to improve their efficiency as they are faced with the challenge of meeting the growing energy requirement. Data centres use a significant amount of electricity and water, and are now being heavily investigated by governments, communities and the environment. The developers are therefore required to strike a balance between the desire for speedy expansion and worry about resource consumption, local infrastructure, etc.
The need to cool has become increasingly critical with the advent of increased heat generation in computing devices. For some high density AI applications, traditional air-cooling systems are becoming increasingly appropriate, prompting companies to innovate in heat removal from servers. For high-performance computing systems, for instance, liquid cooling can offer more efficient thermal management, and for high-performance AI deployments, it is increasingly being considered.
Other technologies that improve the efficiency of power distribution are being considered by electrical equipment manufacturers. Solid-state transformer is one of the technologies that is gaining more interest. Solid state transformers act as a semiconductor device to control and transform electrical power, unlike conventional transformer technology, which is heavily dependent on magnetic components and copper windings.
These technologies are resonating with the changes the industry is going through in the data centre. Infrastructure is not just a backstop function of the business. Availability of power, cooling capacity and electrical efficiency are increasingly becoming issues that are important to consider and may make the difference in the speed at which a facility can be constructed and the efficiency in which it can run.
The offer, however, is accompanied by risks. Demand for transformers, cooling equipment and other specialty components may serve as an enticement for new entrants and expand the capacity of existing manufacturers. Eventually, if supply increases, pricing power may start to decrease. Firms can also incur increased raw materials, component and manufacturing expenses.
