AI data center, at the forefront
Artificial Intelligence (AI) has become mainstream. With the launch of solutions like OpenAI's ChatGPT, which won hundreds of millions of users overnight, AI models are no longer working "quietly" in the background. Instead, they have taken center stage.
01
The Next Big Thing
Data centers are physical facilities used for transmitting, accelerating, displaying, computing, and storing data information over networks, primarily used by organizations with significant demands for data computation and storage. A complete data center consists of data center IT equipment and data center infrastructure.
Data center IT equipment mainly includes connectors (optical fibers, optical modules), network devices (switches, routers), computing devices (servers), and storage devices (memory devices). Data center infrastructure is a collective term for various systems that support the normal operation of data centers, mainly including power distribution and control equipment (UPS, batteries, diesel generators, distribution units), and temperature control equipment (cooling source equipment, computer room air conditioning, fresh air systems), etc.
Advertisement
Fundamentally, artificial intelligence and machine learning algorithms are very adept at discovering patterns in data sets. Then, they apply the knowledge they have learned to future tasks—automating and simplifying many routine operations, which is known as predictive analysis.
In recent years, data center operators have been adopting artificial intelligence to help streamline the daily operation of services. In a recent survey, 57% of data center owners indicated that they would trust AI models to make operational decisions, a figure that has increased by nearly 20% from the previous year.
From the perspective of the number of large data centers operated by hyperscale operators, the overall number of global hyperscale data centers has been growing as the industry concentration gradually increases. According to the latest data from Synergy Research Group, as of 2021, the total number of large data centers operated by hyperscale providers increased to around 700, a year-on-year increase of 17.25%. Based on the latest forecast from Synergy Research Group, with the current planning of 314 future new hyperscale data centers known, the installed base of operational data centers is expected to break through the 1000 mark within three years and continue to grow rapidly thereafter.Specifically, operators can use predictive analytics to improve data center cooling systems and other areas in real time. Providing the power and storage needed for modern computing demands generates an incredible amount of heat. By using artificial intelligence to cool hardware more efficiently, providers can reduce costs and improve energy efficiency. For instance, Google's AI implementation has reduced its cooling costs by 40%.
Artificial intelligence can also help reduce the inefficiencies of IT infrastructure. Predictive analytics can assist providers in fine-tuning power distribution and rack space. The result is reduced operational costs, improved power usage effectiveness (PUE), and smarter data-driven decisions.
Modern companies run extremely demanding workloads on data center infrastructures. Take ChatGPT as an example; a recent report predicts that by 2028, the cost of training and implementing generative AI models will reach $76 billion. This figure is more than double the annual cost of the world's largest public cloud provider, Amazon Web Services (AWS).
What does this mean for data centers? To provide the power, storage, and connectivity required for new technologies and current technologies, they must become more efficient in many areas. As a result, some data center service providers have already started taking action.
Artificial intelligence has already had a significant impact on our world, but it's important that it keeps driving innovation forward. Future data centers must evolve with the development of technology to provide efficient and effective services. For example, we may see advanced artificial intelligence, quantum computing, and other emerging technologies shaping the next generation of data centers. These cutting-edge technologies are expected to offer higher efficiency and advanced functionalities.
However, as artificial intelligence becomes an integral part of data center operations, issues of transparency and accountability will naturally come into play. Sustainability will also play a huge role in data center AI decision-making. Workloads with resource constraints utilize advanced CPUs and GPUs, requiring advanced liquid cooling systems to prevent hardware damage. Therefore, data center energy consumption is projected to increase by 12% by 2030.
Utilizing artificial intelligence will ensure that data center providers evolve with these changing demands, providing the future hyperscale digital backbone in a thoughtful and ethical manner.
02
Staying One Step Ahead
In some data center markets, such as Northern Virginia and Phoenix in the United States, the demand for data center space is extremely strong, and data center service providers and hyperscale cloud vendors have already purchased large amounts of land in anticipation of potential demand surges.Although we cannot attribute 100% of the current data center boom to artificial intelligence, it is reasonable to assume that the growth of AI is part of the reason. With hyperscale enterprises in a race to accelerate the construction of AI infrastructure (investment in infrastructure supports this view), construction teams may be busy for a considerable time to come. Will AI drive the demand for the data center industry like public cloud? Let's first look at the movements of the market's front-line manufacturers.
Relevant data shows that there are currently about 8,000 data centers worldwide, mainly distributed in the United States, Asia, and Europe. Northern Virginia in the United States is the world's largest data center hub, with about 300 data centers. From an energy consumption perspective, in 2023, data centers in Northern Virginia consumed 2,552MW of electricity, Dallas consumed 654MW, Silicon Valley 615MW, Beijing data centers consumed 1,799MW, London 1,052MW, and Frankfurt 864MW.
By 2025, the total capital expenditure of Alphabet, Amazon, Meta, and Microsoft is expected to reach $200 billion, significantly surpassing the capital expenditure of several major oil companies.
In 2023, Amazon (AWS) has been devising new and improved strategies to make its data centers and hyperscale facilities more sustainable. Entering 2024, businesses around the world will continue to use AWS to reduce operational costs and become more digitally agile, thereby accelerating the pace of innovation.
Amazon has invested $7.8 billion in building data centers in central Ohio. Moreover, with the popularity of remote work, a substantial amount of funding has also flowed into data centers, communication infrastructure, fiber optics, cell towers, and related technology industries.
In January of this year, Amazon announced a partnership with Capgemini to further focus on the adoption of enterprise-generated artificial intelligence (Gen AI). Reports suggest that increasing the adoption of AI will allow companies to gain a deeper understanding of their environmental strategies, thereby paying more attention to more sustainable areas.
With its optimized large language model (LLM), AWS hopes to continue efforts to help businesses fully leverage the potential of Gen AI while helping customers address digital challenges such as cost and scale. Within data centers, if fully leveraged, Gen AI and the cloud will continue to revolutionize.
Microsoft also plans to expand the scale of its data centers in London and Cardiff and expand into the northern region of England. This infrastructure investment will help meet the growing demand for efficient, scalable, and sustainable AI-specific computing capabilities, as well as the needs of the private and public sectors waiting to take advantage of the latest cloud computing and AI breakthroughs.
In addition to infrastructure investment, Microsoft will also allocate millions of pounds for training personnel to acquire the skills needed to build and use artificial intelligence. Furthermore, Microsoft will introduce more than 20,000 of the most advanced graphics processors to the UK, which are key technologies required for machine learning and AI development. This investment also includes a training program to help ensure that Britons have the skills needed to build and use AI.
On January 25, the Indiana Economic Development Corporation announced that Meta plans to build an $800 million AI-focused data center campus in Indiana. Meta will build a nearly 700,000 square foot facility in Jeffersonville, Indiana, which will begin construction this month and is expected to be operational by 2026. In addition to 100 operational jobs, the company also expects to support more than 1,250 jobs at the peak of construction.This investment signifies that Meta's new data center in Indiana will focus on the research and development and application of artificial intelligence technology. This move aims to enhance Meta Platforms' competitiveness in the field of artificial intelligence and further consolidate its position in the global technology market. It will also have a positive impact on local economic development, enhance the cultivation of local technology talent and innovation capabilities, and inject new vitality into the development of artificial intelligence industry in Indiana and globally.
On March 13th, Meta announced through an official press release the construction of two new data center clusters, hoping to stand out in AI-focused development with NVIDIA's GPUs. It is reported that the sole purpose of these two data centers is to conduct AI research and develop large language models in consumer-specific application fields. Each cluster includes 24,576 NVIDIA H100 AI GPUs, which will be used to train their own large language model, Llama 3. Meta officials stated that the high-performance network structure of these clusters, along with key storage decisions and the H100 GPUs in each cluster, can support larger and more complex models, paving the way for the development of general artificial intelligence products and the advancement of AI research.
Alphabet's Google will invest $1 billion to build a new data center in Waltham Cross, Hertfordshire, UK. The company stated at a press conference that the 33-acre site will create construction and technology jobs for the local community.
Debbie Weinstein, Google's Vice President and Managing Director for the UK and Ireland, said: "Once completed, this investment will bring critical computing power to businesses across the UK, support artificial intelligence innovation, and help ensure reliable digital services for Google Cloud customers and Google users in the UK and abroad."
They added that this investment is based on three foundations: the Central Saint Giles office acquired for $1 billion in 2022, a 1 million square foot development project in King's Cross, and the launch of an accessibility exploration center. Google pointed out that it intends to power all data centers and campuses with carbon-free energy by 2030. In 2022, Google announced a power purchase agreement with Engie to buy offshore wind energy generated by the Moray West wind farm in Scotland.
Google stated that the agreement will add 100 MW of energy to the grid and make Google's operations in the UK有望在2025年达到或接近90%的无碳能源。Furthermore, Google said it is exploring ways to utilize the heat generated by data centers, and the new facility will also provide regulations for off-site heat recovery.
Artificial Intelligence (AI) has become mainstream. With the launch of solutions like OpenAI's ChatGPT, which won hundreds of millions of users overnight, AI models are no longer working "quietly" in the background. Instead, they have taken center stage.
01
The Next Big Thing
Data centers are physical facilities used for transmitting, accelerating, displaying, computing, and storing data information over networks, primarily used by organizations with significant demands for data computation and storage. A complete data center consists of data center IT equipment and data center infrastructure.
Data center IT equipment mainly includes connectors (optical fibers, optical modules), network devices (switches, routers), computing devices (servers), and storage devices (memory devices). Data center infrastructure is a collective term for various systems that support the normal operation of data centers, mainly including power distribution and control equipment (UPS, batteries, diesel generators, distribution units), and temperature control equipment (cooling source equipment, computer room air conditioning, fresh air systems), etc.
Advertisement
Fundamentally, artificial intelligence and machine learning algorithms are very adept at discovering patterns in data sets. Then, they apply the knowledge they have learned to future tasks—automating and simplifying many routine operations, which is known as predictive analysis.
In recent years, data center operators have been adopting artificial intelligence to help streamline the daily operation of services. In a recent survey, 57% of data center owners indicated that they would trust AI models to make operational decisions, a figure that has increased by nearly 20% from the previous year.
From the perspective of the number of large data centers operated by hyperscale operators, the overall number of global hyperscale data centers has been growing as the industry concentration gradually increases. According to the latest data from Synergy Research Group, as of 2021, the total number of large data centers operated by hyperscale providers increased to around 700, a year-on-year increase of 17.25%. Based on the latest forecast from Synergy Research Group, with the current planning of 314 future new hyperscale data centers known, the installed base of operational data centers is expected to break through the 1000 mark within three years and continue to grow rapidly thereafter.Specifically, operators can use predictive analytics to improve data center cooling systems and other areas in real time. Providing the power and storage needed for modern computing demands generates an incredible amount of heat. By using artificial intelligence to cool hardware more efficiently, providers can reduce costs and improve energy efficiency. For instance, Google's AI implementation has reduced its cooling costs by 40%.
Artificial intelligence can also help reduce the inefficiencies of IT infrastructure. Predictive analytics can assist providers in fine-tuning power distribution and rack space. The result is reduced operational costs, improved power usage effectiveness (PUE), and smarter data-driven decisions.
Modern companies run extremely demanding workloads on data center infrastructures. Take ChatGPT as an example; a recent report predicts that by 2028, the cost of training and implementing generative AI models will reach $76 billion. This figure is more than double the annual cost of the world's largest public cloud provider, Amazon Web Services (AWS).
What does this mean for data centers? To provide the power, storage, and connectivity required for new technologies and current technologies, they must become more efficient in many areas. As a result, some data center service providers have already started taking action.
Artificial intelligence has already had a significant impact on our world, but it's important that it keeps driving innovation forward. Future data centers must evolve with the development of technology to provide efficient and effective services. For example, we may see advanced artificial intelligence, quantum computing, and other emerging technologies shaping the next generation of data centers. These cutting-edge technologies are expected to offer higher efficiency and advanced functionalities.
However, as artificial intelligence becomes an integral part of data center operations, issues of transparency and accountability will naturally come into play. Sustainability will also play a huge role in data center AI decision-making. Workloads with resource constraints utilize advanced CPUs and GPUs, requiring advanced liquid cooling systems to prevent hardware damage. Therefore, data center energy consumption is projected to increase by 12% by 2030.
Utilizing artificial intelligence will ensure that data center providers evolve with these changing demands, providing the future hyperscale digital backbone in a thoughtful and ethical manner.
02
Staying One Step Ahead
In some data center markets, such as Northern Virginia and Phoenix in the United States, the demand for data center space is extremely strong, and data center service providers and hyperscale cloud vendors have already purchased large amounts of land in anticipation of potential demand surges.Although we cannot attribute 100% of the current data center boom to artificial intelligence, it is reasonable to assume that the growth of AI is part of the reason. With hyperscale enterprises in a race to accelerate the construction of AI infrastructure (investment in infrastructure supports this view), construction teams may be busy for a considerable time to come. Will AI drive the demand for the data center industry like public cloud? Let's first look at the movements of the market's front-line manufacturers.
Relevant data shows that there are currently about 8,000 data centers worldwide, mainly distributed in the United States, Asia, and Europe. Northern Virginia in the United States is the world's largest data center hub, with about 300 data centers. From an energy consumption perspective, in 2023, data centers in Northern Virginia consumed 2,552MW of electricity, Dallas consumed 654MW, Silicon Valley 615MW, Beijing data centers consumed 1,799MW, London 1,052MW, and Frankfurt 864MW.
By 2025, the total capital expenditure of Alphabet, Amazon, Meta, and Microsoft is expected to reach $200 billion, significantly surpassing the capital expenditure of several major oil companies.
In 2023, Amazon (AWS) has been devising new and improved strategies to make its data centers and hyperscale facilities more sustainable. Entering 2024, businesses around the world will continue to use AWS to reduce operational costs and become more digitally agile, thereby accelerating the pace of innovation.
Amazon has invested $7.8 billion in building data centers in central Ohio. Moreover, with the popularity of remote work, a substantial amount of funding has also flowed into data centers, communication infrastructure, fiber optics, cell towers, and related technology industries.
In January of this year, Amazon announced a partnership with Capgemini to further focus on the adoption of enterprise-generated artificial intelligence (Gen AI). Reports suggest that increasing the adoption of AI will allow companies to gain a deeper understanding of their environmental strategies, thereby paying more attention to more sustainable areas.
With its optimized large language model (LLM), AWS hopes to continue efforts to help businesses fully leverage the potential of Gen AI while helping customers address digital challenges such as cost and scale. Within data centers, if fully leveraged, Gen AI and the cloud will continue to revolutionize.
Microsoft also plans to expand the scale of its data centers in London and Cardiff and expand into the northern region of England. This infrastructure investment will help meet the growing demand for efficient, scalable, and sustainable AI-specific computing capabilities, as well as the needs of the private and public sectors waiting to take advantage of the latest cloud computing and AI breakthroughs.
In addition to infrastructure investment, Microsoft will also allocate millions of pounds for training personnel to acquire the skills needed to build and use artificial intelligence. Furthermore, Microsoft will introduce more than 20,000 of the most advanced graphics processors to the UK, which are key technologies required for machine learning and AI development. This investment also includes a training program to help ensure that Britons have the skills needed to build and use AI.
On January 25, the Indiana Economic Development Corporation announced that Meta plans to build an $800 million AI-focused data center campus in Indiana. Meta will build a nearly 700,000 square foot facility in Jeffersonville, Indiana, which will begin construction this month and is expected to be operational by 2026. In addition to 100 operational jobs, the company also expects to support more than 1,250 jobs at the peak of construction.This investment signifies that Meta's new data center in Indiana will focus on the research and development and application of artificial intelligence technology. This move aims to enhance Meta Platforms' competitiveness in the field of artificial intelligence and further consolidate its position in the global technology market. It will also have a positive impact on local economic development, enhance the cultivation of local technology talent and innovation capabilities, and inject new vitality into the development of artificial intelligence industry in Indiana and globally.
On March 13th, Meta announced through an official press release the construction of two new data center clusters, hoping to stand out in AI-focused development with NVIDIA's GPUs. It is reported that the sole purpose of these two data centers is to conduct AI research and develop large language models in consumer-specific application fields. Each cluster includes 24,576 NVIDIA H100 AI GPUs, which will be used to train their own large language model, Llama 3. Meta officials stated that the high-performance network structure of these clusters, along with key storage decisions and the H100 GPUs in each cluster, can support larger and more complex models, paving the way for the development of general artificial intelligence products and the advancement of AI research.
Alphabet's Google will invest $1 billion to build a new data center in Waltham Cross, Hertfordshire, UK. The company stated at a press conference that the 33-acre site will create construction and technology jobs for the local community.
Debbie Weinstein, Google's Vice President and Managing Director for the UK and Ireland, said: "Once completed, this investment will bring critical computing power to businesses across the UK, support artificial intelligence innovation, and help ensure reliable digital services for Google Cloud customers and Google users in the UK and abroad."
They added that this investment is based on three foundations: the Central Saint Giles office acquired for $1 billion in 2022, a 1 million square foot development project in King's Cross, and the launch of an accessibility exploration center. Google pointed out that it intends to power all data centers and campuses with carbon-free energy by 2030. In 2022, Google announced a power purchase agreement with Engie to buy offshore wind energy generated by the Moray West wind farm in Scotland.
Google stated that the agreement will add 100 MW of energy to the grid and make Google's operations in the UK有望在2025年达到或接近90%的无碳能源。Furthermore, Google said it is exploring ways to utilize the heat generated by data centers, and the new facility will also provide regulations for off-site heat recovery.