The explosion of AI big models is sparking a new wave of data center expansion.
According to the Wall Street Journal on June 30th, Amazon plans to invest over $100 billion in building data centers over the next 10 years. Nowadays, this e-commerce giant has invested more in data center infrastructure than in retail warehouses.
According to market research firm Dell'Oro Group, Amazon's expenditure on data center capital expenditures (including leasing) accounted for more than half (53%) of total capital expenditures last year, reaching a new high in nearly a decade.
John Felton, Chief Financial Officer of Amazon Cloud Services (AWS), has stated that building data centers to support the development of artificial intelligence is like building a massive logistics network in the past few years. Previously, the company had launched corresponding data center investment plans for New Zealand, India, Singapore, Japan, and other regions, with investment amounts ranging from $2.8 billion to $9 billion.
As the fastest-growing core business under the company, AWS has been Amazon's "growth engine" in recent years, leading the rapid development of the cloud computing industry.
However, starting from 2022, the global economy has entered a downward cycle due to the US dollar interest rate hike, causing companies to tighten their IT service spending and reduce their investment in cloud services. The growth of the cloud computing industry has significantly slowed down, dropping below 30% to 28% for the first time in 2022 and further to 16% in 2023.
Amazon has also taken a series of measures to reduce costs and increase efficiency, such as layoffs, hiring freezes, cutting off branch businesses, and reducing unnecessary capital expenditures. In 2022, Amazon's overall capital expenditure was $59 billion and it stated that it will reduce its capital expenditure in 2023.
Nevertheless, Amazon has not reduced its investment in cloud computing. Especially last year, the explosion of ChatGPT brought new opportunities to the slowing growth of the cloud computing industry. Due to the fact that the training and inference of AI models cannot be separated from cloud computing services, major cloud vendors are striving to improve their AI service capabilities to attract and retain customers, further intensifying competition in the cloud computing market.
According to Amazon's plan, last year, while reducing overall capital expenditures, a portion of the expenses saved from e-commerce and logistics businesses were transferred to cloud businesses. At the same time, capital expenditures for generative artificial intelligence and big language model related businesses were increasing.
As an infrastructure for cloud computing services, Amazon is increasing investment in building data centers.
In the United States, data center bases in Virginia and Oregon are the main strongholds of Amazon's "cloud computing empire", and AWS invests $4 out of every $5 in infrastructure expenditures in the United States from both regions. Last year, in addition to expanding data center facilities in Virginia and Oregon, the company also planned to expand into other states in the United States and more places in Europe, Saudi Arabia, Malaysia, and more.
Not only Amazon, but also Google, Microsoft, and more cloud giants are expanding their data center presence. In April this year, Google announced that it would invest $3 billion to establish a new data center campus in Indiana and expand existing facilities in Virginia. Microsoft has also been exposed by the media as planning to invest $110 billion to jointly customize a large-scale data center project with OpenAI. This project includes a super artificial intelligence computer representing the Stargate internally, with millions of built-in servers, and is expected to be officially launched in 2028.
According to NVIDIA CEO Huang Renxun, as the demand for computing in AI continues to grow, data centers will further become "AI factories" in this wave of technology. These factories operate intensively around the clock, constantly processing massive and continuous data to train and improve AI models. At present, cloud giants such as Amazon and Microsoft are striving to become the owners of "AI factories".
The domestic market is also vigorously promoting the construction of data centers.
This year's government work report pointed out that the construction of digital infrastructure should be moderately advanced. A domestic data center manufacturer told Interface News reporters that due to the outbreak of the AI big model last year, the domestic computing power continued to be in short supply. Currently, it has become a consensus to build more computing infrastructure, and the new data center mainly based on AI intelligent computing centers is one of the representatives, which is being implemented and constructed on a large scale nationwide.
With numerous technology companies expanding the construction scale of data centers, the energy consumption problem is becoming increasingly significant. How to balance the rapid growth of data centers with energy conservation and emission reduction has become a major challenge. At the same time, specific areas may face power rationing issues due to power supply capacity limitations, affecting the stable operation of data centers.
Moreover, if there is a phenomenon of oversupply in the data center market, it is likely to trigger a price war for cloud services, thereby affecting the healthy development of the industry. How to maintain a reasonable and orderly expansion while competing for the data center market will be a topic that technology giants need to jointly address.