① This investment will be used for chips and servers, while supporting the development of AI services such as the Gemini model; ② Alphabet CEO is optimistic that the opportunities in AI are vast; ③ Global technology giants are investing heavily in building AI infrastructure.
According to the Star Daily on April 10 (Editor Song Ziqiao), although returns remain unclear and global trade conflicts may increase costs, Google's parent company Alphabet has reiterated that it will invest approximately 75 billion USD this year to build Datacenter capacity, doubling down on generative AI.
On April 9 local time, Alphabet CEO Sundar Pichai stated at the annual meeting of Google Cloud Computing that this investment will be used to improve the chips and servers needed for its core products, including search, while supporting the development of AI services such as the Gemini model.
He said, “The opportunities in AI are vast.” He added that these investments will also benefit the company's enterprise clients.
In February of this year, Alphabet outlined its spending plan of 75 billion USD. This target is 29% higher than the expectations of Wall Street analysts at the time.
$Alphabet-C (GOOG.US)$ Demis Hassabis, the CEO of the AI laboratory DeepMind, gave an optimistic forecast in March this year: general artificial intelligence (AGI) is expected to be achieved within 5 to 10 years, and it will possess cognitive abilities equivalent to humans. "I think society needs to be prepared for this and the impact it will have. Moreover, you know, ensuring that we benefit from it, that society as a whole benefits, but we should also mitigate some risks."
This AI leader also stated that Deepseek's AI model "may be the best work in China."
Since OpenAI launched ChatGPT in 2022, investment in the AI field has surged, and global technology giants are heavily investing in building AI infrastructure.
Domestically, Internet giants and the three major operators continue to actively increase their capital expenditures for computing power. Both Alibaba and Tencent have clearly focused on the AI sector, with both expected to exceed 100 billion in capital expenditure by 2025.
$Alibaba (BABA.US)$The announcement states that at least 380 billion yuan (53 billion USD) will be invested over the next three years to build Cloud Computing and AI infrastructure. If averaged over the years, capital expenditures in 2025 are expected to exceed 120 billion yuan.
Tencent President Liu Chiping stated at a performance communication meeting that the main reason for the significant increase in capital expenditure in Q4 2024 is the purchase of more GPUs to meet inference needs, with plans to further increase capital expenditures in 2025, expected to account for a low double-digit percentage of revenue. Tencent's revenue growth rate for 2024 is approximately 8%. Assuming the revenue growth rate for 2025 remains at 8%, and calculating with a median of 15% for the low double-digit percentage (usually between 10%-20%), the capital expenditure for 2025 is expected to be around 107 billion yuan.
However, under the overall decline in capital expenditure plans for the three major Operators in 2025, there is an increase in investments in computing power, including$CHINA TELECOM (00728.HK)$a preliminary plan to invest 21.96 billion in the fields of Cloud Computing and computing power.$CHINA UNICOM (00762.HK)$A planned investment of 18 billion, representing a growth of 22% and 28% compared to the same period last year, while$CHINA MOBILE (00941.HK)$The computing power Capex plan, although only increased by 0.5% to 37.3 billion, mainly focuses on pre-training computing power, while the resources for inference will be invested according to market demand without any upper limit.
Abroad,$Microsoft (MSFT.US)$Noel Walsh, President of Cloud Computing and Innovation, stated in a LinkedIn post on Tuesday: “Microsoft will invest more than 80 billion USD (approximately 588.742 billion yuan) this year to continue building the datacenter infrastructure, with these investments made based on short-term and long-term demand.”
$Meta Platforms (META.US)$ The platform indicates that it plans to invest up to 65 billion USD (approximately 478.353 billion RMB) this year to expand its AI infrastructure.
Orient believes that although there have been few significant catalysts in the computing power sector recently, the iteration speed of AI models remains rapid, and the applications of multimodal and Agent will consume more computing power. The increase in computing power consumption will continue to enhance the prosperity of the computing power industry. In addition, domestic policies promoting controllable computing power and requirements for data security will draw more attention and procurement to domestically produced AI Chips and integrated machines.
Wanlian Securities also stated that, based on the latest capital expenditure plans disclosed by major domestic cloud companies and the three major operators, domestic demand for intelligent computing will continue to rise. Moreover, as the demand for computing power shifts from training to inference, it will further accelerate the application scenarios of large AI models. It is suggested to pay attention to:
1) The increased investment in domestic capital expenditure will boost the demand in the computing power Industry Chain, including the construction of intelligent computing centers, servers, optical modules, optical chips, etc.;
2) The accelerated landing of generative AI applications will further boost the demand for intelligent computing in inference;
3) The accelerated landing of AI applications on the application side.
Editor/rice
