Google CEO Warns of Compute Shortage: Annual Compute Rental Market Expected to Exceed 260 Billion Yuan, Earnings of Multiple Stocks Poised to Double
Google CEO Pichai said customers are seeking compute for 2027, and it is "extremely difficult" for Google to meet demand. In the first quarter of 2026, the domestic compute rental market reached 68 billion yuan, up 62% year on year, and is expected to exceed 260 billion yuan for the full year; the one-year rental price of H100 rose from $1.70/hour in October 2025 to $2.35/hour in March 2026. Institutions forecast that the earnings of 56 compute concept stocks are expected to double this year.
Body · AI translation
Article body
Google CEO warns of computing power shortage! Computing power leasing is booming, and many stocks are expected to see their performance double (list)
Google CEO Pichai said that customers are seeking computing power for 2027, and it is "extremely difficult" for Google to meet demand. Institutions expect the full-year market size of computing power leasing to exceed 260 billion yuan.
Google CEO: New computing power demand in 2027 will be "extremely difficult to meet"
In an interview on the Big Technology podcast released on October 20 local time, Alphabet CEO Sundar Pichai revealed that if customers want to obtain computing power for 2027 now, it is "extremely difficult" for Google to meet such demand, and companies across the industry all wish the computing power scale they had originally planned could be twice as large. He also admitted that Google even briefly leased external computing power last year to meet the needs of some customers, while the cloud business is giving up some short-term revenue at any point in time to safeguard computing power investment for frontier AI research and development.
The overseas computing power shortage is not an isolated case. Amazon CEO Andy Jassy previously said that the company's computing power resources being in short supply will continue into 2027, and even orders for 2028 have already piled up in large numbers. Financial report data show that the capital expenditure guidance for the four major North American cloud vendors in 2026 totals approximately 720 billion to 745 billion US dollars, of which Google's full-year capital expenditure has been raised to 195 billion to 205 billion US dollars, Amazon's is about 220 billion US dollars, and Meta's is about 130 billion to 145 billion US dollars.
Domestic computing power supply is also accelerating its expansion. Information released at the 2026 China Computing Conference shows that as of June 2026, the number of standard racks in use nationwide reached 15.56 million, the scale of intelligent computing power in use reached 2185 EFLOPS (FP16), a year-on-year increase of 177%, and storage capacity exceeded 2000 EB; the conference also held the launch ceremony for the national integrated computing power monitoring of the China Computing Platform, and the development pattern of "one network, one chessboard, and integration" for China's computing power has basically taken shape.
The computing power supply-demand gap continues to widen
Computing power leasing refers to service providers renting server computing power resources equipped with high-performance chips such as GPUs to enterprises or institutions with AI training and inference needs through cloud-based, pay-as-you-go, and elastic leasing models. It is a key bridge connecting upstream computing power hardware and downstream AI applications. At present, the explosion of AI inference demand combined with limited supply of high-end chips has continued to widen the domestic computing power supply-demand gap, and the center of computing power leasing prices has continued to move upward.
English text is being prepared.
"The supply shortage of computing power in the second half of the year is expected to remain unresolved effectively, and leasing prices will remain high." Shanxi Securities believes that the rapid penetration of AI Agents has driven explosive growth in token usage, while the supply constraints on Nvidia's high-end GPUs have not eased and domestic computing power chips still need time to scale up. The supply-demand gap is expected to further widen, and rental prices for high-end cards such as H200 and B300 will remain high.
Against the backdrop of continued demand expansion, Founder Securities further pointed out that as of the end of June 2026, China's intelligent computing power scale reached 2185 EFLOPS, a year-on-year increase of 177%, and estimates that the domestic computing power leasing market size is expected to approach one trillion yuan by 2030.
This judgment echoes the view of Northeast Securities, which believes that with the continued iteration of domestic computing power software and hardware, the supply and demand of domestic computing power chips is expected to achieve overall balance in 2028, and Ascend is expected to capture 50% of the domestic chip market share in 2026.
Many concept stocks are expected to double their performance this year
East Money concept sectors show that there are currently 249 stocks in the A-share market involved in the computing power concept, with a total market value of approximately 10.38 trillion yuan.
China Mobile ranks first with a size of 2095.606 billion yuan, Industrial Fulian ranks second with 986.251 billion yuan, and Zhongji Innolight follows closely with 761.837 billion yuan. China Telecom, Cambricon, Hygon Information, and Eoptolink all have market values exceeding 400 billion yuan.
Over the past month, the computing power concept sector has generally trended weakly, with 36 stocks rising and 213 falling. Nanwei Software surged 34% to rank first, Decai Shares rose 29% to rank second, Haitian Ruisheng rose 27%, and Bohui Shares, Huace Film & TV, and Huichen Shares all rose more than 19%. On the decline side, Eoptolink, Weice Technology, and Sino-New Energy and many other stocks fell more than 31%.
From a capital perspective, East Money Choice data show that over the past month, a total of 30 computing power concept stocks received contrarian bets from margin traders. Among them, Xingyun Technology received 84 million yuan in additional leverage funds, Leo Planning and Shensangda A attracted 59 million yuan and 51 million yuan respectively, and Dongyangguang, Runjian Shares, and Alt all received more than 34 million yuan in net margin purchases.
In terms of future growth potential, according to consensus institutional forecasts, a total of 56 computing power concept stocks are expected to see their performance double this year. Among them, institutions expect Yushun Electronics' net profit to surge 31.7 times year-on-year, Kaile Shares' net profit to grow more than 28.7 times, Hongjing Technology and China Science Xingtu are both forecast to see net profit growth of more than 14 times, and Jiechuang Intelligent and Runjian Shares are also expected to double their performance.
Zhongji Innolight achieved operating revenue of 41.778 billion yuan and net profit attributable to the parent company of 13.651 billion yuan in the first half of the year, becoming the first in the industry to ship 1.6T DR8 OSFP224 products; 1.6T optical modules have been shipped in batches and continue to ramp up. According to company disclosures, orders from almost all customers currently cover the full year of 2026, with some customer orders scheduled into 2027, and leading overseas cloud vendors have provided quarterly shipment procurement guidance for 2027; on the capacity side, phases I and II of the Thailand base are at full production, and bases in Tongling and Suzhou are expanding production simultaneously.
Cambricon achieved operating revenue of 5.996 billion yuan in the first half of the year, a year-on-year increase of 108.13%, and net profit attributable to the parent company of 2.311 billion yuan, a year-on-year increase of 122.61%, of which cloud product line revenue was 5.994 billion yuan, accounting for 99.98% of revenue. The company's new-generation flagship cloud AI chip, Siyuan 690, was mass-produced in early 2026, with FP16 computing power exceeding 700 TFLOPS, and has achieved Day 0 adaptation for mainstream domestic models such as DeepSeek, Qwen, and Kimi. The product is accelerating large-scale deployment in fields such as finance and the internet.
(Article source: East Money Research Center) Mid-body embedded advertisement