After years on the sidelines, VNET seizes on China’s AI moment with new CATL tie-up - FT中文网
登录×
电子邮件/用户名
密码
记住我
请输入邮箱和密码进行绑定操作:
请输入手机号码,通过短信验证(目前仅支持中国大陆地区的手机号)。如果该手机号尚未注册,将自动创建 FT中文网账号。短信可能需要几分钟送达,验证码15分钟内有效,请耐心等待:
请您阅读我们的用户注册协议和隐私权保护政策,点击下方按钮即视为您接受。
咏竹坊

After years on the sidelines, VNET seizes on China’s AI moment with new CATL tie-up

China’s oldest independent data center operator is finding its footing in the AI era, helped by a new $1 billion investment from battery champion CATL
00:00

{"text":[[{"start":8.85,"text":"This article only represents the author's own views."}],[{"start":12.149999999999999,"text":"When Chen Sheng, who also goes by Josh, founded VNET Group Inc. (VNET.US) in 1996, China had yet to build its first independent, privately owned data center. Three decades later, including a spotty track record filled with ups-and-downs, VNET thinks it has finally found a formula for success by leaning into an AI boom that has made data center capacity a key asset in the U.S.-China technology race."}],[{"start":39.8,"text":"That new reality was on display last week, when a consortium linked to leading battery maker Contemporary Amperex Technology Ltd. (CATL)(300750.SZ; 3750.HK) agreed to buy about 38% of VNET’s shares for nearly $1 billion. VNET’s shares surged roughly 25% on the announcement, reflecting not only a big vote of confidence in its prospects, but also CATL’s belief that the stock was highly undervalued at the time."}],[{"start":64.25,"text":"From pioneer to AI contender"}],[{"start":67.4,"text":"VNET built its early franchise by offering hosting services outside China’s big three state-run telecoms carriers that dominated the data center business at that time. Today it runs more than 50 data centers across over 30 Chinese cities, with 889 MW of wholesale capacity and more than 49,000 retail cabinets. Its customers include internet and cloud service giants like ByteDance, Alibaba and Tencent, whose AI ambitions are driving their demand for data center capacity."}],[{"start":98.35000000000001,"text":"But before the recent AI boom, VNET spent much of the past decade falling behind its faster-growing and savvier rivals. Heavy debt, governance concerns and repeated failed privatization efforts battered investor confidence in the company, leaving its stock undervalued compared with domestic peers GDS (GDS.US; 9698.HK) and Chindata."}],[{"start":121.85000000000001,"text":"VNET’s comeback rests on a pivot to wholesale facilities custom designed for AI workloads. In 2025, it delivered a record 404 MW of new capacity, taking its wholesale capacity in service to 889 MW, with another 452 MW under construction. New wholesale orders totaled 135 MW, lifting the commitment rate for in-service capacity to 95%, while its utilization rate reached 70.1%."}],[{"start":153.70000000000002,"text":"With guidance to deliver another 450 MW to 500 MW in 2026 and projected capex of 10 billion yuan ($1.4 billion) to 12 billion yuan this year, VNET’s AI-centric wholesale business has rapidly become its main growth engine."}],[{"start":172.60000000000002,"text":"CATL’s leap beyond batteries"}],[{"start":176.10000000000002,"text":"Then there’s CATL, which has suddenly found a new AI dance partner in VNET. CATL built its dominance in EV batteries and energy storage, but intense competition is now pushing it to look beyond its core areas. This spring it invested in a parent of high-voltage direct-current supplier Hangzhou Zhongheng Electric and signed a large sodium-ion battery contract, before its latest investment in VNET."}],[{"start":203.20000000000002,"text":"The strategic logic is straightforward: pair CATL’s energy storage technology with data center power systems to build a “battery-to-power-to-compute” ecosystem. The model is premised on the reality that data centers are huge energy consumers, and that consumption is rapidly growing with the rise of power-hungry AI applications. Many data centers use onsite solar arrays to power their operations, which also requires energy storage facilities to keep the power flowing even when the sun isn’t shining."}],[{"start":232.20000000000002,"text":"CATL sold 541 GWh of EV batteries and 121 GWh of energy-storage batteries in 2025. But its shipments of batteries for energy storage systems rose only 29%, far slower than the market’s 79% expansion, as rivals including BYD (002594.SZ; 1211.HK) and EVE Energy (300014.SZ) scaled aggressively. The VNET deal gives CATL a captive customer for its energy storage system products, as well as a real-world laboratory to trial new technologies."}],[{"start":272.25,"text":"A transformative deal, with Chen still in control"}],[{"start":276.35,"text":"Under the transaction announced last Wednesday, two CATL-linked investment vehicles will buy up to 650.4 million VNET Class A shares for $1.4486 each, equivalent to $8.6914 per American depositary share (ADS). If fully executed, the purchase will give CATL about 38.1% of VNET’s outstanding shares."}],[{"start":301.90000000000003,"text":"But CATL is far from taking control of VNET, since a concurrent agreement requires it to vote its shares according to Chen’s instructions. An investor-rights agreement also restricts transfers for a period. In plain English: CATL gets a large strategic foothold, while Chen keeps his place calling the shots at his company."}],[{"start":321.40000000000003,"text":"That may disappoint investors hoping for a full governance reset, given the company’s laggard status in the past. But the market’s reaction suggests investors see more than a secondary share sale. With CATL’s vote of confidence, VNET’s AI push looks more credible, and tied to a partner that understands one of the bottlenecks that matters most in AI: power."}],[{"start":344.05,"text":"Financing an AI expansion"}],[{"start":347.1,"text":"The strategic backing arrives as VNET ramps up its investment in the notoriously capital- intensive business of building data centers. The company has promoted a “conveyor belt” financing model: it develops projects, then sells mature assets into private real estate investment trusts (REITs) to recover capital. In March 2026, it listed two REITs in Shanghai, raising 6.36 billion yuan. That funding is meant to support VNET’s 2026 delivery plan. Without asset recycling, VNET has suggested it would need to borrow more than 8 billion yuan a year to fund its pipeline."}],[{"start":383.40000000000003,"text":"And other risks remain as well. VNET’s wholesale business depends heavily on a handful of large customers, and any slowdown in their AI build-outs, or a shift toward self-built capacity, could pressure bookings. Delays in REIT sales could also strain the company’s balance sheet."}],[{"start":401.15000000000003,"text":"Still, the market backdrop is favorable. Frost & Sullivan expects China’s carrier-neutral data center service market to grow at an 18.5% annual rate from 2025 to 2030, supported by IT outsourcing, 5G, cloud services and AI."}],[{"start":420.3,"text":"Energy, AI and the global rivalry"}],[{"start":423.5,"text":"CATL’s interest in VNET underscores a broader shift: in the AI race, electricity is becoming as strategic as the GPUs that power actual AI computing. The International Energy Agency estimates U.S. data center electricity demand will more than double to 426 TWh by 2030, while China’s demand could more than double to about 277 TWh."}],[{"start":448.05,"text":"China’s edge is its ability to scale power infrastructure quickly. It already generates more than twice the electricity of the U.S. and can build transmission at a pace unrivaled in the West. Pairing VNET’s operational base with CATL’s storage and power-conversion technology dovetails with China’s goal of building a national network of high-capacity data centers to power its AI ambitions."}],[{"start":472.25,"text":"That matters as the U.S. and China remain locked in a technology standoff. The Trump-Xi summit in Beijing last week produced modest economic deals, but no breakthrough on advanced AI chip exports. Against that backdrop, VNET’s tie-up with CATL looks like more than a corporate transaction. It hints at how China will try to build up an infrastructure advantage as it concurrently plays catchup in GPU chips in the global AI contest."}],[{"start":508.40000000000003,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1779259793_5195.mp3"}

版权声明:本文版权归FT中文网所有,未经允许任何单位或个人不得转载,复制或以任何其他方式使用本文全部或部分,侵权必究。

马克•扎克伯格是不是太贪心?让股东来决定

如果这位亿万富翁创始人在意人们的看法,就应该说服其他投资者支持他的愿景。

AI监管不能照搬银行业模式

金融监管体系历经一个世纪的拼凑修补才逐渐成形。AI需要一套针对自身特点设计的监管机制。

FT社评:星链风波凸显欧洲亟须在太空领域迎头赶上

美国科技遥遥领先,但布鲁塞尔仍需建立自己的卫星网络并具备发射能力。

新加坡寻求追究相关公司责任,应对十年来最严重空气污染

政府威胁将对10家企业采取法律行动,称这些企业是邻国印尼山火的幕后推手,而山火造成的烟霾正在困扰新加坡。

美债收益率冲击下,欧洲必须自寻出路

埃里安:欧洲不能只盼着大西洋彼岸的借贷成本回落,它必须着手解决自身问题。

AI与通胀:央行行长该担忧到什么程度?

凯恩斯:这取决于一个价值万亿美元问题的答案:AI是否真能带来预期的生产力提升?
设置字号×
最小
较小
默认
较大
最大
分享×