📰 The InformationThe InformationCodex · 实盘不做把握 8/10
北京时间·🕒 发布 13:00:41(美东 01:00)·⟳ 抓取 13:00:59 +18s·⚡ 决策 07-25 13:03:11 +2m12s·94.3 天前
英伟达与SK集团披露5000亿美元级AI工厂、HBM和Naver数据中心合作
事件本身对SK海力士HBM长期需求、SK集团AI基础设施布局和Naver数据中心扩容方向偏正面。
📄 原文The Information↗ 原文
Nvidia Forms $500 Billion AI ‘Partnership’ With Memory Chip Giant SK
Nvidia on Friday announced a $500 billion partnership with the conglomerate that owns SK Hynix, whose high-bandwidth me…
不做 SK海力士（000660.KS）
同一英伟达-SK集团5000亿美元AI工厂/HBM合作已在近72小时处理库中由company_newsroom、Reuters和Twitter等多源处理过，本条The Information未提供可重开交易的新事实，不重复推卡。
不做 SK Inc.（034730.KS）
同一具体催化已处理且未推卡；SK集团层面收益路径和金额归属仍不清晰，不因重复转述重新建仓。
不做 Naver（035420.KS）
Naver投资和数据中心扩容内容已包含在同一英伟达-SK合作报道簇内，且Naver 7月23日已大涨、7月24日回落，当前无新增未定价事实。
展开详情(成交明细 / 研究档案 / 原文全文 / 决策记录链)
研究档案
🔎 新鲜度:最早出现 2026-07-25 · 扩散 high · price-in 风险 high ↗
关键理由
- 事件本身对SK海力士HBM长期需求、SK集团AI基础设施布局和Naver数据中心扩容方向偏正面。
- 内部已处理消息库在近72小时内多次命中同一具体催化，包括公司新闻源、Reuters、Twitter和The Information同题转述，当前信号不应重复推卡。
主要风险
- 该合作金额口径包含AI工厂、融资、数据中心和长期存储协作，Nvidia与SK方未明确5000亿美元由谁支出、何时确认收入。
反向证据
- 做多前查反向 · 来源 kb_search · 日期 2026-07-25 · 影响 weakens
内部KB未发现本条之前未处理的新交易性增量；命中多条存储长单和HBM景气旧背景，支持该主题已广泛讨论。
检索:Nvidia SK Hynix Naver Brookfield 2GW AI cloud $500 billion partnership HBM
工具调用 · 历史记录,结果未知(5 种)
- search_processed_events · 结果未知
- supply_chain_lookup · 结果未知
- stock_quote · 结果未知
- query_modeling_kb · 结果未知
- kb_search · 结果未知
原文全文
展开原文 · 2,035 字
Nvidia on Friday announced a $500 billion partnership with the conglomerate that owns SK Hynix, whose high-bandwidth memory chips are critical to Nvidia’s AI servers. The partnership aims to help Nvidia access more memory chips and fill more data centers with its servers.
Nvidia’s representatives didn’t elaborate on which firms would spend such a large amount or precisely what it would be for. Nvidia and the SK Hynix parent company said they would build two gigawatts of AI cloud capacity together, which could cost around $100 billion, based on today’s prices for data centers that consume such power. Their engineers also plan to work more closely together to develop new high-bandwidth memory products.
An Nvidia executive said the company’s biggest priority is to get its AI servers online as fast as possible, so it’s helping more customers find land, power, memory and raw materials to make AI data centers happen quickly.
“ What’s limiting the access to compute capacity is the fact that it’s very hard to find powered data centers, powered land,” said Raj Mirpuri, Nvidia’s vice president of global AI clouds and infrastructure, in a briefing with reporters. For instance, he said Nvidia helped Naver, the South Korean search firm, secure access to land and power to expand an AI data center from 55 megawatts to 200 MW by 2028. (Such a facility is relatively small compared to AI facilities underway in the U.S.)
Nvidia also plans to invest $1 billion in Naver and asset manager Brookfield has committed to fund the project with $9 billion. Naver is set to pay for the rest. The new data center is expected to start with Nvidia’s Grace Blackwell server racks and later add next-generation Vera Rubin racks.
Nvidia is increasingly using its balance sheet to help its customers afford more AI chips and help its cloud provider partners get financing to build new facilities for the chips.
The AI giant is continuing to use its balance sheet to help its biggest customers and partners buy and manufacture its AI servers.
Nvidia’s representatives didn’t elaborate on which firms would spend such a large amount or precisely what it would be for. Nvidia and the SK Hynix parent company said they would build two gigawatts of AI cloud capacity together, which could cost around $100 billion, based on today’s prices for data centers that consume such power. Their engineers also plan to work more closely together to develop new high-bandwidth memory products.
An Nvidia executive said the company’s biggest priority is to get its AI servers online as fast as possible, so it’s helping more customers find land, power, memory and raw materials to make AI data centers happen quickly.
“ What’s limiting the access to compute capacity is the fact that it’s very hard to find powered data centers, powered land,” said Raj Mirpuri, Nvidia’s vice president of global AI clouds and infrastructure, in a briefing with reporters. For instance, he said Nvidia helped Naver, the South Korean search firm, secure access to land and power to expand an AI data center from 55 megawatts to 200 MW by 2028. (Such a facility is relatively small compared to AI facilities underway in the U.S.)
Nvidia also plans to invest $1 billion in Naver and asset manager Brookfield has committed to fund the project with $9 billion. Naver is set to pay for the rest. The new data center is expected to start with Nvidia’s Grace Blackwell server racks and later add next-generation Vera Rubin racks.
Nvidia is increasingly using its balance sheet to help its customers afford more AI chips and help its cloud provider partners get financing to build new facilities for the chips.
The AI giant is continuing to use its balance sheet to help its biggest customers and partners buy and manufacture its AI servers.
决策记录链 · 只追加
每次领取产生一条记录,重新决策追加一条并以 supersedes 指向被替代的记录(ADR 0014)。
这条决策早于决策账本(2026-10-08 起逐条追加),没有追加记录。