内容来源：https://aibusiness.com/generative-ai/nvidia-semiconductor-vendor-expand-partnership

内容总结：

英伟达斥资35亿美元入股联发科，加码AI定制芯片与平台化布局

全球AI芯片巨头英伟达周一宣布，将扩大与中国台湾半导体大厂联发科的合作伙伴关系，并向其投资35亿美元购买可转换债券。此举标志着英伟达正试图摆脱对GPU销售的单一依赖，转而成为AI市场各类玩家的核心基础设施平台供应商。

根据协议，联发科将采用英伟达的NVLink Fusion平台，该平台允许云服务商将定制化处理器直接集成到英伟达的AI超算生态中。双方将合作开发整合英伟达GPU与联发科系统级芯片（SoC）的PC、AI开发者超级计算机及企业级工作站。

此次合作正值英伟达客户群发生结构性变化之际。包括大型云厂商和前沿模型公司在内的诸多客户，正越来越多地自研ASIC（专用集成电路）以降低对英伟达GPU的依赖，OpenAI近期就推出了自研AI推理芯片Jalapeno。与此同时，AI市场中的大量中小型玩家则转向支持ASIC的基础设施供应商，联发科正是其中的重要代表。

分析师指出，英伟达此举意在“确保自己成为市场上尽可能多解决方案的顶级品牌和技术供应商”。通过入股联发科，英伟达不仅能够覆盖AI芯片市场的中端需求，更向外界传递出“不仅是GPU公司，也能提供CPU或混合芯片”的信号。Signal65总裁兼分析师瑞安·施劳特表示，英伟达已将联发科视为切入AI ASIC业务快速增长的重要跳板，并借此从单纯的芯片销售商转型为面向全行业、尤其是中小AI厂商的服务与平台提供方。

华盛顿大学教授奇拉格·沙阿认为，英伟达正借此进入AI市场的“中间层”，与谷歌、OpenAI等基于其芯片构建平台的公司展开直接竞争。“英伟达正在成为AI一切的平台，即便你不是大厂。”他说。

此外，英伟达还希望通过与联发科的技术整合，继续推动其“AI工厂”（AI Factories）战略。该战略将AI工厂定位为超越传统数据中心的新阶段，旨在持续生产“智能”。英伟达正将NVLink及此类合作视为对冲GPU销量波动、并将其技术影响力渗透到AI工厂各个关键环节的重要手段。

尽管英伟达进军AI服务市场的时间较晚，但其充沛的资本储备为其保持领先地位提供了底气。沙阿同时提醒，这种大规模投资也存在潜在风险——英伟达已在多家AI供应商中广泛持股，若其自身遭遇财务问题或市场动荡，可能对AI行业产生连锁反应。

中文翻译：

由谷歌云赞助
选择您的首批生成式AI应用场景
要开始使用生成式AI，首先要关注那些能够改善人类获取信息体验的领域。

此举表明这家AI硬件巨头如何旨在通过第三方维持其主导地位。

英伟达周一扩大了与联发科的合作关系。联发科是一家总部位于台湾的大型半导体供应商，为智能手机、电视和无线通信设计片上系统。

此举表明这家AI硬件和软件提供商正试图超越单纯销售GPU，转而成为AI市场中大小玩家都会使用的供应商。

作为扩大合作的一部分，英伟达投资35亿美元购买了联发科发行的可转换债券。联发科将采用英伟达的NVLink Fusion平台。该平台允许云服务提供商将定制处理器直接集成到英伟达的AI超级计算生态系统中。两家供应商将携手推动搭载英伟达GPU与联发科SoC（片上系统）的个人电脑、AI开发者超级计算机和企业级工作站的性能提升。

此次合作正值越来越多的英伟达客户（包括超大规模云服务商和前沿模型制造商）将更多精力放在训练和部署自己的ASIC（专用集成电路）上。最近，OpenAI推出了其首款定制ASIC“Jalapeno”，专为处理AI推理而设计。虽然英伟达的客户正在打造自己的芯片以减少对GPU的依赖，但AI市场中的许多中小玩家正转向支持ASIC的基础设施供应商，例如联发科。

因此，通过这次合作，英伟达希望确保自己仍然是整个AI市场首选的AI基础设施提供商。

“这就是为了确保他们尽可能成为市场上最多解决方案的顶级品牌和顶级技术提供商。”Signal65总裁兼分析师瑞安·施劳特表示。

他补充说，英伟达已经注意到联发科业务中增长最快的部分就是作为AI ASIC供应商。因此，华盛顿大学教授奇拉格·沙阿表示，这家AI芯片巨头看到了联发科作为合作伙伴的机会，可以帮助自身成为小型AI供应商更广泛的平台提供商。

“这实际上是英伟达试图成为一切AI的平台，即使你不是大玩家也可以使用。”沙阿说。他补充说，通过这样做，英伟达变得更加注重服务和平台提供，与谷歌、OpenAI等公司展开竞争——这些公司一直在使用英伟达的芯片并在其基础上构建平台。

“英伟达正在进入中间层级。”沙阿说，“这是一个拥挤的市场。”他表示，通过联发科提供XPU，英伟达传达了这样一个信息：它不仅仅是一家GPU公司，同样可以提供CPU或混合处理器。

英伟达还将与联发科的技术整合视为持续推动AI工厂建设的一种方式。

在过去几年中，英伟达将AI工厂定位为超越传统数据中心的下一个阶段，能够持续制造智能。因此，NVLink以及与联发科这样的合作伙伴关系可以帮助其继续推动这一叙事。

“英伟达将NVLink和这类技术视为对GPU销售的一种对冲，并将自身技术的定制ASIC集成视为对尽可能多的AI工厂计划施加影响的下一个最佳途径。”施劳特说。

虽然英伟达进入AI服务市场较晚，但它拥有保持领先所需的资金实力，其对联发科的大规模投资就是明证。

“他们占据优势，因为他们拥有大量涌入的资金可以注入这个市场。”沙阿说。

他补充说，英伟达的资金也可能成为弱点，因为它投资了许多AI供应商，如果英伟达遭遇财务问题或动荡，可能会在AI市场产生连锁效应。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The move shows how the AI hardware giant is aiming to maintain its dominance, even through third parties.
Nvidia on Monday expanded its partnership with MediaTek, a large Taiwan-based semiconductor vendor that designs system-on-chip systems for smartphones, TVs and wireless communications.
The move shows how the AI hardware and software provider is trying to move beyond selling GPUs to become the provider that both big and smaller players in the AI market will use.
As part of the expanded partnership, Nvidia invested $3.5 billion in convertible bonds issued by MediaTek. MediaTek will adopt the Nvidia NVLink Fusion platform. The platform lets cloud providers integrate custom processors directly into Nvidia’s AI supercomputing ecosystem. The two vendors will work together to power PCs, AI developer supercomputers and enterprise-class workstations that incorporate Nvidia GPUs with MediaTek SoCs (system-on-chip).
The partnership comes as a growing number of Nvidia customers, including hyperscalers and frontier model makers, focus more on training and deploying their own ASICs (application-specific integrated circuits). Most recently, OpenAI introduced its first custom ASIC, Jalapeno, designed to handle AI inference. While Nvidia customers are creating their own chips to rely less on GPUs, many smaller players in the AI market are turning to infrastructure vendors that support ASICs, such as MediaTek.
Therefore, through the partnership, Nvidia is looking to ensure it remains the go-to AI infrastructure provider across the market.
“It's making sure that they are the top-level brand and top-level technology provider for as many of the solutions on the market as possible,” said Ryan Shrout, president and analyst at Signal65.
He added that Nvidia has noticed that the fastest-growing part of MediaTek’s business is serving as an AI ASIC provider. Therefore, the AI chip giant sees an opportunity in MediaTek as a partner to help it become more of a platform provider for smaller AI vendors, said Chirag Shah, a professor at the University of Washington.
“This is really Nvidia trying to be a platform for anything AI, even if you’re not a big player,” Shah said. He added that by doing this, Nvidia becomes even more of a service and platform provider, competing with Google, OpenAI and others, which have been taking Nvidia chips and building platforms on them.
“Nvidia is getting into that middle tier,” Shah said. “It’s a crowded market.” He said that by offering XPU through MediaTek, Nvidia conveys that it is more than a GPU company and can offer a CPU or a hybrid as well.
Nvidia also sees integrating its technology with MediaTek as a way to continue pushing AI factories.
In the past couple of years, Nvidia has positioned AI factories as the next stage beyond traditional data centers, enabling the continuous manufacturing of intelligence. Therefore, NVLink and partnerships such as the one with MediaTek could help it continue to push this narrative.
“Nvidia sees NVLink and these types of technologies as a hedge against GPU sales and custom ASIC integration of their technology as the next best thing to have influence and effect on as much of this AI factory directive as possible,” Shrout said.
While Nvidia is entering the AI services market late, it has the capital it needs to stay on top, as evidenced by its sizable investment in MediaTek.
“They have an upper hand in the sense that they have tons of this influx of capital that they can inject into this market,” Shah said.
He added that Nvidia’s capital can also be a weakness because it has invested in many AI vendors, and if Nvidia encountered financial problems or turbulence, that could have a cascading effect in the AI market.