Google fait entrer Marvell dans son écosystème de puces IA
Google 与 Marvell 达成了一项同时涉及 AI 芯片供应和资本合作的商业协议，核心围绕 Google 的 TPU（Tensor Processing Units）展开。根据协议，Marvell 将为 Google 提供定制芯片方案；与此同时，协议还包含一项认股权证机制，允许 Google 逐步增持 Marvell 股份，最多可达到 5,890 万股。
这意味着，Google 不仅在其自研 AI 芯片生态中引入了 Marvell 作为供应与技术合作伙伴，也为双方关系增加了资本层面的绑定。对 Google 来说，这有助于强化 TPU 相关供应链和定制化能力；对 Marvell 而言，则意味着其在 AI 芯片市场中的角色进一步提升，并获得来自大型云与 AI 平台客户的更深度合作机会。
目前公开信息显示，这笔交易的重点在于两部分：一是面向 Google TPU 的定制芯片合作，二是通过认股权证为 Google 提供未来入股 Marvell 的路径。协议细节尚未完全披露，但其结构表明，Google 正在通过产业合作与潜在股权参与，进一步巩固其 AI 芯片生态。
Google has struck a commercial agreement with Marvell Technology that spans both custom silicon for its Tensor Processing Units (TPUs) and a warrant mechanism that could let Google gradually build a stake in Marvell of up to 58.9 million shares.
The deal strengthens Google’s AI chip ecosystem by tying Marvell more closely to the hardware stack behind its TPUs, the specialized processors used to accelerate machine learning workloads. In parallel, the equity component gives Google a potential path to increase its ownership in Marvell over time, adding a strategic financial dimension to the partnership.
No further terms were disclosed in the brief announcement, but the structure of the agreement suggests a deeper collaboration between the two companies at a time when demand for custom AI infrastructure continues to rise.
L’accord porte à la fois sur des puces sur mesure pour les TPU (Tensor Processing Units) de Google et sur un mécanisme de bon de souscription permettant à Google de monter progressivement au capital de Marvell à hauteur d’un maximum de 58,9 millions d’actions. Marvell Technology et Google ont récemment conclu un accord commercial relatif […]
Cet article Google fait entrer Marvell dans son écosystème de puces IA a été publié par VIPress.net.
Google signed a commercial deal with Marvell that includes custom AI chips for Google’s TPU ecosystem and an option structure allowing Google to gradually build a stake in Marvell up to 58.9M shares, strengthening control over AI compute supply.
Google — designs TPU AI accelerators and cloud AI infrastructure, based in the US.
Marvell Technology — designs semiconductors and networking/compute silicon, based in the US.
[Strongly track] — equity build + custom chip commitments could shift competitive dynamics in AI accelerator supply chains and pricing.