GEO

OpenAI与微软修订合作后,收入分成设上限和AGI条款取消意味着什么?

2026/5/8
OpenAI与微软修订合作后,收入分成设上限和AGI条款取消意味着什么?

AIAI Summary (BLUF)

OpenAI and Microsoft have revised their partnership, introducing a cap on revenue share payments from OpenAI to Microsoft, and allowing OpenAI to serve customers across any cloud provider including AW

核心洞察

OpenAI与微软的合作关系正在经历一次根本性的重塑:从“金主-明星创业公司”的单一依赖,转向以“竞争性合作”为核心的成熟企业联盟。协议中收入分成设上限和AGI条款的删除,标志着双方从“押注未来”的共生关系,进化为更务实、更具商业弹性的多供应商框架。这一调整不仅回应了OpenAI对独立性的渴求,更预示了未来AI巨头的云基础设施选择将不可避免地走向多云与混合。

The partnership between OpenAI and Microsoft is undergoing a fundamental transformation: from a single-thread dependency akin to "patron-star startup" to a "co-opetition" centered mature corporate alliance. The introduction of a revenue share cap and the removal of the AGI clause signal a shift from a symbiotic relationship built on betting on the future to a more pragmatic, commercially flexible multi-vendor framework. This adjustment not only addresses OpenAI's demand for independence but also foreshadows an inevitable move towards multi-cloud and hybrid infrastructure strategies for major AI players.

协议核心变更:收入分成设上限,独家性全面松动

OpenAI 与 Microsoft 于周一宣布了一项经过大幅修订的合作协议。核心变化在于,OpenAI 将对其向 Microsoft 支付的分成设定一个“总额上限”,而作为交换,Microsoft 放弃了“AGI(通用人工智能)条款”中的单方面解除权。

OpenAI and Microsoft announced a significantly revised partnership agreement on Monday. The core change involves OpenAI capping the total revenue share payments made to Microsoft, while Microsoft, in exchange, relinquishes its unilateral triggering rights under the controversial “AGI clause.”

财务条款详解

根据新协议,OpenAI 向 Microsoft 支付的分成将继续适用原有的 20% 比例,但支付金额已确定一个上限。这一上限将覆盖到 2030 年,期间无论 OpenAI 的技术进展如何,该比例依然有效。值得注意的是,用户通过 Azure 访问 OpenAI 模型时,Microsoft 此前需要向 OpenAI 支付费用,但现在这一反方向的分成已被取消。

Under the new terms, the revenue share paid by OpenAI to Microsoft will continue at the original 20% rate but is now subject to a hard cap. This cap will remain in effect until 2030, regardless of OpenAI’s technological milestones. Notably, Microsoft will no longer pay a revenue share to OpenAI for instances where users access OpenAI models through Azure, reversing a previous practice.

对于两家公司而言,这一财务框架本质上是一次“去风险化”操作:

  • 对于 OpenAI:上限的设定意味着其长期增长带来的红利不会被无限分走,保留了更多利润用于再投资和研发。
  • 对于 Microsoft:虽然放弃了未来的潜在超额收益,但确保了 2030 年前稳定的现金回流,将双方绑定在更确定的商业轨道上。

For both companies, this financial framework is essentially a “de-risking” maneuver:

  • For OpenAI: The cap ensures that the benefits from long-term growth won’t be indefinitely diluted, allowing more profit to be retained for reinvestment and R&D.
  • For Microsoft: While forgoing potential future excess returns, it secures a stable cash flow stream until 2030, locking both parties into a more predictable commercial trajectory.

独家性调整与多云战略

新协议彻底改写了两家公司的排他性条款。长期以来,OpenAI 的理论服务载体是 Azure,导致其对亚马逊 AWS 和谷歌云的使用受限。现在,OpenAI 不仅可以在任何云平台上提供服务,还与 AWS 签署了一项重大合作,使 AWS 成为其企业平台“Frontier”的“独家第三方云分发商”。

The new agreement completely rewrites the exclusivity clauses between the two companies. Historically, Azure was the de facto service vehicle for OpenAI, limiting its ability to use Amazon AWS or Google Cloud. Now, OpenAI can deliver its services on any cloud platform and has signed a major deal with AWS, designating AWS as the “exclusive third-party cloud distribution provider” for its enterprise platform “Frontier.”

以下表格梳理了主要云服务商在当前格局中的角色变化:

云服务商 角色定位 关键条款变化
Microsoft Azure 主要云提供商(优先权) 仍为“主要”云商,OpenAI 产品将在 Azure 上首发,除非 Microsoft 放弃该权利。
Amazon AWS 核心多云合作伙伴 获得“独家第三方分发”权,承担 OpenAI 企业平台“Frontier”的云分发。
Google Cloud 基础设施供应商 不再是禁区,OpenAI 可以自由使用其算力资源。
其他(CoreWeave, Nebius等) 补充算力 开放给 OpenAI 使用,以补充 Azure 的产能。

战略转型的动因:13亿美元投资下的“竞争性合作”

Strategic Drivers: "Co-opetition" Under a $13 Billion Investment

此次调整并非突然之举。2025年10月,OpenAI 完成资本重组并承诺在 Azure 上投入 2500 亿美元后,双方的关系已出现微妙变化。微软当时的投资在转股后估值 1350 亿美元,占 OpenAI 稀释后约 27% 的股份。这既是巨额资本注入,也构成了对 OpenAI 运营自由的隐性约束。

This adjustment is not a sudden move. Since OpenAI completed its recapitalization in October 2025 and committed $250 billion to Azure, the relationship has been subtly shifting. Microsoft’s investment, valued at $135 billion on a converted diluted basis (roughly a 27% stake), was both a massive capital injection and an implicit constraint on OpenAI’s operational freedom.

资源与独立性的矛盾

OpenAI 的收入负责人 Denise Dresser 在内部备忘录中指出,原有的合作关系“限制了我们在客户所在之处服务他们的能力”。这直接点明了核心矛盾:作为一家追求企业级客户的市场化公司,OpenAI 必须服务那些已经在 AWS 或 Google Cloud 上建立深度基础设施的客户。微软既是最大投资人,又是最强的 AI 竞对(通过 Copilot 和 Azure AI 平台),这种双重身份使得纯粹的独家绑定难以为继。

OpenAI's revenue chief, Denise Dresser, noted in an internal memo that the old partnership agreement “limited our ability to meet enterprises where they are.” This highlights the core conflict: as a market-driven company pursuing enterprise clients, OpenAI must serve customers who already have deep infrastructure commitments with AWS or Google Cloud. Microsoft’s dual role as both the largest investor and a formidable AI competitor (via Copilot and the Azure AI platform) made pure exclusivity untenable.

客户需求的倒逼

企业客户对“多云”和“多模型”的需求日益强烈。大型企业不希望将所有 AI 负载绑定在同一家超大规模云提供商或其旗舰模型上。Meta 近期承诺向 CoreWeave 和 Nebius 等云厂商投资 480 亿美元以补充算力,这便是一个信号。OpenAI 必须能够像 AWS、Azure 和 GCP 上的任何其他 SaaS 工具一样,被灵活调用。

Enterprise demand for "multi-cloud" and "multi-model" strategies is intensifying. Large enterprises are reluctant to bind all their AI workloads to a single hyperscaler or its flagship model. Meta's recent $48 billion commitment to CoreWeave and Nebius to supplement its own computing power serves as a strong signal. OpenAI must be as flexible and callable as any other SaaS tool running on AWS, Azure, or GCP.

影响与展望:开源模型、多云竞争的行业新格局

Impact and Outlook: A New Industry Landscape of Open Source and Multi-Cloud Competition

单一云依赖风险的降低

对于行业观察者而言,OpenAI 与 Azure 的“松绑”实际上降低了单一技术供应商风险。无论 OpenAI 的模型迭代速度如何,其底层算力将不再被动受限于 Azure 的扩张节奏。通过引入 AWS 和自有算力投资,OpenAI 获得了更强的算力冗余和议价能力。

For industry observers, the "decoupling" of OpenAI from Azure actually reduces single-vendor risk. Regardless of OpenAI's model iteration speed, its underlying compute power will no longer be passively limited by Azure's expansion pace. By introducing AWS and their own compute investments, OpenAI gains greater redundancy and bargaining power.

竞争压力与开源生态

与此同时,开源模型(如 Llama、Mistral)和新兴的 AI 云厂商(如 CoreWeave)正在快速侵蚀闭源模型的领地。微软之所以愿意调整协议,部分原因也在于它通过 Copilot 和 Azure AI 服务已经建立了强大的技术护城河,不再完全依赖 OpenAI 的独家 IP。事实上,微软将继续持有 OpenAI 模型的非独家 IP 许可至 2032 年。

Meanwhile, open-source models (like Llama, Mistral) and emerging AI cloud providers (like CoreWeave) are rapidly encroaching on the territory of proprietary models. Microsoft’s willingness to adjust the agreement is partly because it has built a strong technological moat through Copilot and the Azure AI service stack, no longer relying entirely on OpenAI’s exclusive IP. In fact, Microsoft will retain a non-exclusive license to OpenAI’s IP on AI models until 2032.

编辑点评

这一协议标志着 AI 行业“浪漫主义投资期”的终结。未来的 AI 头部公司,将不再寻求与一家云厂商形成“所有权绑定”,而是通过资金、算力和 IP 许可构建一个多极化的合作伙伴网络。对于初创企业而言,这是一记响亮的警钟:与巨头的深度绑定是双刃剑,迟早需要做出类似的结构性调整。OpenAI 正从科技巨头的实验室化身,走向一家真正独立的、拥有多云选择权的科技巨头。

This agreement marks the end of the “romantic investment era” in the AI industry. Future AI leaders will not seek an “ownership-driven” binding with a single cloud provider but will build a multipolar partner network through capital, compute, and IP licensing. For startups, this is a stark warning: deep ties with a giant are a double-edged sword, and a similar structural adjustment will likely be inevitable. OpenAI is transitioning from being the protégé of a tech giant to a truly independent, multi-cloud-enabled tech powerhouse.

常见问题(FAQ)

OpenAI与微软新协议的主要变化是什么?

新协议设定了OpenAI向微软的收入分成上限,允许OpenAI使用AWS和Google Cloud等任何云服务商,移除了AGI条款,微软保留非独占IP许可至2032年。

为什么OpenAI要与多家云服务商合作?

为了减少对Azure的单一依赖,提升独立性和商业弹性,满足客户对多云和混合部署的需求,并灵活配置算力资源。

微软在OpenAI的投资估值和股份占比如何?

微软投资转股后估值1350亿美元,占OpenAI稀释后约27%股份,标志着双方从共生关系转向竞争性合作。

← 返回文章列表
分享到:微博

版权与免责声明:本文仅用于信息分享与交流,不构成任何形式的法律、投资、医疗或其他专业建议,也不构成对任何结果的承诺或保证。

文中提及的商标、品牌、Logo、产品名称及相关图片/素材,其权利归各自合法权利人所有。本站内容可能基于公开资料整理,亦可能使用 AI 辅助生成或润色;我们尽力确保准确与合规,但不保证完整性、时效性与适用性,请读者自行甄别并以官方信息为准。

若本文内容或素材涉嫌侵权、隐私不当或存在错误,请相关权利人/当事人联系本站,我们将及时核实并采取删除、修正或下架等处理措施。 也请勿在评论或联系信息中提交身份证号、手机号、住址等个人敏感信息。