Alliances are conventionally framed as a method of offense combining forces to multiply power. Less appreciated, however, is the defensive role alliances play. Forming alliances can constrain your friends. In the case of OpenAI and Microsoft, viewing the alliance from a lens of defense and containment provides insight into the multibillion-dollar deal.

Microsoft, the epitome of tech incumbents and oldest of the MAANG, has much to lose to new hot upstarts such as OpenAI. Together, Microsoft’s two biggest revenue sources Azure and Office account for more than one half the company’s revenue. These two services also happen to be particularly threatened by OpenAI.

On the Office front, the biggest threat to the core suite of products is arguably large language models (LLMs) such as GPT which have the potential to make 10x product improvement. Significant improvements like these are often the only means to break the tight grasp of incumbents. While OpenAI seems intent on remaining a backend service rather than primarily a consumer-facing product, it nonetheless will service startups, some of which may hope to displace Office.

In fact, OpenAI is making efforts to support these startups. The company has launched an investment fund, giving startups cash and access to the latest and greatest of their models. This set OpenAI on a trajectory to displace Microsoft and own a part of the new regime replacing Office. Thanks to Microsoft’s investment and their newly strengthened partnership, OpenAI is now constrained. Instead of owning part of the new software regime, the old regime remains and now owns a part of OpenAI.

Microsoft will have not only OpenAI’s best LLMs but also at the cheapest price, making it difficult for upstarts to compete. Furthermore, the fact that Microsoft is partnered with OpenAI should send a chilling signal to anyone trying to use LLMs for a product competing with Office. Already Microsoft’s Github code pilot seems to have been a success, and the company seems even more intent on replicating it across its suite of apps.

On the cloud and Azure front, OpenAI is constrained as to not build their own customized hardware or their own data centers. While rival Google and DeepMind may be innovating on the hardware, introducing the TPU, and later commoditizing it as a rentable service, OpenAI will for the most part will not be touching their own hardware. This staves off potential competition for Azure, which sells their compute to AI researchers at places like here at Stanford.

The move by Microsoft also helps contain its rival Google. Microsoft’s biggest competitor, seemed to have the edge in AI with custom hardware, extensive libraries, and acquisition of DeepMind. Microsoft seems to have been lacking in this department until now.

This new rush has been harnessed to draw in talent. Shortly after releasing ChatGPT on Nov 30, OpenAI changed their call to action on their landing page from reading about DALLE-E to the “explore careers” page on Dec 7. Chat GPT is a massive outbound sales effort, but also as this landing page indicates a way for OpenAI to hire the best AI talent. AI talent that could otherwise be ending up at Google. Microsoft funding OpenAI’s multimillion dollar ChatGPT extravaganza is a way of diverting potential talent from Google and to its new partner.

Thus, the OpenAI Microsoft partnership can been seen less as an offensive play—trying to revive Bing search is a nice coincidental and fringe benefit. Instead, the deal can be seen as an effort to protect Microsoft’s incumbency and constrain OpenAI and their customers. This is not to downplay OpenAI’s benefit and increased resources and distribution, though while the deal may be offensive for the young OpenAI, the same does not apply to Microsoft. ∎