The short version

It's a simple dilemma: AI benefits from network effects. If Microsoft integrates OpenAI into its products, it will increase the capabilities of OpenAI--more data, training, and apps. As AI features become "must-have" features, Microsoft's business success will increasingly depend on OpenAI. In the short run, AI features would give Microsoft a competitive edge, but in the long run, those features would shift shareholder value creation to OpenAI away from Microsoft. Microsoft will have no choice but to acquire OpenAI (or rights equivalent to ownership).

The long version

Following up on my earlier post, it has now been publicly reported that Microsoft has been working with OpenAI about integrating their AI technology into Microsoft Office and into Bing: https://appleinsider.com/articles/23/01/07/microsoft-plans-adding-openai-chatbot-tech-to-productivity-apps, https://www.theguardian.com/technology/2023/jan/05/microsoft-chatgpt-bing-search-engine.

If done well, this would create immense value for Microsoft's customers. I've already talked about how extensively I use AI myself and how I no longer use Microsoft Word as it's nowhere near as efficient as writing with AI assistance. These assistance-type features will rapidly become essential for any competitive product.

However, integrating AI from OpenAI creates a fundamental business challenge for Microsoft. Typing, printing, document sharing, and so forth--the "base" features of Microsoft's productivity tools--are effectively commodity features now. Google Docs, Apple Pages/Keynote/Mail, and so on are perfectly fine substitutes for Microsoft Office. But those products have yet to demonstrate features that dramatically shift market share away from Microsoft Office. Google arguably has come the closest, but even then, many Google users still maintain licenses to Microsoft Office.

AI could change that--and change it fast. How fast it changes, though, is orthogonal to the business dilemma. If AI features become the competitive battleground for knowledge worker productivity tools, Microsoft will have to have those features, plain and simple. But if it gets those features from a third-party company (OpenAI), that shifts value creation away from Microsoft and to OpenAI.

This value transfer problem is seen over and over in business. Ironically, Microsoft itself benefited from it with their original MS-DOS deal with IBM for the IBM PC: https://thisdayintechhistory.com/11/06/ibm-signs-a-deal-with-the-devil/. With that deal for the MS-DOS operating system, Microsoft captured the network effects and value creation for owning the underlying software platform. The IBM PC business became just one of many hardware manufacturers.

(source: https://commons.wikimedia.org/wiki/File:Original_IBM_PC_%2851332334711%29.jpg)

AI similarly has powerful network effects. OpenAI's current business model is genius. By allowing free usage of chatGPT, they are getting millions of "free" employees, all actively and eagerly providing more data and more training to the system. At the most fundamental level, AI benefits from more data and training. It's one of the reasons why Google has been able to maintain a substantial lead in search (until maybe now), despite the billions of dollars Microsoft has invested in Bing. Google simply had more data--and specifically, more usage data. Google had more queries than Bing and more clicks (indicating what websites are better results) than Bing. It was such a formidable advantage to overcome that Microsoft eventually resorted to outright just copying Google (https://www.wired.com/2011/02/bing-copies-google/).

More importantly, OpenAI is, at its core, a software platform for applications. They have done a great job creating easy-to-use and capable APIs for other companies to leverage in their products. I'm using those APIs myself extensively-more on that in upcoming posts. Being the fundamental AI platform will only further accelerate the network effect: more apps on OpenAI means more data, more training, and better results for OpenAI, which in turn will lead to more apps on the platform.

Positive feedback loops are exceptional-exceptional for the loop owners and exceptionally difficult for others to disrupt. Microsoft knows this firsthand with Bing (and recall that Satya Nadella, Microsoft's CEO, was the executive in charge of Bing for many years). Even if competing AI platforms emerge (and they will), those competitive platforms must be exceptionally inventive to overcome the early lead OpenAI is achieving. Microsoft has outstanding in-house AI talent and has invested in AI for decades. However, all that talent does not change the network effect that OpenAI is enjoying now.

So why has Microsoft yet to announce an acquisition? They may well be already in discussions. Either way, the internal debate is likely discussing the following dilemma: if Microsoft buys OpenAI too soon, it could open up the door for strengthening the competition. It's one thing for companies worldwide to rely on an independent third-party platform, but those companies might pause before being solely dependent on Microsoft and want a viable alternative. We see a similar dynamic with Amazon and Shopify. If you are an e-commerce vendor, there are a lot of advantages to being on the Amazon platform. But Amazon is aggressive--do you really want to build on Amazon if Amazon is going to learn from you and build competing products to your own? Maybe it's better to use Shopify and not have the competition from your platform provider: https://www.cnbc.com/2022/07/06/amazon-investigated-by-uk-watchdog-over-its-marketplace-practices.html.

Additionally, there are countless examples of companies acquiring disruptive technologies too soon--Yahoo's acquisition of Overture in 2003 is a particularly noteworthy example. Even though Overture at the time was dominant, it did not stop Google from rapidly overtaking the search market: https://www.wired.com/2007/02/yahoo-3/. Even though Overture had by far the early lead, Google learned from Overture and effectively built the better mousetrap.

Conversely, from a shareholder value creation perspective, Microsoft would have a tough time justifying having their knowledge worker business depend on OpenAI as an external, independent company. Microsoft is supposed to generate value for its shareholders, not OpenAI shareholders! The more OpenAI's AI technology becomes crucial to Microsoft's business, the more pressure there will be for Microsoft to acquire OpenAI.

There would also be legal and regulatory reasons for Microsoft to acquire OpenAI. Many exciting use cases for AI in Microsoft Office involve analyzing or creating data (e.g., writing assistance for documents or analyzing a spreadsheet)--data that could be proprietary and confidential to a company. That customer data would need to be kept separate and isolated between customers in some fashion. Imagine an Airbus executive writing a business plan for a new airliner, only to have the AI make suggestions that include the details of Boeing's new airplane!

This privacy and rights problem for AI is a deep topic. For this note, I'll simply comment that those challenges need both technological solutions as well as business solutions. Let's say a company is using OpenAI-powered Microsoft Office, but sensitive PII customer data is sent to OpenAI in a way that violates European GDPR regulations. Who is liable? How would that be handled? Sure, all sorts of contracts, indemnifications, and so forth could be crafted, but at some point, it would be much easier for Microsoft to simply own OpenAI.

A subtle nuance here is Microsoft's one-billion-dollar investment into OpenAI: https://openai.com/blog/microsoft/. Microsoft has a highly talented business development team. Likely, the OpenAI investment came along with contractual terms that would facilitate an acquisition (or potentially make an acquisition by others more difficult). Those terms may enable Microsoft to postpone deciding on an acquisition. But at the end of the day, if AI becomes fundamental to Microsoft's business, from a shareholder perspective, they will have to buy.

Of course, for the completeness of the discussion, Microsoft could construct a partnership agreement that effectively gives them the same value as an outright acquisition. It would be complex but conceptually possible. Such an agreement would need to result in Microsoft having rights similar to those they would enjoy as owners (e.g., Google couldn't buy OpenAI and shut Microsoft off).

Bottom line: I am quite sure this discussion is happening now inside Microsoft. As an outside observer, I think it would be a smart move for Microsoft. But regardless of the outcome, it's going to be an exciting year!

Microsoft will have to buy OpenAI in 2023