The AI Talent Wars

When is the line being crossed?

Microsoft has been under an anti-trust microscope before.

The Federal Trade Commission started an inquiry in 1990 focusing on their bundling of Internet Explorer, and after years in court, the Department of Justice announced in 2001 that it was no longer seeking to break up Microsoft and would instead seek a lesser antitrust penalty. This resulted in what many call a “lost decade” for Microsoft. They were trying to navigate the antitrust action while, at the same time, transforming the company. Fast forward to the last ten years and, under the current leadership, Microsoft has become one of the leading Cloud providers and is now playing a major role in Artificial Intelligence. They deserve significant praise for that. It is not easy to completely retool a large business and deal with the technology and cultural aspects of that change. Believe me, I have been there.

What is happening today in the AI space, however, is concerning.

Microsoft has devised a robust strategy to monetize AI be it as “copilots” on top of their core Office franchise or available as services on their Azure Cloud platform, and has also positioned itself as one of the major cloud compute capacity providers for those training AI models or using AI in general. So far, so good - great innovation and masterful business strategy.

Is this coming at the expense of too much concentration in this segment of the industry?

Theories of harm in anti-trust legal doctrine evaluate the potential threats resulting from increased market power and compare them to the efficiencies that can be inherent to a potential horizontal or vertical merger of rivals. If the harms of increased market power outweigh the benefits that the consumers can get from the efficiencies, there is a case to be discussed and, maybe, brought to court. To assess this, what is key is the definition of the market - cases are won and lost depending on how the parties convince a judge to look into the market.

One can say that we are in the relatively early days of the current AI wave and there are still several companies jostling for leadership. The market is fragmented with:

• Multiple players - Amazon, Anthropic, Cohere, Google, Hugging Face, IBM, Meta, Microsoft, Mistral, OpenAI …  

• Several approaches - open source, closed models, …

• Emerging technologies - LLMs, liquid AI, objective-driven AI, GFlowNets, ...

But what if we define the market as the market for AI talent? Have you noticed what is going on? First Microsoft, after their $13B investment and the board events from last year, essentially took over OpenAI. This week, they announced that they are hiring two of the three co-founders of Inflection AI and most of their 70-person staff. Inflection AI’s roots go back to the British startup DeepMind previously acquired by Google. The third leg of the stool is the partnership announced in late February with the French national AI champion Mistral, including a research and development cooperation to access public sector accounts in the European market. Mistral provides open-source models, as opposed to OpenAI proprietary ones, which positions Microsoft well in these two camps.

Are these masterful strategic moves or are we approaching a point in which there is too much talent concentrated with one player? These are aggressive moves to capture talent and know-how while leaving an empty shell company behind as in the case with Inflection AI. The same would have happened with Open AI if their team had fled to Microsoft last November.

Maybe it is time for the FTC to visit Redmond again.

PS: As of this writing, Microsoft’s market cap is close to $3.2T (trillion dollars) making it the most valuable company in the world and the “6th largest country” behind only the GDPs of the US, China, Japan, Germany, and India.