Microsoft’s deal with OpenAI was just the first step. They want (probably) to get rid of them.

Satya Nadella quickly took over the role consigliere from Microsoft. His predecessor, Steve Ballmer, was a major contributor to failure in both the search engine war and (especially) the mobile device war. Mr Nadella suggested straighten your march immediatelyand soon began to throw off ballast on one side and change focus on the other.

His strategy worked. His company focused on mobile devices (albeit indirectly) and, above all, on the cloud, where Azure managed to become the absolute hero of this panorama. His pulse never wavered when it came to making decisions, or when it came to acquiring companies or investing in new infrastructure.

The numbers are dizzying. Since its arrival, Microsoft has invested 170 billion euros in acquiring companies (e.g. Activision Blizzard – 68,700 million, LinkedIn – 26,200, GitHub – 7,500), but it has also been outlining in recent months what its big bet is for the future. The company has invested more than $30 billion in future data centers, adding to the $10,000 investment in OpenAI. In the last two cases with one purpose:

Victory in the battle of artificial intelligence.

However, for now, he needs help to achieve this. Although the company has been working in this field for many years, the development of generative artificial intelligence models has caught Microsoft, like other major players, in the wrong situation. Satya Nadella decided to solve this problem before anyone else and soon began stimulate the adoption of OpenAI technology on all Microsoft products.

Here’s how we got here: We were inundated with Microsoft co-pilots, and that strategy was cemented at the recent Build 2024 conference with the reveal of the promising Copilot+ PC. Artificial intelligence is a great argument for Microsoft to continue to lead the technology battle in the coming years, and Open AI is the perfect candidate to achieve this goal.

At least for now.

As explained in The Wall Street Journal, various movements seem to make it clear that Microsoft’s goal is to get rid of OpenAI at some point in the future. The company is taking advantage of ChatGPT while also taking small but important steps to eventually run your own LLM programs which manage to compete with those that now set the standard (GPT-4, Claude 3 Opus, Llama 3).

This journey has already begun with the Phi-3, a model capable of running on mobile devices and which, of course, has many competitors. Google has the Gemini Nano, Anthropic has the Claude 3 Haiku, and Apple has its own model without a specific name that debuted in Apple Intelligence on the iPhone 15 Pro/Max (among other devices).

But it is clear that Microsoft is not going to agree to this. A few months ago, he hired Mustafa Suleiman, who co-founded DeepMind and who left Google – with some controversy – to co-found Inflection AI, maker of the Pi chatbot. Despite raising $1.3 billion, the company hasn’t rallied, so Microsoft’s “come and lead our commitment to artificial intelligence” pitch comes at the right time.

A few months ago, Suleiman led Microsoft’s efforts to create your own models artificial intelligence. Their work seems to be taking shape, and with it, Microsoft already has its first GPT-4 competitor called MAI-1 (MicrosoftAI). The model is huge (500B, 500 billion parameters), but it is curious that it was not discussed at the Build conference dedicated to the Copilot+ PC.

The freedom of maneuver they gave Suleiman is unusual: his team communicates not with Microsoft Teams, its own product, but with Slack. There is nothing there. The artificial intelligence division is Redmond’s favorite, and so far the hardware seems to be losing weight.

Panos Panay, its top manager, left the ship in September 2023. According to the WSJ, budgets were cut that governed this division, as well as the division that worked on HoloLens, which effectively canceled HoloLens 3 in February 2022.

At the moment, yes, everything is unknown about these attempts to compete with OpenAI and the rest of the alternatives. The capabilities of the MAI-1 are currently unclear, but it seems clear that Microsoft wants to follow in Apple’s footsteps. Yes, the Cupertino firm has an agreement with OpenAI, but it is almost symbolic because they only use ChatGPT as a last resort.

Redmond seems to want to get into that future, too. For now, the alliance with OpenAI is a win-win for everyone: it allows Microsoft to give its customers easy access to one of the best AI models on the market. The question, of course, is How long will Microsoft’s love affair with OpenAI last?.

If things move as fast as the rest of AI is progressing, perhaps not so much.

In Hatak | Apple Intelligence aims to be the perfect excuse for something important for the future: selling more iPhones