Microsoft prefers its own 7 that a 10 of OpenAi. The 13,000 million invested in Openai have just gosses meaning
Microsoft has launched Mai-1his first model of the fully developed at home. This In the 13th position of Lmarenabelow those of Anthropic, Depseek, Google and, of course, OpenAi. It is not the best model and is not even close, but that could be exactly what Satya Nadella had in mind. Why is it important. Technological do not need absolute excellence to master markets. They need control, integration and margins. Microsoft has understood it since the time of MS-DOS: it is better to have a sufficiently good product than to depend on the excellence of others. Windows was never the best operating system. Internet Explorer was not the best browser. Excel is the best, but it took years to overcome Lotus 1-2-3. All, in any case, ended up dominating their markets because Microsoft controlled development, distribution and, above all, integration with the rest of their ecosystem. The money trail. The 13,000 million that Microsoft has invested in Openai They begin to seem less one bet and more a university enrollment. Microsoft has paid for: Early access to GPT technology while building its own infrastructure. Time to learn what works in the pressure of being pioneers. Instant credibility of offering “the best model” through COPILOT. A perfect excuse to build large gpus clusters that now uses for Mai-1. Suleyman has made it clear: they have trained Mai-1 with 15,000 gpus H100 (Grok uses 100,000, to compare) and have a new generation GB200 cluster operational. This infrastructure was not built to run OpenAi models. It was built for this. The current situation. Mai-1 does not compete in gross abilities. But it has advantages that Openai can never offer: Microsoft completely controls development. They can optimize it specifically for Windows, Office and Azure without asking anyone permission. They can adjust costs, latency and capabilities according to their exact needs. The voice is important. Mai-Voice-1 generates a minute of audio in less than a second with a single GPU. They do not need to be the best in text processing if they dominate The interface they believe of the future: The voice. Yes, but. A model in the 13th position remains a model in the 13th position. Business users who pay thousands of dollars for co -pilot surely expect the best, not “good enough.” Microsoft knows it and that’s why they are not replacing GPT-5 immediately. Mai-1 is gradually introduced in “specific use cases” while improving. This is its 1.0 version. GPT-5 is the fifth great iteration of OpenAI. They have room to grow. The decisive moment. The true test will come when Microsoft has to choose: renew the agreement with OpenAi or bet on their own models? With Mai-1, Microsoft has shown that it has a viable alternative. It does not need to be better than GPT-4. You just need to be good enough for the 250,000 million annual revenues of Microsoft not depending on Sam Altman’s whims. In a negotiation, the best position is to be able to get up from the table. Microsoft has just bought the chair. In Xataka | China’s self -sufficiency test in chips for AI is already here: it has not bought Nvidia or a single H20 GPU in the last quarter Outstanding image | Microsoft
