Arms Race: Microsoft to Create Its Own Ultra-Large AI Models by 2027

The vector for separation is confirmed by insiders. Simultaneously with the release of the lightweight MAI models, on April 3 it became known about Microsoft’s strategic decision to develop its own "frontier" AI models (cutting-edge ultra-large networks) by 2027.

According to Bloomberg, Satya Nadella has tasked engineering teams with creating a foundation that will not yield in performance to GPT-Next or Claude Opus. This is a logical consequence of massive infrastructural investments (recall the $15.5B poured into Singapore and Japan over the last 3 days). If you have the best server capacities in the world, giving all the intellectual margin to a third-party startup is unprofitable. The race for AGI (Artificial General Intelligence) is ceasing to be the domain of labs and is moving into the workshops of trillion-dollar corporations.

Source: Bloomberg / AFR