Microsoft launches Microsoft Research AI lab in Redmond, involving 100+ AI scientists, focused on developing general purpose AI, challenging Google's DeepMind
Jeremy Kahn / Bloomberg :
Context & Ripple Effects
Microsoft is formalizing what had been an ad hoc effort: after partnering with OpenAI on Azure the year before, it now stands up its own general-purpose AI lab in Redmond with 100+ scientists, explicitly benchmarked against Google's DeepMind. The move answers an internal critique that would resurface years later, when Nadella voiced concern that Microsoft was slow to commercialize its own research.
The lab gives Microsoft a second track on general-purpose AI — an internal bench alongside its external OpenAI bet — and sets up the in-house-versus-partner tension that defines its AI strategy through the decade.
First-order effects
- Google's DeepMind gains a named corporate rival inside Microsoft, with 100+ scientists dedicated to the same general-purpose AI goal rather than scattered product research.
- Microsoft's researchers get an institutional home for general-purpose AI work, reducing dependence on the OpenAI partnership as the company's only frontier capability.
Second-order effects
- Competition for AI scientists intensifies between Redmond and DeepMind, pushing compensation and lab autonomy upward across big-tech research organizations.
- Azure's role as compute provider to AI labs — established with OpenAI — becomes more strategically valuable as Microsoft now runs both a customer lab and its own on the same cloud.
Third-order effects
- If the dual-track pattern holds, major cloud vendors converge on owning both an internal frontier lab and equity/compute ties to independent ones — a structure visible later in Microsoft's advanced AI research team absorbing OpenAI staff during the 2023 leadership crisis and its ~500B-parameter in-house model effort.
The trend: Big-tech incumbents are responding to DeepMind-style centralized AI labs by building their own general-purpose research organizations alongside external partnerships, making frontier AI capability a two-track corporate asset.