Microsoft has hired Ali Farhadi, the former CEO of the Allen Institute for AI (Ai2), along with researchers Hanna Hajishirzi and Ranjay Krishna and former Ai2 COO Sophie Lebrecht. The group joins Mustafa Suleyman's Superintelligence team, Microsoft's internal effort to build frontier AI capabilities in-house.
The hire is a significant talent acquisition and a clear strategic signal: Microsoft is investing in the ability to develop world-class AI models independently of OpenAI.
Who Microsoft Just Hired
Ali Farhadi ran Ai2 for years and built it into one of the most credible open AI research organizations in the world. Under his leadership, Ai2 published influential work on vision-language models, reasoning systems, and open-weight model development — including the OLMo family of models, which demonstrated that open research labs could produce competitive language models without trillion-dollar training budgets.
Hanna Hajishirzi is one of the leading researchers in natural language processing and has been a key contributor to open model development. Ranjay Krishna is known for work on vision and multimodal AI. These aren't just credentialed hires — they're people who have built the specific capabilities Microsoft wants: open-weight, efficient, and competitive models.
The Suleyman Context
Mustafa Suleyman joined Microsoft in 2024 after Inflection AI, the company he co-founded, effectively sold its team and intellectual property to Microsoft. He's been building out Microsoft's internal AI organization with the mandate of creating capabilities the company doesn't have to source entirely from OpenAI.
The Farhadi group joins a team that already includes significant internal AI research capacity but has lacked the specific expertise in open-source frontier model development that Ai2 was known for. The combination of Suleyman's commercial AI experience and Farhadi's research pedigree creates a credible internal lab that Microsoft didn't have before.
What This Means for OpenAI
The OpenAI-Microsoft relationship is the most consequential partnership in the AI industry. Microsoft has invested roughly $13 billion in OpenAI, and Azure is the infrastructure backbone on which OpenAI's products run. The relationship is financially and technically deep.
But it's also complex. Microsoft's ability to deploy OpenAI models in its products depends on an agreement that includes clauses around AGI thresholds, profit caps, and commercial use rights. As that agreement approaches its natural renegotiation points — and as OpenAI continues its transition from nonprofit to capped-profit to for-profit — Microsoft's leverage in those negotiations depends partly on having credible alternatives.
Hiring Ali Farhadi and building a serious internal AI team doesn't threaten the OpenAI relationship in the short term. But it gives Microsoft options. That's exactly what a sophisticated negotiating position requires.
The Open-Source Angle
One subtext of this hire is open-source model development. Ai2 was one of the few major AI organizations genuinely committed to open-weight model releases — making the actual model weights publicly available, not just APIs. Microsoft has historically relied on OpenAI's closed models for its flagship products.
With Farhadi and his Ai2 colleagues in-house, Microsoft now has the people who know how to build and release open-weight models. Whether it will pursue that strategy is an open question. But the capability — and the institutional knowledge — is now inside the building.
