Microsoft superintelligence ambitions took a concrete step forward on March 23, 2026, when the company confirmed it had hired Ali Farhadi, former CEO of the Allen Institute for AI (AI2), as a corporate vice president reporting directly to Mustafa Suleyman, Microsoft’s AI CEO. The hire caps a broader talent sweep that also pulled several key AI2 researchers into Microsoft’s orbit, signaling that Suleyman’s push to build proprietary frontier models is no longer a vague aspiration — it’s a staffed operation.
Key Takeaways
- Ali Farhadi joined Microsoft as corporate vice president under Mustafa Suleyman after stepping down as AI2 CEO on March 12, 2026.
- Microsoft formed an in-house AI team in November 2025 to train frontier models using its own data and compute.
- Farhadi previously co-founded Xnor.ai, acquired by Apple in 2020 for an estimated $200 million.
- Additional AI2 and University of Washington researchers joined Suleyman’s team, including Hanna Hajishirzi and Ranjay Krishna.
- Microsoft’s stated goal is “humanist superintelligence” — safer, more controllable AI systems built for humanity’s hardest problems.
Why Microsoft Is Building Its Own Frontier Models
Microsoft formed a dedicated in-house AI research team in November 2025, with Suleyman describing its mission as training “frontier models of all scales with our own data and compute at the state-of-the-art level” to make Microsoft “self-sufficient in AI”. That’s a significant strategic declaration from a company that has spent years as OpenAI’s most prominent backer and distribution partner. Building in-house capability isn’t just a hedge — it’s a bid for independence.
A recent internal restructure made the direction even clearer. An employee email stated that Suleyman would focus all of his “energy on our Superintelligence efforts and be able to deliver world class models for Microsoft over the next five years”. Five years is a long runway, but the pace of hiring suggests Microsoft isn’t waiting around to staff up.
Who Is Ali Farhadi and Why Does Microsoft Want Him
Ali Farhadi is one of the more credible hires Microsoft could have made for this role. He led AI2 for more than two and a half years after taking the CEO position in July 2023, steering a nonprofit that has historically punched above its weight on open AI research. Before that, he co-founded Xnor.ai, an on-device AI model startup that Apple acquired in 2020 for an estimated $200 million, after which Farhadi led machine learning efforts at Apple. That’s a career arc that spans open-access research, efficient model deployment, and big-tech infrastructure — exactly the profile you’d want running a frontier model team.
AI2 board chair Bill Hilf told GeekWire that Farhadi wanted to pursue research at the “extreme frontier of AI, where for-profit companies are spending billions on training the most advanced models”. That’s a candid acknowledgment that nonprofit research budgets, however well-intentioned, can’t match what Microsoft, Google, or Meta are deploying. Farhadi’s move is less a betrayal of open AI values and more a pragmatic recognition of where the most consequential work is now happening.
Microsoft Superintelligence Strategy: The Full Team Taking Shape
Microsoft’s superintelligence team isn’t just Farhadi. The company also brought in Hanna Hajishirzi, Ranjay Krishna — who led AI2’s Molmo multimodal models — and Sophie Lebrecht, former AI2 COO. Hajishirzi, Krishna, and others are expected to retain their University of Washington faculty positions, which means Microsoft gains serious research credibility without fully pulling these figures out of academia. It’s a smart arrangement, and one that top AI labs have used before to attract researchers who value institutional affiliation.
Microsoft described the collective mission as pursuing “humanist superintelligence: safer, controllable, more capable AI systems in service of humanity and our toughest problems”. That framing is deliberate — it positions Microsoft’s superintelligence work as distinct from raw capability racing, though whether the substance lives up to the branding remains to be seen.
How This Fits Into the Broader AI Talent War
The AI talent market is moving at a pace that makes traditional recruiting look quaint. On the same day Microsoft’s Farhadi hire became public, Meta confirmed it had acquired the entire founding team of AI startup Dreamer, which had been building AI agent tools. These aren’t isolated moves — they reflect a systematic effort by the largest tech companies to consolidate the pool of researchers capable of working at the frontier.
For AI2, the departures sting. The institute built a reputation as a place where researchers could do serious work without the commercial pressures of for-profit labs. Interim CEO Peter Clark stated that AI2 remains committed to its mission and its partnerships, including with the NSF and Nvidia’s OMAI initiative. That’s the right thing to say, and AI2’s track record gives it credibility. But losing a CEO, a COO, and multiple senior researchers in a short window is a structural challenge that goodwill alone won’t solve.
Is Microsoft’s superintelligence push a threat to OpenAI?
Microsoft building its own frontier models inevitably raises questions about its relationship with OpenAI, which it has backed heavily. The two companies have not publicly addressed any tension, but Microsoft’s stated goal of AI self-sufficiency speaks for itself. Having your own capable models reduces dependence on any single external partner — that’s basic strategic logic, not a declaration of rivalry.
What happened to AI2 after Farhadi left?
Farhadi stepped down as AI2 CEO on March 12, 2026, with Peter Clark stepping in as interim CEO. The institute stated it remains committed to open AI research and its existing partnerships. AI2 has a strong institutional foundation, but the departure of multiple senior figures in quick succession will test its ability to maintain research momentum.
What does Ali Farhadi’s role at Microsoft involve?
Farhadi joined as corporate vice president, reporting to Mustafa Suleyman, and is already listed in Microsoft’s internal systems with that title. His role sits within Suleyman’s in-house AI team, which is focused on training frontier models using Microsoft’s own data and compute infrastructure.
Microsoft’s superintelligence play is now more than a memo and a mission statement — it’s a team. Whether Suleyman, Farhadi, and their growing roster of AI2 and UW alumni can actually close the gap with the most capable models in the world is the real question. The talent is credible. The ambition is clear. The next five years will show whether the execution matches both.
Edited by the All Things Geek team.
Source: Windows Central
