Staying Off the Frontier to Save Costs and Optimize Utility
Mustafa Suleyman, the newly appointed CEO of Microsoft AI, has shed light on the tech giant’s deliberate decision not to chase the absolute frontier of AI model development. In an interview with CNBC, Mustafa Suleyman explained that Microsoft prefers to build models that are “three or six months behind” the cutting-edge offerings in the industry. The rationale? Cost savings, reduced redundancy, and a sharper focus on usability.
“It’s cheaper to give a specific answer once you’ve waited for the first three or six months for the frontier to go first,” Mustafa Suleyman said. “We call that off-frontier. That’s actually our strategy — to really play a very tight second, given the capital-intensiveness of these models.”
A Shift from the Race for First to Strategic Efficiency
Mustafa Suleyman, co-founder of DeepMind and former CEO of Inflection, joined Microsoft in 2024 along with several Inflection employees. His leadership signals a pragmatic turn in Microsoft’s AI roadmap. Rather than competing to build the most advanced models immediately, the company focuses on creating refined, cost-effective AI systems that can be tailored to specific use cases.
This approach allows Microsoft to leverage insights from early entrants in the AI space and build systems that meet user needs without incurring the steep costs of developing state-of-the-art models from scratch. “Maybe we don’t develop the absolute frontier, the best model in the world first,” Suleyman acknowledged. “That’s very, very expensive to do and unnecessary to cause that duplication.”
OpenAI Partnership Remains Strong — For Now
Microsoft’s AI strategy is deeply intertwined with its ongoing partnership with OpenAI. Since investing $13.75 billion in the startup, Microsoft has integrated OpenAI’s models across products like Bing, Windows, and its Copilot assistant. Suleyman emphasized the strength of the partnership, even as some cracks have begun to emerge.
“It’s absolutely mission-critical that long-term, we are able to do AI self-sufficiently at Microsoft,” Suleyman said. “But until 2030 at least, we are deeply partnered with OpenAI, who have [had an] enormously successful relationship for us.”
Despite OpenAI’s recent engagement with rival cloud provider Oracle on the $500 billion Stargate project, Microsoft reaffirmed in a blog post that OpenAI had “recently made a new, large Azure commitment.”
Copilot Gets Smarter With “Memory” Feature
At Microsoft’s 50th anniversary event held at its Redmond, Washington headquarters, Suleyman also announced new updates to Copilot, the company’s AI-powered assistant. One major addition is the introduction of a “memory” feature, which allows Copilot to retain key facts about users over time — a capability already available in OpenAI’s ChatGPT.
This enhancement reflects Microsoft’s broader push to bring more personalized, helpful interactions to its AI ecosystem, without necessarily being first to launch new features.
Focus on Practical, Scalable AI
Microsoft continues to invest in open-source small language models that run efficiently on PCs, bypassing the need for costly server-grade GPUs. These lightweight models serve a different purpose from the more complex systems developed by OpenAI, but they align with Mustafa Suleyman’s vision of targeted, efficient AI deployment.
In a rapidly evolving AI landscape, Suleyman’s approach suggests that winning the AI race isn’t just about being first — it’s about being smart, sustainable, and strategic.
Get more such blog on The Enterprise Globe