Microsoft’s AI Journey: A Balancing Act
Microsoft has made it abundantly clear: they’re “all in” on artificial intelligence. However, the company’s pursuit of self-sufficiency in the AI realm, particularly the development of homegrown models, has been a longer process than anticipated. While Microsoft continues to lean heavily on its partner, OpenAI, recent signs suggest a potential shift in strategy and a push towards greater independence.
Reports indicate that Microsoft has completed the training of a new family of generative AI models, internally referred to as MAI. These models have reportedly achieved performance “nearly as well” as those developed by OpenAI and Anthropic, according to The Information. Furthermore, the company is working on reasoning models designed to compete with models from OpenAI, DeepSeek, and Alibaba. These reasoning models are critical for simulating human cognitive abilities like logical reasoning, decision-making, and complex problem-solving, often employing techniques such as chain-of-thought prompting to improve performance.
Reportedly, a point of contention between Microsoft and OpenAI was the latter’s alleged reluctance to share technical details about the chain-of-thought process, a key component in the development and performance of reasoning models.
Delays and Departures
Almost a year ago, the news of Microsoft’s internally developed MAI models sparked speculation that they would be released at the 2024 Microsoft Build conference. This, however, did not come to pass. According to reports, the delay was due to “technical setbacks, abrupt changes in strategy, and the departures of top talent who disagreed with [Microsoft AI CEO] Suleyman’s management and technical approach.” During this period of internal struggle, OpenAI continued to advance its models, releasing previews of their o3 and GPT-4.5 models.
A Competitive API on the Horizon?
Recent reports suggest Microsoft may be preparing to release the MAI models later this year as an application programming interface (API). This move would position Microsoft in direct competition with APIs from both its rivals and partners, including OpenAI. Moreover, Microsoft is already experimenting with integrating MAI models into its Copilot platform, potentially substituting OpenAI’s models. Furthermore, testing of models from Anthropic, xAI, DeepSeek, and Meta within Copilot is also underway.
A Partnership’s Evolving Landscape
A potential point of strain in the OpenAI-Microsoft relationship appeared with the Stargate Project, an initiative where OpenAI, in collaboration with Oracle and Nvidia, will build and operate considerable AI infrastructure. While OpenAI stated that the project will build off its current partnership with Microsoft, the move has caused some analysts to suggest Microsoft adjusted some of its expansion plans, canceling leases on new data centers despite maintaining its expected capital expenditure.
Diverging Priorities
Industry analysts provide insights into the dynamic between the two companies. Jason Andersen, vice president and principal analyst at Moor Insights & Strategy, highlights that OpenAI and Microsoft have distinct priorities. “OpenAI focuses on delivering the best models in service of the mission of achieving AGI [artificial general intelligence]. There are not many (or any) multi-billion dollar revenue engines with that level of single-mindedness. I believe that Microsoft is dealing with a wide range of AI priorities and that’s precisely why [Microsoft AI CEO] Suleyman was brought in to balance these efforts,” stated Anderson. “This range of priorities is also why we have seen Microsoft partner with other LLM [large language model] companies despite having its own products and the OpenAI partnership. Shifting to a multi-model mindset is significant and reinforces the point that right now model pure plays will have an innovation advantage in areas where they focus.”
Justin St-Maurice, technical counselor at Info-Tech Research Group, expands on the technological differences. “Instead of operating through a series of functions with predictable inputs and outputs, LLMs accept any arbitrary input. They behave chaotically, with natural system properties,” he observed. “Operating this new type of technology has been OpenAI’s bread and butter since day one, whereas Microsoft has a legacy of thinking, processing, and designing that are fixed in an old paradigm. That said, Microsoft Research has been working on large-scale AI models for years now, so while OpenAI has had a head start on embracing the new paradigm, Microsoft isn’t starting from zero.” St-Maurice further noted that OpenAI’s edge comes from a broader set of innovations like improvements in reinforcement learning and customer architectures.
Can Microsoft Catch Up?
While Microsoft has encountered setbacks, analysts do not discount the company’s long-term chances of achieving its goals. As Andersen puts it, whether Microsoft will catch up “will ultimately come down to balancing across priorities, including innovating in areas where OpenAI is not focused, such as developer tooling, application integration and management, and governance.” St-Maurice adds, “Betting against the turtle in favor of the hare was never a great bet. Microsoft will endure, iterate, and optimize in mature labs with world-class engineers. If OpenAI isn’t interested in sharing all of its underlying IP, the open-source market is providing more than enough output to ensure it doesn’t really matter.”