Microsoft is preparing to reduce its reliance on OpenAI as it accelerates plans to build its own frontier artificial intelligence models. The shift signals a major strategic change for the USD$3tn software giant and deepens uncertainty around OpenAI’s financial future.
Mustafa Suleyman, Microsoft’s AI chief and a co-founder of Google DeepMind, confirmed the move in a recent interview. He said the company must create its own cutting-edge foundation models using vast computing power and elite training teams.
Microsoft has relied heavily on OpenAI’s systems, including ChatGPT and DALL·E 3, to power Copilot tools across its product suite. Those tools support enterprise software such as Microsoft 365 and GitHub Copilot.
However, the companies reworked key parts of their partnership last October. The new structure gives OpenAI greater freedom to seek outside computing partners and investors. Consequently, Microsoft can now limit some of its long-term exposure.
Microsoft Corp. (NASDAQ: MSFT) remains one of OpenAI’s largest backers. It owns roughly 27 per cent of OpenAI’s for-profit arm and holds intellectual property rights to its models until 2032. Additionally, Microsoft retained access to OpenAI’s most advanced systems under the revised agreement. At the same time, OpenAI gained flexibility to raise capital and secure infrastructure from rival cloud providers.
Suleyman said Microsoft is pursuing what he described as “true self-sufficiency” in AI. He explained that the company is assembling massive datasets and reorganizing internal teams to support that goal. Furthermore, he suggested Microsoft’s most advanced in-house models could launch later this year. The company has already invested billions into expanding its data centers and compute capacity.
Read more: NuScale taps DOE-backed AI research to optimize fuel across 12-reactor SMR plant
Read more: Sidus Space partners with Simera Sense to enable real-time satellite data processing
OpenAI has faced considerable financial challenges
Microsoft expects to spend roughly USD$140bn in capital expenditures during its fiscal year ending in June. Much of that investment targets AI infrastructure, including servers, networking equipment, and energy supply.
However, investor concerns about an AI spending bubble have pressured major technology stocks. Microsoft shares have fallen more than 13 per cent over the past month. OpenAI faces its own financial challenges. The company has committed to enormous future computing contracts that some estimates place above USD$1tn over time.
Additionally, OpenAI has not yet reported sustained profits. It relies heavily on continued capital injections from large investors to fund operations and model training. Amazon.com Inc. (NASDAQ: AMZN) and Japan’s SoftBank Group have also supported OpenAI through partnerships and financing arrangements. Meanwhile, those companies continue building their own AI capabilities.
Microsoft has diversified its AI bets beyond OpenAI. It has invested in Anthropic and Mistral while accelerating its internal research programs. However, competition in enterprise AI remains intense. Anthropic has gained traction in AI-powered coding tools, challenging Microsoft’s GitHub Copilot dominance.
Alphabet Inc. (NASDAQ: GOOGL), Google’s parent company, is also pushing aggressively into corporate AI services. Additionally, OpenAI itself now courts large enterprise customers more directly.
Suleyman predicted that white-collar work could change dramatically within the next 12 to 18 months. He said many computer-based tasks performed by lawyers, accountants, and project managers could become automated.
Furthermore, he argued that AI agents will coordinate workflows across large organizations within two to three years. These systems, he said, will learn from experience and gradually act with greater autonomy.
Read more: Anthropic AI safety lead resigns, warns of ethical strain and world “in peril”
Read more: Global AI safety review flags deepfakes, job risks, and growing emotional reliance
Microsoft must balance its existing agreements
He compared building custom AI models to creating digital content. In the near future, he suggested, organizations could design AI systems tailored to their specific needs. Consequently, Microsoft wants to control its core technology stack rather than depend entirely on an external partner. Internal models would reduce strategic risk and strengthen negotiating power.
However, Microsoft must balance independence with its existing agreements. The company still benefits from privileged access to OpenAI’s latest breakthroughs. Additionally, OpenAI’s brand and developer ecosystem remain powerful assets. Millions of users and businesses rely on ChatGPT and related tools. The revised partnership allows OpenAI to pursue other infrastructure providers. Conversely, it positions Microsoft to step back if OpenAI’s risk profile worsens.
Tensions between the two companies have surfaced before. Leadership disputes and governance controversies have periodically strained the relationship. Meanwhile, the broader AI industry continues to expand at a rapid pace. Corporations are racing to deploy generative AI tools across customer service, coding, marketing, and internal operations.
Microsoft has reported strong enterprise demand for Copilot products. However, its consumer AI initiatives have delivered more mixed results. Additionally, the company must justify its heavy capital spending to shareholders. Large data centers require enormous electricity supplies and long-term contracts.
Suleyman framed the strategy as necessary for long-term competitiveness. He argued that frontier AI requires gigawatt-scale computing and top-tier research teams. Consequently, Microsoft appears determined to own more of that capability directly. The move could reshape the balance of power between Big Tech and the world’s most prominent AI start-up.
