Microsoft AI Chief Says Company Is Now Free to Pursue Superintelligence Independently of OpenAI

At Microsoft's Build 2026 developer conference in early June, Mustafa Suleyman made a statement that would have been unthinkable two years ago. "We were only sort of set free from our contract with OpenAI about six months ago to formally pursue superintelligence," said the CEO of Microsoft AI [1]. The remark, delivered with studied casualness, announced the end of a contractual arrangement that had kept the world's most valuable company tethered to a startup for its most ambitious AI ambitions.

The original 2019 partnership agreement between Microsoft and OpenAI contained a clause that prohibited Microsoft from independently pursuing artificial general intelligence (AGI) or superintelligence — the hypothetical point at which AI systems surpass human cognitive abilities across all domains [1]. That restriction shaped Microsoft's AI strategy for half a decade. Now, with a renegotiated contract signed in September 2025 and a new team already shipping models, Microsoft is signaling that it intends to build the future of AI on its own terms.

The Contract That Bound a Giant

The restriction embedded in the 2019 deal was unusual by any standard of corporate partnerships. Microsoft committed roughly $13 billion to OpenAI between 2019 and 2023, providing both direct capital and Azure compute credits [5]. In exchange, Microsoft received exclusive rights to commercialize OpenAI's models through Azure and its consumer products. But the agreement also carved out AGI as OpenAI's exclusive domain — a provision that effectively barred Microsoft from competing at the frontier of AI research [2].

That arrangement made strategic sense when OpenAI's models were years ahead of anything Microsoft could build internally. By the time GPT-4 launched in March 2023, however, the competitive landscape had shifted. Google DeepMind, Meta AI, Anthropic, and xAI were all producing frontier models, and Microsoft's own research division — historically one of the most productive in the industry — was constrained from joining the race at its highest level.

The renegotiation that freed Microsoft came in two phases. In October 2025, both companies signed a new definitive agreement that restructured the partnership. Microsoft's IP rights were extended through 2032 and expanded to include models developed after AGI is achieved — a change that removed OpenAI's previous ability to cut off Microsoft's access once AGI was declared [2]. Then in April 2026, a further amendment ended the Azure exclusivity arrangement, allowing OpenAI to serve its products across any cloud provider while capping revenue share payments from OpenAI to Microsoft through 2030 [6]. Microsoft retained only a four-month first-mover window on new frontier model releases [6].

The $13 Billion Bet and Its Returns

Microsoft's total investment in OpenAI stands at approximately $13 billion, of which $11.8 billion had been funded as of March 31, 2026 [7]. The financial returns have been extraordinary. Following OpenAI's February 2026 funding round — $110 billion at a $730 billion pre-money valuation, expanding to $122 billion at an $852 billion post-money valuation — Microsoft's 26.79% stake on a fully diluted, as-converted basis is now valued at roughly $228 billion [7]. That represents a 17.6x return on invested capital.

Source: SEC filings, media reports

Data as of Jun 1, 2026CSV

Microsoft also reported $7.6 billion in gains from OpenAI in a single quarter [8]. The financial relationship remains deeply intertwined: OpenAI has contracted to purchase $250 billion of Azure services over the life of the partnership [2].

Competing directly with OpenAI while holding a 27% stake in the company creates obvious tensions. But Microsoft's leadership appears to view the arrangement less as a contradiction than as a hedge. If OpenAI wins the race to superintelligence, Microsoft profits through its equity stake. If Microsoft's own MAI division gets there first, the company captures the full value internally. The financial structure ensures Microsoft benefits in either scenario — a position no other company in the AI race currently holds.

Building the Machine: Microsoft's Independent Capabilities

The MAI Superintelligence Team was formally established in November 2025 under Suleyman's leadership, with Karén Simonyan — who joined Microsoft alongside Suleyman from Inflection AI in March 2024 — serving as chief scientist [9]. The team's stated mission is to develop "humanist superintelligence" focused on three application areas: an AI companion for personal productivity and learning, healthcare diagnostics, and scientific breakthroughs in areas like clean energy [9].

At Build 2026, the team unveiled seven new in-house models under the MAI brand. The headline release was MAI-Thinking-1, Microsoft's first reasoning model — a 35-billion-parameter system with a 256K context window, trained from scratch on commercially licensed data with no distillation from OpenAI or any other third-party model family [3]. The remaining models cover speech, multimodal processing, and specialized inference workloads [3].

Suleyman acknowledged at Build that the superintelligence team is still maturing. "It's still going to be a good year or two before the superintelligence team is producing frontier models," he told Axios in November 2025 [10]. But the pace of investment suggests Microsoft is not treating this as a long-term research project. Capital expenditures reached $37.5 billion in a single quarter (Q2 FY26, ending December 31, 2025), roughly two-thirds of which went to GPUs and CPUs [3]. For full fiscal year 2026, analysts project total capex between $97.7 billion and $120 billion [3].

Source: Company earnings reports

Data as of Jun 1, 2026CSV

The Talent War

Microsoft's ability to pursue superintelligence independently hinges on whether it can assemble a research team capable of competing with OpenAI, Google DeepMind, and Anthropic. The global pool of frontier AI researchers is estimated at fewer than 5,000 individuals, making recruitment a zero-sum competition [4].

Microsoft has been aggressive. In mid-2025, the company hired approximately 20 engineers from Google DeepMind, including Amar Subramanya, a 16-year Google veteran who served as vice president of engineering on the Gemini assistant, and Adam Sadovsky, a distinguished software engineer with nearly 18 years at Google [4]. The MAI Superintelligence team has drawn researchers from Google DeepMind, Meta FAIR, and academic institutions with compensation packages that reportedly include guaranteed equity grants worth $2 million or more over four years [4].

The recruitment pipeline extends beyond poaching. Ali Farhadi, formerly of the Allen Institute for AI, is among the prominent hires [4]. Every researcher who joins Microsoft's superintelligence effort is, by definition, one fewer available to OpenAI, Google, or Anthropic — a dynamic that Microsoft's compensation budget, backed by a $3 trillion market capitalization, is well-positioned to exploit.

A Negotiating Tactic or a Genuine Pivot?

The steelman case that Microsoft's independence declaration is primarily a negotiating tactic rests on several pieces of evidence.

First, the timing aligns with regulatory pressure. The FTC opened a formal antitrust investigation into the Microsoft-OpenAI relationship, sending Microsoft a civil investigative demand [11]. The European Commission and the UK's Competition and Markets Authority (CMA) have both been scrutinizing the partnership as well [11]. The CMA's initial report, released in late January 2025, found that both Microsoft and Amazon's AWS had been "generating sustained returns from their cloud services substantially above their cost of capital" [11]. By publicly distancing itself from OpenAI, Microsoft weakens the core premise of these investigations — that the partnership constitutes a de facto merger or anticompetitive arrangement.

Second, internal history suggests Microsoft's original motivation for partnering with OpenAI was as much defensive as it was offensive. Court filings in the Musk v. Altman case revealed that Microsoft executives in 2018 harbored serious doubts about OpenAI's technical direction and business viability. One executive wrote that backing OpenAI was "less about what they've built and more about what we prevent Amazon from accessing" [12].

Third, the financial entanglement remains deep. With a $228 billion stake, $250 billion in committed Azure purchases, and ongoing revenue sharing through 2030, the two companies are far from independent in any meaningful economic sense [2][7].

Against this reading, the concrete actions Microsoft has taken — forming a dedicated superintelligence team, hiring dozens of frontier researchers, shipping seven in-house models trained without OpenAI data, and spending nearly $100 billion annually on AI infrastructure — are difficult to dismiss as mere positioning. These represent real resource commitments that make sense only if Microsoft genuinely intends to compete at the frontier.

The Competitive Landscape

Microsoft is entering a crowded field. Google has committed over $75 billion in AI infrastructure investment for 2026 [13]. Meta established its own Superintelligence Labs and projects achieving AGI by 2027, followed by superintelligent capabilities by 2029 [13]. Meta also announced a multiyear deal to purchase up to $100 billion worth of AMD chips to support its AI infrastructure [13]. Elon Musk's xAI, despite reporting $2.4 billion in losses in Q1 2026 while spending $7.7 billion on capital expenditures, has positioned itself as a major player — and signed a deal renting 300 megawatts of compute capacity to Anthropic for $1.25 billion per month [13].

Academic research output on superintelligence has tracked the industry's growing ambitions. According to OpenAlex data, publications on superintelligence peaked at 2,002 papers in 2025, up from just 87 in 2015 [14].

Source: OpenAlex

Data as of Jan 1, 2026CSV

Timeline predictions from industry leaders vary widely. Anthropic CEO Dario Amodei has suggested human-level AI could arrive by 2026 or 2027. Shane Legg, chief AGI scientist at Google DeepMind, assigns roughly 50% odds to "minimal AGI" by 2028 [13]. Suleyman himself told Fortune that he expects AI to automate all white-collar work within 18 months [15].

Regulatory Implications

The Microsoft-OpenAI relationship is under active investigation by regulators on three continents. The FTC's probe focuses on whether Microsoft's investment in OpenAI and the accompanying exclusivity arrangements constitute anticompetitive behavior [11]. The European Commission has been examining similar concerns. The CMA's cloud computing market investigation extends beyond the AI partnership itself to broader questions about market concentration in cloud infrastructure [11].

Microsoft's "independence" framing serves a clear regulatory purpose. By ending Azure exclusivity and publicly declaring its freedom to compete with OpenAI, Microsoft can argue that the partnership does not confer the kind of market control that would trigger antitrust intervention. The April 2026 amendment, which allows OpenAI to serve customers across any cloud provider, directly addresses one of the FTC's central concerns [6][12].

Whether regulators accept this framing remains to be seen. The financial ties between the companies — a 27% equity stake, hundreds of billions in committed cloud purchases, ongoing revenue sharing — suggest a level of interdependence that formal contract terms alone cannot dissolve.

Safety and Governance

Suleyman has articulated a safety philosophy that distinguishes between two concepts: containment (establishing boundaries on AI systems) and alignment (ensuring AI shares human values). "Containment has to come first," he has said. "Otherwise, alignment is the equivalent of asking nicely" [16]. Microsoft's superintelligence initiative focuses on domain-specific systems that are "carefully calibrated, contextualised, within limits" rather than unbounded autonomous entities [16].

Both Microsoft and OpenAI were among the original seven companies that signed the White House's voluntary AI safety commitments in July 2023, pledging to conduct internal and external security testing before releasing AI systems, share safety information across the industry, invest in cybersecurity protections, and develop technical mechanisms like watermarking for AI-generated content [17]. Both companies subsequently co-founded the Frontier Model Forum, a nonprofit aimed at facilitating industry-wide AI safety discussions [17].

OpenAI's charter includes a commitment to ensuring AGI "benefits all of humanity" and provisions for halting competitive efforts if a safety-aligned project nears AGI first. Microsoft has made no equivalent public commitment for its independent superintelligence work, though Suleyman has stated publicly that he would "halt AI work if it imperils humanity" [18]. The gap between voluntary pledges and enforceable governance frameworks remains significant — a gap that neither company has yet moved to close with binding commitments.

What Comes Next

The Microsoft-OpenAI relationship has entered a new phase defined by managed competition. Microsoft retains a massive financial interest in OpenAI's success while simultaneously building the organizational capacity to render that partnership optional. OpenAI, freed from Azure exclusivity, gains the ability to diversify its cloud infrastructure and customer relationships.

For the broader AI industry, Microsoft's entry as an independent superintelligence competitor adds another well-capitalized player to a race that already includes Google, Meta, Anthropic, and xAI. The concentration of AI research among a handful of technology companies — each spending tens of billions annually on compute infrastructure alone — raises questions about whether meaningful safety governance can keep pace with the speed of deployment.

The contract clause that once prevented Microsoft from competing at the frontier of AI is now a historical footnote. What Microsoft builds with its newfound freedom will determine whether Suleyman's "set free" rhetoric was a statement of strategic intent or the opening move in a negotiation that is far from over.