At its annual Build developer conference in San Francisco, Microsoft unveiled seven proprietary AI models, marking a decisive move to lessen its reliance on the AI startups it has heavily funded. The announcement reflects a broader trend among major tech firms to develop internal capabilities rather than solely licensing third-party technology.
Mustafa Suleyman, chief executive of Microsoft AI, stated that after fine-tuning the models for consulting firm McKinsey, the company achieved superior quality compared to OpenAI's GPT-5.5, with an estimated tenfold improvement in cost efficiency based on public pricing data across model sizes.
“We believe the time has come for every company to move from consuming a frontier model to fully participating at the frontier,” said chief executive Satya Nadella at the conference.
New Reasoning and Coding Models
The centerpiece is MAI-Thinking-1, Microsoft's first reasoning model, trained from scratch on clean, commercially licensed data without distillation from third-party systems. With 35 billion active parameters and a 256,000-token context window, it is designed for complex multi-step instructions, long-context reasoning, and code generation.
Alongside it, the company launched MAI-Code-1-Flash, a coding model that converts text descriptions into source code for applications and websites, now rolling out across GitHub Copilot and Visual Studio Code. By running these models on Azure infrastructure, Microsoft can bypass fees paid to third-party providers and pass savings to developers.
In blind evaluations by Surge, an independent human rating partner, MAI-Thinking-1 was preferred over Anthropic's Claude Sonnet 4.6 and matched Claude Opus 4.6 on coding benchmarks.
Quantum Computing Milestone
Microsoft also announced that its Majorana 2 quantum chip is 1,000 times more reliable than its predecessor, bringing it closer to a commercially viable quantum computer. Qubits on the new chip survive for an average of 20 seconds, compared to milliseconds on the original—an improvement likened to upgrading from a phone needing daily charging to one lasting years.
“We will have a quantum machine in 2029 that can solve commercially viable, reasonable problems,” said Zulfi Alam, corporate vice president of Microsoft Quantum. The chip currently has only 12 qubits, while a useful machine would require millions. Microsoft's approach uses topological qubits based on a quasi-particle first theorized in the 1930s by Italian physicist Ettore Majorana. The company retracted a 2018 Nature paper claiming evidence for the particle but has pressed on, replacing aluminum with lead as a superconductor in the second-generation chip. The chip and its research have not yet been peer-reviewed, and some physicists have called for more information.
IPO Race and European Implications
Microsoft's push for model independence comes as its investees prepare for blockbuster IPOs. Anthropic filed confidentially for an IPO on 1 June, just days after raising $65 billion (€59bn) in a Series H round that pushed its valuation to $965 billion (€877bn). OpenAI is also readying its own confidential filing. Microsoft has committed $13 billion (€11.8bn) to OpenAI and up to $5 billion (€4.5bn) to Anthropic, while making both companies' models available through Azure.
For European developers and businesses, Microsoft's move could reduce costs and increase competition in the AI model market, potentially benefiting startups in Berlin, Paris, and Stockholm. However, it also raises questions about data sovereignty and regulatory oversight, especially as the European Commission scrutinizes AI investments. MEPs have previously accused the EU Commission of copying Microsoft lobbying into data centre law, highlighting tensions between Big Tech and European regulators.
As Microsoft deepens its AI capabilities, the landscape for European tech firms may shift, with opportunities for collaboration or competition. The company's quantum ambitions also align with EU efforts to develop quantum computing infrastructure, though the timeline remains uncertain.
