Why is Microsoft replacing OpenAI with MAI models in Excel and Outlook?

Microsoft is replacing OpenAI and Anthropic with its own MAI (Microsoft AI) models in software products like Excel and Outlook to reduce AI costs, per Bloomberg reporting in July 2026 (Yahoo Finance, July 7, 2026). Tens of thousands of AI prompts in the widely used spreadsheet and email applications are now being completed each week with Microsoft's internally built MAI models. While still accounting for a small share of overall AI usage, the move shows Microsoft is making progress in its efforts to build competitive artificial intelligence models at a lower cost. The change is part of Chief AI Officer Mustafa Suleyman's strategy to reduce and ultimately eliminate Microsoft's dependence on third-party AI providers.

What are Microsoft's MAI models?

Microsoft's MAI (Microsoft AI) models are the company's in-house model family announced at Build 2026, intended to reduce dependence on OpenAI and Anthropic (Yahoo Finance, July 7, 2026). The headline model is MAI-Thinking-1, Microsoft's first reasoning model, trained from scratch on clean, commercially licensed data without distillation from third-party systems (Euronews, June 3, 2026). MAI-Code-1-Flash is a coding model now rolling out across GitHub Copilot and Visual Studio Code. Microsoft has also announced a Microsoft-built model for transcription that will start being used with the Teams videoconferencing apps and other products in the coming months. The MAI model family is positioned as the long-term replacement for OpenAI and Anthropic models in Microsoft products.

How is MAI-Thinking-1 different from OpenAI's GPT models?

MAI-Thinking-1 is a mid-sized reasoning model with 35 billion active parameters and a 256,000-token context window, designed for complex multi-step instructions, long-context reasoning, and code generation (Euronews, June 3, 2026). It was trained from scratch on clean, commercially licensed data without distillation from third-party systems, which is a different training-data approach from OpenAI's GPT models. In blind evaluations run by Surge, Microsoft's independent human rating partner, MAI-Thinking-1 was preferred over Anthropic's Claude Sonnet 4.6, and the company says it matches Claude Opus 4.6 on coding benchmarks. Microsoft also said the model outperforms OpenAI's GPT-5.5 on quality with what it projects as ten times better cost efficiency, based on public pricing data scaled across model sizes.

What is MAI-Code-1-Flash and where is it used?

MAI-Code-1-Flash is Microsoft's in-house coding model that converts text descriptions into source code for applications and websites, now rolling out across GitHub Copilot and Visual Studio Code (Euronews, June 3, 2026). The model is designed for low-latency, high-volume code completion tasks and is integrated with the existing GitHub Copilot and VS Code extensions. The rollout began in late June 2026 and is expected to be available to all GitHub Copilot and Visual Studio Code users by end of Q3 2026. The MAI-Code-1-Flash model is intended to reduce Microsoft's dependence on OpenAI for code completion, which is one of the highest-volume AI use cases across the Microsoft developer ecosystem.

Microsoft MAI Models Replace OpenAI in Excel, Outlook

Q: How does the MAI model strategy affect OpenAI and Anthropic?

The MAI model strategy is a competitive threat to both OpenAI and Anthropic, particularly in the enterprise productivity market. Microsoft is one of the largest enterprise software vendors in the world, and the integration of MAI into Excel, Outlook, Teams, and other Office apps gives the models a built-in distribution channel that OpenAI and Anthropic do not have. The cost-reduction angle is also strategically important: Microsoft currently gets a large amount of AI technology from OpenAI at discounted rates because of their long-term partnership, but the company knows this discounted arrangement may not last forever. By investing in MAI, Microsoft hedges against the risk that OpenAI raises prices or that the partnership deteriorates. The strategy also weakens Anthropic's position in the AI market because Microsoft may buy fewer AI services from it, which could affect Anthropic's valuation by the expected 2027 IPO.

Published September 10, 2026 - Redmond, WA. Microsoft is replacing OpenAI and Anthropic models with its own MAI (Microsoft AI) models in software products like Excel, Outlook, and other Office apps, per Bloomberg reporting in July 2026 (Yahoo Finance, July 7, 2026). Tens of thousands of AI prompts in the widely used spreadsheet and email applications are now being completed each week with Microsoft's internally built MAI models. The change is part of Chief AI Officer Mustafa Suleyman's strategy to reduce and ultimately eliminate Microsoft's dependence on third-party AI providers, and is the clearest signal yet that the AI industry is moving from a 'use the best third-party model' model to a 'build your own' model for large strategic customers.

Microsoft AI strategy data last verified September 10, 2026 from Bloomberg reporting via Yahoo Finance (July 7, 2026), Microsoft Build 2026 announcements, and Euronews coverage (June 3, 2026).

Quick Answer

Microsoft is replacing OpenAI and Anthropic with its own MAI models in Excel, Outlook, and other Office apps. Tens of thousands of AI prompts per week are now completed with MAI models. The headline MAI-Thinking-1 reasoning model was trained from scratch on clean data, with 35B active parameters and 256K-token context. Microsoft says it matches Claude Opus 4.6 on coding and outperforms GPT-5.5 on quality at 10x lower cost. The strategy weakens OpenAI and Anthropic's position in the enterprise productivity market.

What Was Announced at Build 2026

Microsoft unveiled seven in-house AI models at its annual Build developer conference in June 2026, in the clearest sign yet that the tech giant is moving to reduce its dependence on the AI companies it has poured billions into (Euronews, June 3, 2026). The announcement was framed as the moment Microsoft stops being a customer of the AI industry and becomes a competitor. The headline release is MAI-Thinking-1, Microsoft's first reasoning model, trained from scratch on clean, commercially licensed data without distillation from third-party systems.

A mid-sized model with 35 billion active parameters and a 256,000-token context window, MAI-Thinking-1 is designed for complex multi-step instructions, long-context reasoning, and code generation (Euronews, June 3, 2026). Alongside it, Microsoft 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 its own models on Azure infrastructure, Microsoft can sidestep the fees it currently pays to third-party providers and pass the savings to developers.

Quality and Cost Claims

In blind evaluations run by Surge, Microsoft's independent human rating partner, MAI-Thinking-1 was preferred over Anthropic's Claude Sonnet 4.6, and the company says it matches Claude Opus 4.6 on coding benchmarks (Euronews, June 3, 2026). Mustafa Suleyman, Microsoft AI chief executive, said that after tuning its models for consulting firm McKinsey, the company outperformed OpenAI's GPT-5.5 on quality with what it projects as ten times better cost efficiency, based on public pricing data scaled across model sizes.

The cost-efficiency claim is significant. Microsoft currently gets a large amount of AI technology from OpenAI at discounted rates because of their long-term partnership, but the company knows this discounted arrangement may not last forever (Hindustan Times, July 7, 2026). By investing in MAI, Microsoft hedges against the risk that OpenAI raises prices or that the partnership deteriorates. The strategy also weakens Anthropic's position in the AI market because Microsoft may buy fewer AI services from it, which could affect Anthropic's valuation by the expected 2027 IPO.

Strategic Investor Implications

Microsoft's MAI model strategy is a competitive threat to both OpenAI and Anthropic, particularly in the enterprise productivity market. Microsoft is one of the largest enterprise software vendors in the world, and the integration of MAI into Excel, Outlook, Teams, and other Office apps gives the models a built-in distribution channel that OpenAI and Anthropic do not have.

Investors may now have less confidence that Anthropic will reach its expected high valuation by December 31, 2026, per CryptoBriefing analysis cited in Hindustan Times coverage (Hindustan Times, July 7, 2026). Any new partnerships, funding announcements, or business updates from Anthropic could affect how investors value the company. The Microsoft shift adds an additional variable to the Anthropic IPO pricing expected in 2027.

Why the Strategic Shift Now

The shift to in-house models is driven by three forces. First, the cost of training and running competitive AI models has come down enough that large companies can afford to do it themselves, rather than depend on OpenAI's API. Second, the OpenAI partnership has been a source of strategic uncertainty for Microsoft, particularly after the November 2023 board drama and the ongoing relationship management required to keep both companies aligned. Third, the AI capabilities required for Microsoft's product roadmap - agentic workflows, deep Office integration, computer use, enterprise data grounding - are best built with deep product integration that a first-party model enables.

For Microsoft, the cost of training and operating MAI is also lower than paying OpenAI at full API rates, particularly for the high-volume use cases in Excel and Outlook where the per-prompt cost compounds at scale. The savings can be reinvested in additional model training, infrastructure, or pricing concessions to enterprise customers.

What This Means for the AI Industry

The MAI model strategy is the clearest signal yet that the AI industry is moving from a 'use the best third-party model' model to a 'build your own' model for large strategic customers. Amazon has Anthropic, Google has Gemini, Microsoft has MAI, and Apple is investing in Apple Intelligence. Each of these large tech companies is choosing to invest in their own AI capabilities rather than depend on OpenAI for the long term.

The trend is likely to continue: by 2027, most large enterprises will have their own AI model efforts, and the third-party AI market will be dominated by mid-market companies and the smaller cohort of independent AI labs that have not been acquired or contracted. The MAI rollout also signals that the cost of training and running competitive AI models has come down enough that large companies can afford to do it themselves, rather than depend on OpenAI's API. For OpenAI, the strategic implication is that the company needs to win the developer mindshare and the consumer mindshare to maintain growth, even as the largest enterprise customers build in-house alternatives.

Verify current MAI rollout progress and pricing on the official Microsoft AI portal at microsoft.com/ai and the Azure AI Foundry documentation.

#microsoft mai models 2026 #microsoft replace openai excel outlook #mai thinking 1 reasoning model #mai code 1 flash #microsoft ai cost reduction 2026 #suleyman microsoft ai strategy

Written by

Fazlur RahmanAI, Entrepreneurship, EdTech

Fazlur Rahman is the founder of Tutorsbot, building AI-powered tools for learning and career growth. He writes about applying AI in real products and the practi… Read more

Fazlur Rahman is the founder of Tutorsbot, building AI-powered tools for learning and career growth. He writes about applying AI in real products and the practical side of building an ed-tech startup.