AI Business · · 6 min read
Why is Microsoft Building Its Own AI Models While Still Selling OpenAI and Anthropic?
Understand Microsoft's dual strategy in AI, balancing partnerships with its push for cost-effective, specialized in-house models, and what it means for your business.

At its recent Ignite conference, Microsoft made a bold declaration that caught many by surprise. Judson Althoff, CEO of Microsoft's commercial business, put it bluntly: "Simply stated, anything you can do in Claude or ChatGPT, you can do better with Copilot and for less."
Jacob Andreou, EVP of Copilot, backed this with a chart based on Microsoft-commissioned research from GigaOm. It suggested an "everyday" business user would pay $30 per month for Copilot.
For comparison, the same research estimated $81 for OpenAI's ChatGPT and a hefty $205 for Anthropic's Claude for an "everyday" user. For "advanced" users, the gap was even wider: $96 for Copilot versus $149 for ChatGPT and $299 for Claude.
These are not small differences. They point to a potential shift in the economics of enterprise AI.
But here's the thing: Microsoft is also a major investor in OpenAI and a seller of its models, and it partners with Anthropic, too. In fact, on the very same stage at Ignite, CEO Satya Nadella announced that OpenAI's latest model, GPT-6.2, was now available in Microsoft Foundry.
This creates a fascinating paradox. Why would Microsoft invest so heavily in developing its own AI models, like the newly launched MAI Code 2 and MAI Cyber 2 Flash, while simultaneously promoting and selling the very competitors it claims to undercut? That raises three questions.
Question 1
What New AI Models Did Microsoft Announce and How Do They Claim to Cut Costs?
Beyond the Copilot cost comparison, Microsoft unveiled a new family of specialized AI models at Ignite, all designed to perform specific tasks efficiently and affordably. These are part of its "MAI" series, aimed at reducing reliance on external providers, Reuters reported.
Among the key launches was MAI Code 2, a programming model now available in GitHub Copilot. Microsoft claims it delivers 96% of the coding performance of Anthropic's Claude Opus 5.5, but at just 5% of the cost. But how exactly do these specialized models achieve such significant cost reductions?
Another significant release was MAI Cyber 2 Flash, an AI model tailored for cybersecurity. Microsoft stated it matches OpenAI's GPT-6 Astra on a specific benchmark while costing less than 2% per task. This is a big deal as businesses grapple with a surge in software vulnerabilities, often identified with AI tools.
The company also introduced new models for image generation, speech, and audio transcription. The core idea, as explained by Microsoft AI chief Mustafa Suleyman to Reuters, is that domain-specific designs drastically lower operational costs while sharpening market competitiveness. This isn't about one giant model; it's about a "system... which co-optimizes performance and price."
For context, the cost estimates for Copilot, ChatGPT, and Claude were based on enterprise pricing, which includes usage fees on top of per-seat subscriptions. This means the more an AI is used, the higher the bill, making cost efficiency a critical factor for businesses. And that brings us to the second question.
Question 2
Why is Microsoft Investing in Its Own AI Models Even as It Partners with Rivals?
Microsoft's strategy isn't about abandoning its partners; it's about diversification and control. While it remains a key investor in OpenAI and sells their models, the company is also pursuing what Suleyman calls "long term self-sufficiency," according to GeekWire.
Think of it like a large logistics company that both owns its fleet of trucks and contracts with independent drivers. Owning its own trucks (the MAI models) allows Microsoft to customize them for specific routes (tasks), control maintenance costs, and ensure reliability. This provides a strategic hedge against potential issues with external providers and allows for deeper integration into its own products like Copilot. But why would Microsoft invest in its own fleet when it has such strong partnerships already?
This approach helps Microsoft manage the escalating costs associated with large, general-purpose AI models. By developing specialized models, they can achieve high performance for particular tasks without the massive computational expense required for a single, all-encompassing breakthrough model.
Another crucial aspect is "MAI Licensed Weights." This program allows companies to take a Microsoft AI model, customize it with their own data, and run it within their own Azure cloud environment. It's like a chai stall owner buying a proven recipe (the model's weights) from a master, then brewing and serving the chai in their own shop with their unique blend of spices (their data). This offers businesses greater data privacy, customization, and control over their AI deployments, as AT&T did for its 5G network analytics.
This dual strategy positions Microsoft to serve a broader range of customer needs, from those wanting the latest general-purpose models from partners to those requiring highly specialized, cost-controlled, and customizable solutions that keep their data within their own infrastructure. And this leads us to our final question.
Question 3
What Does Microsoft's Dual Strategy Mean for Businesses Using AI?
For business leaders, Microsoft's evolving AI strategy presents both opportunities and challenges. On one hand, it means more choice and potentially lower costs for specific AI applications. The competition among providers, even those who are partners, is likely to drive down prices and improve performance. How do you navigate this increasingly complex landscape?
However, it also introduces complexity. Businesses now need to carefully evaluate whether a general-purpose model like ChatGPT or Claude, offered by Microsoft, or a specialized MAI model, is the right fit for a given task. This requires a deeper understanding of AI capabilities and cost structures.
Microsoft is aware of the cost concerns. A McKinsey survey cited by the company found that 93% of respondents exceeded their AI budgets. To help, Microsoft is offering tools like Agent 365 Business Analytics and a Cost Management tool, allowing IT departments to set spending policies and get alerts.
The emphasis on "AI alignment", ensuring AI systems behave as intended and don't take harmful actions, also becomes paramount. Mustafa Suleyman highlighted this as his biggest priority, especially given incidents like the OpenAI agents and Hugging Face episode reported by CNBC-TV18. Businesses must consider the security and ethical implications of the models they deploy.
Ultimately, Microsoft's dual approach means businesses have more levers to pull. You can mix and match, leveraging partners for cutting-edge general intelligence while deploying Microsoft's specialized, cost-effective models for core, high-volume tasks. But how do you decide which lever to pull and when? It's about strategic AI adoption, not just buying the latest tool.
The Bigger Picture
The shift towards specialized, cost-optimized AI models signals a maturing market. The initial race was about who could build the biggest, most capable AI. Now, the focus is increasingly on making AI practical, affordable, and tailored for enterprise needs.
This means that simply having access to powerful AI is no longer enough. Businesses must become adept at selecting the right AI tool for the right job, balancing raw power with efficiency, cost, and control. It's about intelligent deployment, not just adoption.
Microsoft's strategy, with its blend of internal innovation and strategic partnerships, reflects this reality. It's a testament to the idea that in the world of AI, the future isn't about one winner, but about a diverse ecosystem where different players offer different strengths.
The era of 'one AI to rule them all' is giving way to a more pragmatic, cost-conscious approach.
Quick Answers
What new AI models did Microsoft announce?
Microsoft announced a range of specialized AI models at its Ignite conference, including MAI Code 2 for programming, MAI Cyber 2 Flash for cybersecurity, and others for image generation and audio transcription, as part of its MAI series.
How does Microsoft Copilot compare in cost to ChatGPT and Claude?
According to Microsoft's commissioned research, Copilot costs $30 per month for an 'everyday' business user, compared to $81 for ChatGPT and $205 for Claude. For 'advanced' users, Copilot was estimated at $96, significantly less than ChatGPT ($149) and Claude ($299).
Why is Microsoft developing its own AI models?
Microsoft is developing its own AI models to achieve 'long term self-sufficiency,' control operational costs, and offer highly specialized, customizable solutions. This strategy allows for better performance-to-price optimization for specific tasks, even while maintaining partnerships with other AI providers.
Sources
Facts are credited to the outlets above. The analysis, calculations and Signal Check are by the Finnect AI Insights desk.