The Brexit of the AI Market Has Begun: The True Meaning of Microsoft MAI

Microsoft has announced the release of its own AI models, the MAI series, declaring a break with OpenAI. This announcement is not merely about a new product launch but signifies a structural transformation of the AI market.

Blue futuristic image of Microsoft MAI models with data center and Copilot UI, symbolizing independence from OpenAI and enterprise AI efficiency

MAI-Voice-1 and MAI-1-preview are expected to merge with Microsoft’s enterprise ecosystem, armed with efficiency as their weapon. By analyzing Copilot’s stagnating user base, its gap with ChatGPT, and the short- and long-term impacts from an investor’s perspective, this article explores the dawn of the AI Warring States Era.

An illustration symbolizing Microsoft MAI’s independence with a blue-toned data center and Copilot UI in the background, highlighting the split with OpenAI and emphasizing an enterprise AI strategy focused on efficiency.

On August 28, 2025, Microsoft shook Silicon Valley with a bold move: the release of its homegrown AI model series, MAI.¹ On the surface, it looked like a new product launch, but in reality, it was essentially the Brexit of the AI market. After six years of a close relationship with OpenAI, Microsoft had declared independence.

Why now? And what message does this move send us?

The End of a Shared Dream: From Partnership to Competition

The relationship between Microsoft and OpenAI was complicated from the start. Beginning with a $1 billion investment in 2019, the partnership looked like a win-win on the surface, but behind the scenes, each had different goals. OpenAI pursued the lofty goal of achieving AGI, while Microsoft sought practical revenue generation.

The seeds of conflict were sown early. In 2024, Microsoft added OpenAI to its list of competitors in its annual report.² Imagine the situation: a partner in which you’ve invested $13 billion suddenly becomes a competitor.

So Microsoft brought in a game-changer: Mustafa Suleyman, co-founder of DeepMind and former leader at Inflection AI. His first words were telling: *“We need the capability to build the world’s most powerful models in-house.”*³ Translated, it meant: “Let’s stop envying someone else’s cake.”

The Technical Ambition of MAI: Efficiency as a Weapon

The MAI series started with two models. MAI-Voice-1 can generate one minute of natural-sounding speech in just one second on a single GPU.⁴ This isn’t just fast — it’s revolutionary. MAI-1-preview is a text-based model trained with 15,000 NVIDIA H100 GPUs and is set to power Copilot in the future.

What stands out here is Microsoft’s strategy. Instead of “bigger and more,” the company chose “more efficient.” Suleyman explained: *“The art and craft of model training is about selecting the right data and not wasting flops on useless tokens that don’t really teach the model much.”*⁵

This is the exact opposite of OpenAI’s expansion-at-all-costs strategy. For comparison, xAI’s Grok was trained with over 100,000 H100 chips, while MAI-1-preview managed with just 15,000.

The Numbers: The Gap Remains

Let’s face reality. Microsoft Copilot has stagnated at 20 million weekly users over the past year.⁶ Not a bad number, but compared to ChatGPT’s 800 million users (as of August 2025),⁷ the difference is overwhelming. A ratio of 40:1.

But what if we look at this differently? Microsoft already holds a dominant position in enterprise software: Office 365, Azure, Teams. If MAI integrates seamlessly into this ecosystem, the rules of the game could change.

Key Points for Investors

Short-term impact:

  • Positive for Microsoft stock: securing AI independence acts as a risk hedge.
  • Negative factor for OpenAI-related investments: weakening partnership could prompt a valuation reassessment.
  • NVIDIA is a beneficiary: as Microsoft accelerates in-house development by deploying next-generation GB200 clusters,⁸ GPU demand will increase.

Long-term perspective:

  • Acceleration of a multipolar AI market: Google, Amazon, Meta may also strengthen independent paths.
  • Strengthening Microsoft’s enterprise AI market presence: building a B2B-focused AI ecosystem.
  • Rise of open-source AI models: intensified competition among big tech could benefit open-source innovation.

Future Scenarios: The Opening of the AI Warring States Era

Microsoft’s MAI announcement is not just about a product launch. It is a signal flare that the structure of the AI market itself is changing.

Scenario 1: Successful Independence

A fully connected AI ecosystem linking Copilot+ PCs, custom AI chips, and edge computing, securing overwhelming dominance in the enterprise market.

Scenario 2: Escalating Chicken Game

Competition with OpenAI intensifies, exhausting both sides. In the end, Google or Amazon could reap the benefits.

Scenario 3: Coexistence of Cooperation and Competition

On the surface, competition continues, but in certain areas, cooperation persists. Suleyman himself said: *“The goal is to deepen our partnership with OpenAI and ensure excellent collaboration for years to come.”*⁹

Personally, I’d bet on Scenario 1. Microsoft’s DNA in enterprise software and MAI’s efficiency-driven strategy seem well aligned.

The Essence We Must Not Miss

At the heart of all these moves lies independence — technological independence, strategic independence, and above all, control over the future. Microsoft’s message is clear: “To survive in the AI era, don’t depend on others. Build your own weapons.”

This lesson applies not only to companies but also to individuals. The AI Warring States Era has just begun. And Microsoft has chosen to become an independent force in this war. Whether their choice proves correct, only time will tell.

Microsoft has begun writing its own story. Now it’s the other players’ turn.

References

  1. Microsoft AI (2025). “Two in-house models in support of our mission” - Official Microsoft AI blog
  2. CNBC (2025). “Microsoft tests MAI-1-preview AI model boost to Copilot, rival OpenAI”
  3. Semafor (2025). “Microsoft unveils powerful new home-grown AI models”
  4. StartupHub.ai (2025). “Microsoft AI Unveils First In-House Models MAI”
  5. Engadget (2025). “Microsoft introduces a pair of in-house AI models”
  6. Born’s Tech and Windows World (2025). “Active user count stagnate with Microsoft’s CoPilot”
  7. Backlinko (2025). “ChatGPT Statistics 2025: How Many People Use ChatGPT?”
  8. Microsoft AI (2025). “Two in-house models in support of our mission”
  9. Semafor (2025). “Microsoft unveils powerful new home-grown AI models”

Q&A

Q: What is the MAI series and why is it important?
A: MAI (Microsoft AI) is Microsoft’s first in-house AI model series. It consists of MAI-Voice-1 (speech generation) and MAI-1-preview (text-based), and represents a strategic move to reduce reliance on OpenAI.

Q: What is the technical differentiator?
A: Efficiency is the focus. MAI-Voice-1 can generate one minute of speech in one second on a single GPU, while MAI-1-preview was trained on 15,000 H100 GPUs — far fewer resources than competing models.

Q: What is the user gap?
A: Copilot has 20 million weekly users, stagnating over the past year, while ChatGPT has 800 million — a 40:1 gap. However, the enterprise market may show different dynamics.

Q: How should this be viewed from an investment perspective?
A: Short-term: positive for Microsoft stock, negative for OpenAI-related investments. Long-term: acceleration of a multipolar AI market and strengthening of Microsoft’s enterprise position.

Q: What are the future scenarios?
A: Three possibilities: successful independence, an escalating chicken game, or coexistence of cooperation and competition. Considering Microsoft’s enterprise DNA, successful independence appears most likely.