Tired of ads? Enjoy an ad-free experience by signing up.
👩‍🍳 How we use AI at Tech in Asia, thoughtfully and responsibly.
🧔‍♂️ A friendly human may check it before it goes live. More news here

Microsoft invests heavily to train its own AI models

Microsoft is making significant investments in infrastructure to train its own AI models, according to Mustafa Suleyman, head of Microsoft AI.

The company recently launched its first in-house AI models and plans to expand AI compute clusters.

The current MAI-1-preview model was trained on 15,000 H100 chips, with future efforts using resources 6x to 10x larger.

Microsoft aims to train frontier models comparable to those from Meta, Google, and xAI, while staying open to external models when needed.

CEO Satya Nadella said Microsoft will continue to support multiple AI models within its products, citing GitHub Copilot as an example.

Microsoft also plans to use Anthropic’s AI models for some Microsoft 365 features, after tests reportedly found them to outperform OpenAI in Excel and PowerPoint.

🔗 Source: The Verge

🧠 Food for thought

Implications, context, and why it matters.

Microsoft shifts from AI partnership dependency to strategic diversification

  • Microsoft’s current approach marks a significant departure from its previous heavy reliance on OpenAI, where it invested $1 billion in 2019 to become the company’s exclusive cloud provider 1.
  • The company is now developing its own MAI-1-preview model while simultaneously integrating Anthropic’s AI models into Microsoft 365 features, particularly for Excel and PowerPoint 2.
  • This multi-model strategy reduces Microsoft’s dependency risk and gives the company more control over its AI roadmap, particularly important as AI becomes central to its product offerings.
  • CEO Satya Nadella’s emphasis on supporting “multiple models” across Microsoft products demonstrates how the company is hedging its bets rather than putting all resources behind a single AI partnership 2.

Massive computational scale required for competitive AI development

  • Microsoft’s current MAI-1-preview model was trained on just 15,000 NVIDIA H100 chips, which Microsoft AI chief Mustafa Suleyman described as “a tiny cluster in the grand scheme of things” 2.
  • To compete with Meta, Google, and xAI’s efforts, Microsoft plans to scale up to clusters that are “six to ten times larger in size,” suggesting they need between 90,000 to 150,000 H100 chips for future models 2.
  • This massive infrastructure investment requirement explains why Microsoft is making “significant investments” in its own computational capacity rather than relying solely on external partners 2.
  • The scale demonstrates the barriers to entry in frontier AI model development, where only companies with substantial cloud infrastructure and capital can compete effectively.

Recent Microsoft developments

Stay ahead in Asia’s tech landscape

You've reached your 2 free content limit for the month. Sign up for free to read the full story.

🏄 For casual readers / 👶 Free

Basic

US$0

Free forever

Get instant access to this article and more every month

0 premium content

Unlimited news briefs

5

5 articles

Ad-free reading experience

Just US$0 per day

⌛Sign up in 20s. No payment details needed.

📖 For learners / 👍 Starter

Lite

US$4.92/month

Billed annually at US$59/year

Get instant access to this article and more every month

4

4 premium content

Unlimited news briefs & articles

Ad-free reading experience

Just US$0.17 per day

Cancel anytime

Our subscriber community includes professionals from these companies:

Stay updated on the go with our mobile app.

Get latest insights with smoother, more personalized experience through TIA mobile app.