Microsoft Is Building Its Own AI. What That Means for OpenAI Investors.

Token prices fell again overnight, and both Google and Microsoft are responsible. Google released Gemini 4 Neon on November 18, priced at $3 per million input tokens and $15 per million output tokens. Google claims the mid-tier model scores ahead of Claude Opus 5.5 and GPT-6.2 Sol in its own benchmark tests. That is notable. What happened at Microsoft’s Ignite conference the day before is more consequential for anyone thinking about where durable value settles in AI.

Sponsored

Own Every Major AI Stock for $38. Get Paid Every Thursday.

The biggest AI stocks have been soaring and now cost hundreds of dollars a share. But there’s a single fund that owns all of them – Nvidia, CrowdStrike, Palantir, and dozens more – for just $38 a share. And it pays income every Thursday. Up to $1,051 a month.

Watch the free presentation here.

Microsoft AI announced a second wave of in-house models on November 17 at Ignite in San Francisco. The seven models are MAI Cyber 2 Flash for defensive security, MAI Code 2 and MAI Code 2 Flash for coding, MAI Image 2.7 for image generation and editing, MAI Realtime 1 for live speech-to-speech, Dragon Copilot Transcribe built on MAI Transcribe 2, and Decision 1.1, a router model. Each one benchmarks against an OpenAI or Anthropic product, and each one undercuts it on price.

MAI Code 2, Microsoft says, delivers 98% of the coding performance of OpenAI’s GPT-6.1 Sol at about 40% of the price, and 96% of Anthropic’s Claude Opus 5.5 at about 5% of the price. MAI Cyber 2 Flash matches GPT-6 Astra’s quality at less than 2% of the cost per task on the CyberGym-E2E benchmark. Those are aggressive numbers, and Microsoft chose to publish them against named rivals rather than unnamed baselines.

Reading the Signal

Investors sometimes treat the Microsoft-OpenAI relationship as a durable asset. The last six months have complicated that view significantly. OpenAI and Microsoft amended their partnership on April 27, 2026, shifting Microsoft’s license to OpenAI intellectual property to non-exclusive and changing how revenue sharing works between the two companies. A renegotiated partnership removed prior limits on exclusivity, and owning proprietary models reduces vendor dependency while giving Microsoft more control over pricing and roadmap.

Sponsored

GOLD Just Had a Historic Run. What Happens Next?

GOLD has made a historic price surge, and financial experts are already looking ahead. With America’s debt growing and market instability a real risk, find out what the forecasts say and why GOLD has remained a sought-after asset through inflation, recessions and financial crises. It’s all inside your FREE Warning Gold Guide.

Get the FREE Guide

The Ignite models are not a first experiment. They are the second major wave. Microsoft’s MAI model family and aggressive development of in-house inference capabilities are the moves of a company that has concluded, after north of $13 billion in cumulative OpenAI investment, that building permanent dependency on a single external AI vendor is not a tenable long-term position for a $3 trillion enterprise.

Peter Lynch would note that Microsoft is converting a fixed external cost into a proprietary capability, the kind of vertical integration that expands margins and defensibility at the same time. Charlie Munger would ask who loses pricing power as that shift compounds. The answer is OpenAI, which remains private, and Anthropic, whose Claude is now being benchmarked publicly and undercut in cost by a customer that once paid it dearly.

Where Value Is Migrating

If Microsoft can position its in-house models as the cost-effective, enterprise-sovereign default across Teams, Office, Azure, and GitHub, it reduces its structural dependence on OpenAI pricing and availability, a dependency that has become increasingly visible as OpenAI pursues its own path and considers broader customer relationships.

The pricing pressure is not coming from Microsoft alone. Google’s Gemini 4 Argon launched at a fraction of what Anthropic charges for Claude Opus 5.5, compressing the premium tier from a second direction.

Sponsored

This AI worked all night. Come see what it found.

While most traders slept, VantagePoint patented AI was parsing 1.4 million data points every day across dozens of related global markets, looking for trends starting to turn. In this free live class, you’ll see what it identified today and learn how to use the forecast to plan the trade before the move.

See what the AI identified. Join the FREE live class.

Microsoft’s reliance on OpenAI defined its AI strategy since 2023, when it built Copilot on GPT models and committed billions to the partnership. Owning models, and co-designing them with its own Maia chips, gives Microsoft room to negotiate on cost and to set its own roadmap rather than wait on a partner.

Google’s Gemini 4 Neon reinforces the same structural point from a different angle. Two of the three largest technology companies on earth are now flooding the market with capable models at prices that erode the premium OpenAI and Anthropic have commanded. The model layer is not becoming worthless, but its economic gravity is shifting toward companies with direct distribution, cloud infrastructure, and enterprise relationships. Microsoft and Google have all three. That is not a new observation, but seven MAI models benchmarked against OpenAI by name at a public conference is a harder kind of evidence than a strategy slide.

The long-term question for owners of MSFT is not whether this strategy works. The evidence suggests it already is. The question is how fast Microsoft replaces third-party model spend with first-party margin, and whether Copilot’s commercial momentum accelerates as the underlying economics improve. Both deserve close attention in the quarters ahead.

More From Author

Walmart Reports Today. Margins Just Hit a New High.

Live Market Pulse

The charting technology is provided by TradingView. Learn how to use theTradingView Stock Screener.

Categories