Microsoft put roughly $13 billion into OpenAI. Now it’s building its own models that go head to head with the company it funded. This isn’t a conflict of interest. It’s the most expensive insurance policy in tech history, and it’s paying off fast.
What Is Actually Happening
For two years, the narrative was simple. Microsoft funds OpenAI, OpenAI builds the best models, Microsoft sells them through Azure. Clean. Profitable. Symbiotic.
That story is over.
According to Bloomberg, Microsoft has built an internal AI research division that now numbers in the thousands. The Phi model family, Microsoft’s own small language models, has been punching well above its weight class. According to Microsoft Research, Phi-4 Mini outperforms models three times its size on key reasoning benchmarks. And according to The Information, Microsoft’s MAI division is developing frontier models designed to sit alongside, and eventually replace, its dependence on OpenAI’s GPT series.
Meanwhile, Anthropic has locked in cloud deals with both Google and Amazon, making it a serious independent player with the compute to back it up. The three-way competition for enterprise AI contracts is now fully open, and there is no obvious winner yet.
Why Microsoft Is Playing Both Sides on Purpose
Here’s what most people miss. Microsoft doesn’t need OpenAI to win. Microsoft needs AI infrastructure to win.
There’s a massive difference.
When you own the pipes (Azure), you make money no matter which model runs through them. OpenAI’s models. Anthropic’s Claude. Meta’s Llama. Your own Phi or MAI. It doesn’t matter. You collect the compute toll either way.
But Microsoft went further. It built its own models so it has a seat at the table when enterprise clients ask which model they should use. That question used to route all traffic to OpenAI by default. Now Microsoft can answer it with its own product, keep the margin internal, and still bill the compute on Azure.
According to Statista, the global cloud AI services market was valued at over $65 billion in 2025 and is projected to grow past $200 billion by 2030. Microsoft wants to capture that revenue at every layer, not just the infrastructure layer.
Rich mindset versus poor mindset framing fits perfectly here. The average tech observer sees a company betraying its partner. The operator sees a company that studied how Amazon built AWS, how Google built internal tools before selling them, and decided to follow the same playbook. You build it for yourself first. Then you sell it to everyone else.
OpenAI knows this. Sam Altman has been public about the tension. And yet OpenAI keeps cashing Microsoft’s checks because it needs the compute credits and distribution. That dependency is the card Microsoft holds, and it’s not going away anytime soon.
If you’re building AI products right now, this competition is directly lowering your costs. Tools like InVideo AI are getting cheaper and better precisely because the underlying model providers are in a price war for your API spend. Serious competition at the top means real savings at the bottom.
What This Means for You
If you’re an investor, the obvious play is already priced in. Microsoft stock reflects the AI trade. What hasn’t been priced in is the margin expansion that comes when Microsoft reduces its OpenAI royalty costs by replacing API calls with its own in house models.
According to Morgan Stanley estimates from late 2025, Microsoft pays OpenAI a revenue share on Azure OpenAI Service usage. As Microsoft’s own models take more of that traffic, that payment shrinks. That’s a quiet margin story most analysts aren’t modeling correctly yet.
If you’re a builder, the message is different. Do not bet your product on a single model provider right now. The switching costs are low. The performance gaps between providers are closing fast. Build with model abstraction layers so you can move when pricing shifts, and it will shift.
Here’s what I would do right now. Audit every AI API call your product makes. Categorize them by task type: summarization, code generation, image understanding, reasoning. Then test whether a cheaper or faster model from Microsoft, Anthropic, or an open source provider can handle that task with equal quality. You will find savings in almost every category.
For content teams specifically, products that aggregate multiple AI providers into one workflow are getting very good very fast. AppSumo regularly features lifetime deals on tools in this category, letting small teams access enterprise AI stacks without enterprise pricing. Worth a look if you haven’t checked lately.
The businesses that win in this environment treat AI providers like utilities, not like partners. You don’t stay loyal to your electricity company. You pay the lowest rate that keeps the lights on.
The Bottom Line
Microsoft spent $13 billion to learn how to build AI. Now it’s building AI. That’s not betrayal. That’s the most rational move a $3 trillion company can make. OpenAI and Anthropic are both excellent companies. They’re also both now competing with their biggest distribution partner. The builders and investors who didn’t see this coming are already behind. The ones who use the price war to their advantage right now are the ones who come out ahead.
Frequently Asked Questions
Is Microsoft really competing with OpenAI?
Yes. Microsoft has developed its own Phi and MAI model families that compete directly with OpenAI’s GPT series in certain use cases. While Microsoft still offers OpenAI models through Azure, it now also offers its own models as alternatives, especially for cost-conscious enterprise workloads.
Will this hurt OpenAI?
It creates serious pressure on OpenAI’s pricing and market position, but OpenAI still has the most recognized brand in the space and a strong developer base. The bigger risk for OpenAI is that Microsoft controls the main distribution channel, and that dependency becomes more uncomfortable over time.
Is Microsoft a better AI investment than OpenAI or Anthropic right now?
Microsoft is publicly traded and benefits from AI adoption across multiple revenue streams including Azure compute, Copilot licensing, and its own model business. OpenAI and Anthropic are private companies. For most investors, Microsoft is the most accessible way to play the AI race at the infrastructure level.
What does Microsoft competing with OpenAI mean for smaller businesses?
More competition means lower prices and faster model improvements. Small businesses should take advantage by shopping AI API providers based on price and performance rather than brand loyalty. The quality gap between Microsoft, OpenAI, and Anthropic is narrowing every quarter.
How is Anthropic positioned in this competition?
Anthropic has backing from both Google and Amazon, giving it serious compute resources and two major enterprise distribution channels. Its Claude model family is widely regarded as among the best for reasoning and safety-focused tasks. Anthropic is the most independent of the three, which is both its strength and its challenge when it comes to distribution scale.
