source avatarWealth Club

Share

Microsoft’s proprietary coding model claims to achieve 98% of its competitor’s performance at just 40% of the cost, while simultaneously selling an additional $500 million worth of OpenAI models to ByteDance: Is Microsoft really OpenAI’s partner—or its biggest rival? At Ignite 2026, Microsoft unveiled seven new MAI-series proprietary models in one go, covering enterprise use cases such as coding, image generation, cybersecurity, speech, transcription, and decision routing. The most anticipated, MAI Code 2 and MAI Code 2 Flash, have already been integrated into GitHub Copilot. Microsoft claims MAI Code 2 delivers approximately 98% of the performance of GPT-6.1 Sol on enterprise coding benchmarks, at only about 40% of the cost. Note that this comparison is based on a 50% promotional pricing tier—$0.50 per million input tokens and $2.50 per million output tokens—valid until May 17, 2027. Almost simultaneously, reports emerged that Microsoft is in advanced negotiations with ByteDance to sell an additional $500 million in Azure cloud services, primarily to enable access to OpenAI models hosted on Azure. OpenAI has expressed concerns that Chinese companies might use model outputs to train competing systems; Microsoft, however, states it has found no evidence of such activity by ByteDance. When viewed together, Microsoft’s strategy is clear: control the model layer internally, while monetizing the platform layer for everyone. Its proprietary MAI models reduce dependence on OpenAI and give Microsoft pricing and cost control over high-volume products like GitHub Copilot. Meanwhile, continuing to sell access to OpenAI models allows Azure to collect platform fees from its partner’s technology. Regardless of which model enterprises ultimately use, as long as it runs on Azure, Microsoft profits. For OpenAI, this relationship is growing increasingly complex. Microsoft is both its largest shareholder and primary compute partner, yet it is now launching lower-cost alternatives in its most lucrative domain—coding—while also selling OpenAI’s models to customers OpenAI itself is reluctant to serve. The divergence in interests will gradually manifest in future contract negotiations, revenue sharing, and compute allocation. From an investment perspective, if Microsoft’s proprietary models can大规模 replace external models in products like Copilot, it will directly improve the gross margin of its AI business. However, the true cost competitiveness after the promotional pricing ends—and the results from independent third-party evaluations—remain to be validated. The ByteDance deal may also introduce geopolitical and regulatory uncertainties. As Microsoft simultaneously undercuts OpenAI in its most profitable segment with cheaper in-house models and charges more customers for access to OpenAI’s models, how long can this partnership—once hailed as the most successful of the AI era—continue to deliver “win-win” outcomes?

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.