Microsoft crushed its third-quarter earnings expectations but sent its stock sideways after revealing it will spend $190 billion on artificial intelligence infrastructure this year, up 61% from the previous year. The company reported earnings of $4.27 per share, beating analyst estimates of $4.06, and revenue of $82.89 billion, ahead of the $81.39 billion target. Yet the massive capital expenditure forecast, combined with weaker-than-expected guidance for the fourth quarter, left investors questioning whether the company's epic spending on AI will actually pay off.

The timing of this announcement is particularly significant because it comes just days after Microsoft fundamentally restructured its relationship with OpenAI. On April 27, 2026, the two companies ended Microsoft's exclusive cloud distribution rights, allowing OpenAI to work with other cloud providers. Within 24 hours, OpenAI announced a $138 billion compute commitment to Amazon Web Services (AWS), signaling a seismic shift in the AI infrastructure market.

Why Is Microsoft Spending So Much on AI Infrastructure?

Chief Financial Officer Amy Hood explained that the $190 billion capital expenditure includes roughly $25 billion in additional costs from rising memory chip prices and other components needed for AI data centers. Analysts had estimated Microsoft would spend $154.6 billion on capital expenditures in 2026, so the higher forecast caught many investors off guard. Hood also blamed supply chain instability in the Middle East for some of the pressure, but emphasized that Microsoft is not letting these challenges slow its long-term AI ambitions.

The spending reflects a broader industry reality: frontier AI development requires massive computational resources. However, investors are increasingly skeptical about whether this spending will translate into revenue growth. Microsoft's stock has declined 12% year-to-date, marking its worst quarterly performance since the 2008 financial crisis. Analyst Rebecca Wettemann from Valoir noted that the market wanted to see a more solid earnings beat to be reassured about Microsoft's AI strategy.

How Is Microsoft Building Its Own AI Models?

In a direct response to investor pressure and the OpenAI partnership restructuring, Microsoft launched three new foundational AI models built entirely in-house: MAI-Transcribe-1 for speech-to-text conversion, MAI-Voice-1 for voice generation, and MAI-Image-2 for image creation. These models represent Microsoft's opening salvo from its superintelligence team, formed just six months ago to pursue what the company calls "AI self-sufficiency".

The transcription model is the headline release. MAI-Transcribe-1 achieves the lowest average word error rate on the FLEURS benchmark across the top 25 languages by Microsoft product usage, averaging 3.8% accuracy. According to Microsoft's benchmarks, it beats OpenAI's Whisper-large-v3 on all 25 languages, Google's Gemini 3.1 Flash on 22 of 25, and ElevenLabs' Scribe v2 on 15 of 25 languages. The model also delivers batch transcription speeds 2.5 times faster than Microsoft's existing Azure Fast offering.

What makes these releases particularly striking is the efficiency with which Microsoft built them. Mustafa Suleyman, who leads Microsoft's superintelligence efforts, revealed that the audio model was built by just 10 people, and the image team is equally small. "The vast majority of the speed, efficiency and accuracy gains come from the model architecture and the data that we have used," Suleyman explained. This lean-team approach challenges the prevailing industry narrative that frontier AI development requires thousands of researchers and billions in headcount costs.

"I'm very excited that we've now got the first models out, which are the very best in the world for transcription. Not only that, we're able to deliver the model with half the GPUs of the state-of-the-art competition," said Mustafa Suleyman.

Mustafa Suleyman, Head of Superintelligence Team, Microsoft

What Changed in Microsoft's Deal With OpenAI?

Until October 2025, Microsoft was contractually prohibited from independently pursuing artificial general intelligence (AGI), a theoretical future state where AI systems match or exceed human intelligence across all domains. The original 2019 deal with OpenAI gave Microsoft a license to OpenAI's models in exchange for building the cloud infrastructure OpenAI needed. When OpenAI sought to expand its compute footprint beyond Microsoft, the two companies renegotiated the terms.

The revised agreement, finalized in September 2025, freed Microsoft to build its own frontier models while retaining license rights to everything OpenAI builds through 2032. The new deal also removed the AGI clause that had tied Microsoft's intellectual property rights and revenue share to OpenAI achieving artificial general intelligence; rights are now fixed to a calendar date rather than a speculative technical milestone. Additionally, the deal caps the revenue share OpenAI pays Microsoft, signaling that Microsoft prefers a known maximum exposure to an uncapped upside scenario.

How Are These Changes Reshaping the AI Market?

The restructuring of Microsoft's OpenAI relationship marks the formal end of an era when one cloud provider could anchor a frontier AI lab. For three years, the assumption was that frontier-lab compute and frontier-lab distribution would consolidate around the same cloud. Now, two clouds have OpenAI models available, two clouds have Anthropic's Claude, and the cloud market for frontier AI is genuinely contested for the first time.

The competitive implications are significant. Microsoft's Copilot is built on OpenAI models served via Azure. OpenAI's new Frontier platform, launched in February 2026 for building and managing AI coworkers in Fortune 500 environments, is now available exclusively on AWS. This creates an awkward dynamic where Fortune 500 buyers can compare Microsoft's Copilot stack directly against OpenAI's own Frontier stack, with OpenAI controlling both the underlying models and one of the distribution channels.

Anthropic, which has relied on Amazon Bedrock as its primary cloud distribution channel, also faces new pressure. After April 28, 2026, AWS sales representatives now have two frontier AI options to pitch to customers, and the OpenAI brand is more recognizable to most enterprise buyers. Anthropic's competitive playbook has narrowed from exclusivity to differentiation, relying on features like Claude Code's native sandboxing and Computer Use capabilities.

Steps to Understanding Microsoft's AI Strategy Shift

• Model Independence: Microsoft freed itself from contractual restrictions on building its own frontier AI models, allowing it to compete directly with OpenAI and Google rather than relying solely on distribution partnerships.
• Efficiency Focus: The company is building state-of-the-art models with small teams of fewer than 10 engineers, using superior model architecture and data rather than massive headcount, which improves profit margins.
• Enterprise Positioning: Microsoft is launching models across multiple modalities (speech, voice, images) and integrating them into existing products like Copilot, Teams, and Office, creating multiple revenue streams from AI capabilities.
• Cost Reduction: By building its own models, Microsoft can reduce its reliance on third-party AI providers and lower its cost of goods sold, addressing investor concerns about profitability.

Microsoft's Copilot business is already showing momentum despite the broader market skepticism. Chief Executive Satya Nadella reported that the company now has more than 20 million seats for its 365 Copilot AI add-on for commercial Office subscriptions, up from 15 million three months earlier. He expects a similar increase in the current quarter. "Weekly engagement is now at the same level as Outlook as more and more users make Copilot a habit," Nadella stated.

Annualized revenue from all of Microsoft's AI services now totals more than $37 billion, up 123% from the same period one year ago. This includes customers running AI services on Azure, revenue from developers building models, and AI tools like Copilot. The Azure cloud infrastructure business also delivered strong results, with revenue from that segment and other cloud services rising 40% in the quarter, ahead of the Street's target of 39.3%.

Yet investor skepticism persists. Analyst Holger Mueller from Constellation Research offered a more measured assessment, noting that Microsoft deserves credit for managing to keep growing and become more profitable while AI spending reaches epic levels. "Its cash flow remains positive despite the substantial spending on AI infrastructure, but investors will want to see this balance maintained in the coming quarters," Mueller said.

"Back in September of last year, we renegotiated the contract with OpenAI, and that enabled us to independently pursue our own superintelligence. Since then, we've been convening the compute and the team and buying up the data that we need," explained Mustafa Suleyman.

Mustafa Suleyman, Head of Superintelligence Team, Microsoft

The broader context is that Microsoft's $190 billion capital expenditure forecast reflects a bet that compute scarcity will continue through at least 2034. The company is essentially locking in massive computational resources to ensure it can compete with OpenAI, Google, and other frontier labs on model development. Whether this spending translates into the revenue growth investors are demanding remains the central question facing the company heading into 2027.