Microsoft AI investments — 11,000 models on the shelf

Azure, Foundry, Copilot and enterprise data let Microsoft profit while OpenAI, Anthropic, Mistral and its own models compete task by task.

Microsoft turns AI investments into enterprise software rivalries with one ruthless move: make every model replaceable inventory. OpenAI, Anthropic, Mistral, xAI and Microsoft’s MAI models compete by task; Microsoft owns the platform and bills everyone.

I heard Satya Nadella explain this on Microsoft’s July 29 earnings call, because investor transcripts count as summer reading. Somewhere in Ivrea, my younger self closed the laptop and went outside.

Microsoft invested in AI labs so the eventual winner matters less.

The strategy has weight: Azure passed $100 billion in annual revenue, Microsoft 365 Copilot has over 30 million paid seats, and Microsoft Foundry offers more than 11,000 models—a catalog with strong Costco energy.

Microsoft monetizes models, sells compute, connects company data and charges for the app showing the answer. My nonna would understand: own the espresso machine and café lease; let others argue about beans.

The $41 billion tollbooth

In its July 29 earnings release, Microsoft reported fiscal 2026 fourth-quarter revenue of $90 billion, up 18% year over year. Microsoft Cloud generated $59.3 billion, up 27%; Azure and other cloud services grew 43%.

AI infrastructure spending resembled a national space program with worse merch. Now revenue comes from both sides.

Activate Consulting CEO Michael J. Wolf told the Associated Press on July 29 that Microsoft was “winning on both fronts”: enterprises pay for Azure infrastructure, then Copilot inside existing software.

A better product can win a demo. The product already on every laptop wins procurement.

Nadella summarized the scale in Microsoft’s July 29 earnings release:

“This year, Azure revenue surpassed $100 billion for the first time, and Microsoft 365 Copilot reached over 30 million paid seats, reflecting the confidence customers are placing in us to power their AI transformation.”

Those 30 million seats let Microsoft test features, prices and model substitutions. Standalone AI companies face procurement; Microsoft adds AI to agreements covering Outlook, Teams, Excel, Windows and enough security products to break an exhausted IT manager.

Then comes the spending.

Microsoft spent $41 billion on quarterly capital expenditures, according to AP. CFO Amy Hood said an accounting change would put calendar 2026 guidance near $175 billion without changing underlying investment expectations. AP also cited Microsoft’s previous roughly $190 billion expectation, including about $25 billion from higher component prices.

I expected discipline sooner, especially when inference economics began resembling a restaurant where everyone orders lobster and pays for a sandwich.

Enterprise contracts tolerate the feast. Commercial remaining performance obligation reached $678 billion, up 84%, according to the earnings release. Those commitments turn GPUs and datacenters into recurring consumption.

Nadella described the expansion:

“We added 31 new datacenters across 5 continents this quarter, bringing the total to 88 this year, as we expand our footprint in response to accelerating demand.”

AI labs provide intelligence. Microsoft increasingly owns its workplace.

OpenAI and Anthropic become inventory

Microsoft Foundry carries more than 11,000 models, including OpenAI, Anthropic, Mistral and xAI products alongside Microsoft’s MAI family.

Nadella described the catalog during Microsoft’s fiscal 2026 fourth-quarter call:

“We offer the broadest model catalog in the cloud, with over 11,000 models, including the latest from OpenAI, Anthropic, Mistral, xAI, as well as our own MAI family.”

A bank can choose separate models for document extraction, coding and regulated local workflows, balancing quality, latency, cost and data location.

Each model strengthens Foundry because Microsoft handles evaluation, deployment, billing and policy enforcement. Task-by-task swapping turns models into ingredients.

Microsoft says customers building with multiple providers increased fivefold during the first half of 2026.

Nadella gave investors the figure:

“Since the start of the year, we have seen a 5X increase in the number of customers building with models from multiple providers.”

Levi Strauss & Co. uses OpenAI and Anthropic models in Foundry while consolidating more than 1,000 domain-specific agents on one enterprise AI platform. Levi’s can switch suppliers; Microsoft stays embedded.

The finances are wonderfully awkward. Microsoft recorded a $3.2 billion quarterly gain from its Anthropic investment. Full-year statements separately showed OpenAI investments increasing net income by $4.963 billion.

Microsoft profits when a lab appreciates while building products to survive without it.

Cold. Also brilliant.

Infrastructure follows. Microsoft says Maia 200 chips support OpenAI and MAI models with 30% better performance per dollar than its latest-generation fleet hardware. Cobalt virtual machines power Microsoft workloads and systems from Adobe, Arm, Elastic, OpenAI, Sprinklr and TomTom.

Every workload adds scale; every model adds choice. Partners compete without quite leaving the alliance, though several lawyers are probably always typing.

Excel knows what the model labs don’t

Microsoft’s nastiest advantage is ordinary user behavior.

A model company can train on code and spreadsheets. Microsoft sees whether developers accept GitHub Copilot suggestions in VS Code, return two days later or rewrite them while muttering words unsuitable for stand-up.

Excel shows whether an agent finished the work or drove users back to manual formulas and quiet despair.

Benchmarks rarely do.

According to a July 23 post from Microsoft AI’s Superintelligence team, MAI-Code-1-Flash achieved an approximately 10% higher VS Code code acceptance rate than GPT-5.4 Mini and Claude Haiku 4.5. Developers were 6% more likely to return across multiple days than with GPT-5.4 Mini and 11% more likely than with Claude Haiku 4.5.

Microsoft AI stated:

“It has an approximately 10% higher code accept rate than GPT 5.4 Mini and Claude Haiku 4.5 in VS Code.”

Acceptance beats benchmark trophies because it reflects deadline pressure, a waiting pull request and increasingly hostile Slack.

Microsoft then trained the MAI-Code-1-Flash checkpoint in an Excel reinforcement-learning environment. Production feedback showed quality comparable to GPT-5.6 on common Excel tasks. The specialized model runs on Nvidia H100 and A100 GPUs, not only the latest accelerators.

That should worry suppliers. Microsoft owns product-specific evaluations showing where frontier models justify their price and smaller MAI models can replace them.

Nadella made the ambition explicit in comments covered by ITPro on July 23:

“We are now seeing MAI models outperform general-purpose frontier models in many use-cases while using a fraction of the tokens.”

The savings are substantial. Microsoft reported an 89% GPU-cost reduction in Dynamics 365 using MAI-Voice-2-Flash. In PowerPoint, MAI-Image-2.5 cut GPU costs by as much as 84%.

We had the physical machine, cloud services and brew telemetry in one product. Problems arose where those layers met, favoring whoever saw the full loop over isolated suppliers.

Yes, coffee telemetry became a business lesson. I was born in Italy; I’m legally required to make caffeine strategic.

GitHub and Excel give Microsoft that loop at scale. Labs build intelligence; Microsoft writes the exam and watches millions take it.

A visual showcasing various AI models, highlighting Microsoft's extensive investment in artificial intelligence technology.

Image alt text: How Microsoft turns AI investments into enterprise software rivalries across Azure, Foundry, Copilot and MAI.

Caption: Partners provide models. Microsoft owns the production loop that decides which ones keep the job.

Permission to press “execute”

Enterprise AI creates value by changing purchase orders, approving access, resolving IT tickets or contacting customers. Chat answers are cute. Permission to alter live processes is money.

The platform must know the employee, policy and current customer record. Every action must survive an auditor arriving six months later with airport-security warmth.

Microsoft’s expanded Databricks partnership targets that control. The July 23 agreement runs into the 2030s and integrates Databricks Genie and Unity AI Gateway across Entra, Power BI, Purview, Foundry, Microsoft 365, Teams and Copilot.

Databricks says more than 20,000 organizations use its platform, including 70% of the Fortune 500. It will expand Azure Databricks for core operations and adopt Cobalt 200, which Microsoft says performs up to 50% better than its predecessor.

Foundry competes for developers; Databricks connects Microsoft to governed enterprise data; Copilot owns the employee interface; Dynamics reaches business applications.

ServiceNow sees the same prize. Its AI business crossed $1 billion in annual contract value during Q2 2026, while production agentic deployments rose ninefold in nine months, according to its July results. It ended the quarter with 658 customers generating more than $5 million each in annual contract value.

Chairman and CEO Bill McDermott opened with characteristic restraint:

“ServiceNow’s exceptional Q2 results solidify our position as the fastest-growing major enterprise software and cybersecurity company.”

ServiceNow calls itself the “AI control tower” and expanded it to govern agents anywhere. Action Fabric lets ServiceNow and third-party AI execute work through ServiceNow workflows, with Anthropic as first design partner.

That challenges Microsoft’s governance through Foundry, Entra and Microsoft 365. They can partner while fighting over customers. Enterprise software is civilized that way.

SAP has a credible attack through systems of record. In its July 23 results, SAP reported current cloud backlog of €22.9 billion, up 27%, and cloud revenue of €6.28 billion, up 22%.

CEO Christian Klein said SAP’s momentum comes from AI grounded in customers’ most critical processes and data. He is right: a supply-chain agent using live inventory and approval rules beats a brilliant chatbot guessing from a PDF.

Salesforce attacks through observability. Agentforce’s Session Tracing Data Model records user input, planner decisions, retrieved sources, action flows, errors and outputs. Waterfall traces follow work across agents; citations link reviewers to sources.

Salesforce also made its standard Agentforce observability stack unmetered, letting builders inspect production traces without extra Data Cloud credits for default tooling.

Model quality wins the demo. Execution rights win the account.

Europe needs to own more than the model

I’m happy about Microsoft’s expanded Mistral agreement. Europe needs serious AI champions, and Mistral is among its few labs operating at globally relevant scale.

The July 21 agreement includes a multibillion-dollar Microsoft commitment to use Mistral’s expanded European GPU infrastructure. Mistral is adding thousands of Nvidia Vera Rubin GPUs; Mistral Medium 3.5 and OCR 4 are entering Microsoft Foundry, and Medium 3.5 is available in Copilot Studio.

Regulated industries need these options. Customers can run Mistral models in Azure’s public cloud, customer-controlled environments or fully disconnected through Azure Local. Disconnected systems support defense and critical infrastructure where external APIs are unacceptable.

Molto bene.

The weakness is clear: Mistral gets compute and global distribution; Microsoft keeps the main enterprise doorway through Azure, Foundry and Copilot Studio. Europe can build superb models yet rent customer access.

Mistral CEO Arthur Mensch said in the July 21 Microsoft-Mistral announcement that the partnership gives Mistral global access to enterprises and public institutions. He is right to take the distribution. I would too.

Europe must build beyond labs: cloud capacity, orchestration, enterprise channels and deployable infrastructure. A European model in an American platform’s dropdown provides choice. Sovereignty requires European ownership across more of the stack.

European Commission President Ursula von der Leyen set the direction when announcing InvestAI on February 11, 2025:

“We want AI to be a force for good and for growth. We are doing this through our own European approach, based on openness, cooperation and excellent talent.”

The European Commission said InvestAI would mobilize €200 billion for AI investment, including €20 billion for European AI gigafactories.

I passionately support it. But public funding must create companies that retain customer relationships. Otherwise Europe funds research, trains talent and supplies strategic models while an American hyperscaler distributes them and captures compounding product data.

Microsoft treats sovereignty as a product feature. European founders and policymakers must treat it as market structure.

Model choice can deepen platform lock-in

An 11,000-model catalog reduces reliance on one AI lab but can deepen reliance on Microsoft’s routing, identity, governance and application stack.

Swappable ingredients do not make the restaurant portable.

Nadella told investors every organization should build its own “continuous learning loop” and avoid outsourcing core intellectual property. Buyers should apply that advice to Microsoft.

Before committing a production agent to Foundry or Copilot, I’d ask five questions:

Where do our product-specific evaluations live?
Can we export agent memory and execution traces?
Who controls identity and permissions?
Can we replace the model without rebuilding the workflow?
Are we buying a measurable outcome, or financing a migration wearing an AI costume?

Ask the same in SAP and Oracle negotiations. A TechRadar Pro analysis by Chad Stewart noted that SAP promotes more than 200 specialized agents coordinated through roughly 50 domain-specific assistants. Oracle’s Fusion agents live in Fusion Cloud, requiring E-Business Suite customers to re-platform before using them.

Upgrades can consume the budget before agents prove value. TechRadar cited Americas’ SAP Users’ Group research in which 61% of members called budget their biggest challenge.

Lock-in enters through the component accumulating operational knowledge. The technology stays theoretically replaceable; its history becomes painfully sticky.

Enterprises should negotiate ownership of evaluations, traces, memories and permission mappings as fiercely as database exports. These assets form the learning loop; whoever controls them improves faster and collects rent longer.

By 2028, buyers will debate benchmark leaders less and ask who controls an agent after it joins the org chart.

OpenAI, Anthropic and Mistral will keep producing extraordinary intelligence. Microsoft is building the workplace where it gets hired, evaluated, permissioned and replaced.

My bet: enterprise software’s most powerful AI company will decide when the smartest model is worth paying for.

Right now, Microsoft is writing that decision into Excel.

Frequently asked questions

How does Microsoft make money from enterprise AI?

Microsoft monetizes enterprise AI at several layers: Azure compute, Foundry model deployment and governance, connections to company data, and Copilot inside workplace software. Enterprises can pay for infrastructure running AI and then pay again for AI features within Microsoft applications already covered by existing agreements.

Does Microsoft Foundry’s model choice reduce vendor lock-in?

Offering more than 11,000 models reduces dependence on any single AI lab, but customers can become more dependent on Microsoft’s routing, identity, governance, evaluation and application layers. Models remain swappable while execution traces, permissions, memories and product-specific learning loops accumulate inside the Microsoft platform.

What advantage does Microsoft have over independent AI model labs?

Microsoft can observe how models perform inside products such as VS Code, Excel, Dynamics 365 and Microsoft 365 Copilot. That product feedback provides evaluations based on acceptance, repeat use, workflow completion and cost, helping Microsoft decide when frontier models justify their price and when smaller specialized models can replace them.