Microsoft integrates more affordable DeepSeek R1 model across Azure AI Foundry and Copilot+ PCs

Microsoft revealed a sweeping AI partnership with Chinese firm DeepSeek on January 31, 2025, integrating its R1 language model into core enterprise and consumer platforms. The strategic move positions Azure as the first major Western cloud service to deploy Chinese-developed DeepSeek R1 AI at scale, offering developers unprecedented cost savings and hardware flexibility.

Azure AI Foundry

The Azure AI Foundry now features DeepSeek-R1 alongside 1,800 other models, enhanced by new developer tools:

1. Low-code customization via drag-and-drop workflows reduces fine-tuning time from weeks to hours.
2. Military-grade security with automatic compliance for FedRAMP High and DoD SRG IL5 standards.
3. Unified monitoring dashboard tracking model performance, data drift, and API costs in real-time.

Early adopters include JPMorgan Chase, which reduced fraud detection AI deployment time by 73% using these tools. Microsoft’s GitHub Copilot will also incorporate R1 next month, providing coding suggestions trained on 30% more Python repositories than current models.

Copilot+ PCs: Local AI processing leap

Microsoft optimized DeepSeek’s 7B and 14B parameter models for Qualcomm’s Snapdragon X Elite processors, enabling:

1. 43 tokens/second throughput on Surface Laptop 7 prototypes – 3× faster than Apple’s M3 chips.
2. Offline functionality for sensitive tasks like medical record analysis.
3. Cross-platform support coming to Intel Lunar Lake and AMD Ryzen AI 300 chips by Q2 2025.

This addresses growing enterprise demand for edge AI, with Microsoft claiming 58% latency reduction versus cloud-only setups in manufacturing quality control tests.

DeepSeek API pricing: 98% cost reduction over OpenAI

DeepSeek’s API pricing through Azure disrupts the AI market:

Metric	DeepSeek R1	OpenAI GPT-4 Equivalent
Input Tokens	$0.14/M	$7.00/M
Output Tokens	$0.28/M	$14.00/M
Fine-tuning Cost	$0.003/GPUhr	$0.48/GPUhr

Analysts estimate this could save Fortune 500 companies $4.7B annually. During a Wall Street Journal interview, CEO Satya Nadella stated: “We’re not here to undercut partners like OpenAI, but to expand what’s possible when AI becomes truly economical.”

Cloud wars escalate

Within hours, AWS added R1 to SageMaker JumpStart, mirroring Microsoft’s pricing. However, Azure retains advantages through:

1. Hybrid deployments syncing cloud and Copilot+ PC AI instances.
2. Pre-negotiated enterprise contracts locking in rates through 2026.
3. China-specific instances already operational in Beijing and Shanghai regions.

Gartner projects this integration could capture 19% of China’s cloud AI market within 18 months, leveraging DeepSeek’s local regulatory expertise.

Get started

DeepSeek is now available via a serverless endpoint through the model catalog in Azure AI Foundry. Get started on Azure AI Foundry here.

On GitHub, you can explore additional resources and step-by-step guides to integrate DeepSeek seamlessly into your applications. Read the GitHub Models blog post.

Customers will be able to use distilled flavors of the DeepSeek R1 model to run locally on their Copilot+ PCs. Read the Windows Developer blog post.

Related Posts

1. Microsoft surpasses Q2 2025 expectations with strong cloud and AI performance
2. Surprise DeepSeek AI App Shocks Tech Markets: Microsoft and Nvidia Stocks Plunge Amid Fierce China AI Tool Competition
3. DeepSeek R1 7B & 14B models now available for Copilot+ PCs expanding capabilities with Azure AI Foundry