Enhancing Scientific Research with AI
Microsoft Discovery is set to integrate NVIDIA’s ALCHEMI NIM microservice, a pivotal step in optimizing AI inference for complex chemical simulations. This integration is poised to dramatically accelerate research within materials science by enabling more precise property prediction and more effective candidate recommendation. The resultant streamlining in identifying novel materials with highly specific characteristics cuts down significantly on the time and resources traditionally needed.
Furthermore, Microsoft Discovery will begin to use NVIDIA BioNeMo NIM microservices. These are specifically made to harness pre-trained AI workflows so that the AI model development process for drug discovery can be sped up significantly. By leveraging these powerful tools, researchers can substantially accelerate the development and refinement of AI models that can predict drug efficacy and potential side effects, making drug development both faster and more effective.
These integrations are meticulously designed to empower scientists and researchers by providing accelerated performance, which cuts down on the time required for significant scientific breakthroughs. They also ensure enhanced capacity to handle vast datasets and intricate simulations with unprecedented speed and accuracy, thus enabling researchers to tackle some of the most pressing challenges currently confronting scientific research.
A real-world application of these remarkable advancements was recently demonstrated by Microsoft researchers, who successfully utilized Microsoft Discovery to pinpoint a new coolant prototype that demonstrated promising properties for immersion cooling in data centers in less than 200 hours, a stark contrast to the months or even years typically required by conventional methods. Immersion cooling, which entails submerging electronic components in a non-conductive liquid coolant, is becoming increasingly essential to effectively manage the significant heat generated by high-performance computing systems.
NVIDIA Blackwell GPUs in Azure Data Centers
In a significant move to bolster both performance and efficiency, Microsoft is rapidly deploying hundreds of thousands of NVIDIA Blackwell GPUs within its AI-optimized Azure data centers globally. These GPUs are integrated within NVIDIA GB200 NVL72 rack-scale systems, meticulously engineered to handle the very most demanding AI workloads with exceptional proficiency.
Several of Microsoft’s key clients, OpenAI included, are currently running their production workloads on this advanced infrastructure. The deployment of NVIDIA Blackwell GPUs enables these organizations to execute complex AI tasks more efficiently and effectively, showcasing Microsoft’s unwavering commitment to providing its clientele with cutting-edge AI capabilities.
Azure’s ND GB200 v6 virtual machines represent a significant leap forward in computational power, delivering up to 35 times more inference throughput compared to the previous ND H100 v5 VMs. The older VMs, accelerated by eight NVIDIA H100 GPUs, now pale in comparison to the enhanced performance offered by the new generation, marking a new benchmark for AI workloads. This radical improvement can dramatically reduce the time and cost associated with running large-scale AI models.
This impressive scale and high performance are supported by custom server designs, high-speed NVIDIA NVLink interconnects, and NVIDIA Quantum InfiniBand networking. These technologies facilitate seamless scaling to thousands of Blackwell GPUs, critical for handling demanding generative and agentic AI applications. The sophisticated interconnectivity ensures low-latency communication between GPUs, enhancing overall system performance.
Satya Nadella, Microsoft’s chairman and CEO, and Jensen Huang, NVIDIA’s founder and CEO, have emphasized that their collaboration is resulting in substantial performance gains through ongoing software optimizations across all NVIDIA architectures on Azure. This strategic approach is meticulously designed to maximize developer productivity, reduce the total cost of ownership, and accelerate all workloads, including AI and data processing, ultimately leading to greater efficiency per dollar and per watt for customers.
Expanding Capabilities with NIM Integration
Building upon the already significant NIM integration in Azure AI Foundry, Microsoft and NVIDIA are further expanding the platform with the NVIDIA Llama Nemotron family of open reasoning models and NVIDIA BioNeMo NIM microservices. These are specifically crafted to deliver enterprise-grade, containerized inferencing for complex decision-making and highly specialized domain-specific AI workloads.
Developers can now leverage optimized NIM microservices for advanced reasoning within Azure AI Foundry. This includes the NVIDIA Llama Nemotron Super and Nano models, designed to provide advanced multistep reasoning, coding, and agentic capabilities. Offering substantially improved accuracy—up to 20% higher—and five times faster inference compared to previous models, developers are empowered to create more sophisticated and efficient AI applications.
BioNeMo NIM microservices, tailored specifically for healthcare applications, address critical needs in digital biology, drug discovery, and medical imaging. They enable researchers and clinicians to accelerate protein science, molecular modeling, and genomic analysis, leading to improved patient care and faster scientific innovation. These sophisticated tools empower healthcare professionals to make more informed decisions and develop more effective treatments.
This expanded integration enables organizations to rapidly deploy high-performance AI agents. By connecting to these models and other specialized healthcare solutions, organizations can achieve robust reliability and simplified scaling, thereby meeting the diverse needs of various industries and applications.
Generative AI on RTX AI PCs
Generative AI is revolutionizing PC software by introducing entirely new experiences ranging from sophisticated digital humans to insightful writing assistants, intelligent agents, and powerful creative tools. NVIDIA RTX AI PCs facilitate experimentation with generative AI and significantly improve performance on Windows 11, thereby making state-of-the-art AI accessible to a far broader audience than ever before.
At Microsoft Build, NVIDIA and Microsoft unveiled an AI inferencing stack designed to simplify development and enhance inference performance for Windows 11 PCs. This invaluable toolset is critical for enabling a seamless AI experience on personal computers, making AI tools more responsive and efficient.
NVIDIA TensorRT has been completely redesigned specifically for RTX AI PCs. It uniquely combines TensorRT performance with just-in-time, on-device engine building, and an eight-times smaller package size for seamless AI deployment to over 100 million RTX AI PCs. This comprehensive optimization enables faster and more efficient AI processing on PCs, paving the way for entirely new applications and capabilities.
Announced at Microsoft Build, TensorRT for RTX is natively supported by Windows ML – a new inference stack that provides app developers with broad hardware compatibility and state-of-the-art performance. TensorRT for RTX is available in the Windows ML preview starting today and will be available as a standalone software development kit from NVIDIA Developer in June. This significant development simplifies the process for developers looking to integrate AI capabilities into their Windows applications, ensuring that AI is accessible to a wide range of software solutions.
In essence, the collaboration between NVIDIA and Microsoft is creating a synergistic ecosystem where advancements in AI technology are rapidly translated into real-world applications, benefiting researchers, developers, and end-users alike. This strategic partnership is strategically positioned to maintain its leadership in the rapidly evolving field of artificial intelligence.
Detailed Breakdown of Advancements
Cutting-Edge Technologies
The collaboration between NVIDIA and Microsoft leverages several cutting-edge technologies to achieve advancements in agentic AI. Here we delve into these key components.
NVIDIA ALCHEMI NIM Microservice: This microservice acts as a highly specialized tool that is meticulously optimized for AI inference in complex chemical simulations. Its primary function is to dramatically accelerate materials science research through highly accurate property prediction and efficient candidate recommendation. By enabling faster and more precise simulations, it empowers researchers to identify promising materials far more quickly and effectively than traditional methods allow.
NVIDIA BioNeMo NIM Microservices: These microservices supply pre-trained AI workflows designed to significantly speed up AI model development specifically for drug discovery. Researchers can leverage them to rapidly develop highly effective models that predict drug efficacy and potential side effects, thereby substantially accelerating the often-lengthy process of new drug development.
NVIDIA Blackwell GPUs: These advanced GPUs deliver drastically enhanced performance and unparalleled efficiency in AI workloads within Azure data centers. Fully integrated into rack-scale systems, they effectively support clients such as OpenAI in smoothly and effectively running highly complex tasks.
NVIDIA NVLink Interconnects: These high-speed interconnects function to ensure extremely low-latency communication between GPUs, dramatically enhancing overall system performance. This results in faster computation and notably improved efficiency across a wide range of AI operations.
NVIDIA Quantum InfiniBand Networking: This networking system supports seamless scaling to thousands of Blackwell GPUs, making it absolutely critical for handling demanding generative and agentic AI workloads with exceptional performance. The advanced networking capabilities ensure that large-scale AI models can be effectively deployed and managed without performance bottlenecks.
NVIDIA Llama Nemotron Models: The NVIDIA Llama Nemotron Super and Nano models are meticulously designed to provide highly advanced multistep reasoning, coding, and agentic capabilities. The improved accuracy and far faster inference speeds empower developers to create significantly more sophisticated and efficient AI applications than previously possible.
Impact on Scientific Research
The seamless integration of NVIDIA’s advanced technologies into Microsoft’s comprehensive platforms has profound implications for scientific research across multiple critical disciplines.
Materials Science: NVIDIA ALCHEMI NIM microservice makes possible highly precise property prediction and more effective candidate recommendation, enabling the significantly faster identification of novel materials with highly desired characteristics.
Drug Discovery: The BioNeMo NIM microservices dramatically accelerate AI model development, empowering researchers to develop highly precise models that effectively predict drug efficacy and potential side effects with unprecedented speed.
Data Center Cooling: Microsoft deployed Microsoft Discovery to detect a novel coolant prototype ideal for immersion cooling in data centers in less than 200 hours instead of the potentially months or years that would have been required with traditional laborious methods. This underscores the dramatic acceleration of scientific discoveries now possible with these tools.
Protein Science, Molecular Modeling, and Genomic Analysis: BioNeMo NIM microservices enable significant acceleration in each of these domains. They can lead to improved patient care and much faster scientific innovation, transforming healthcare research and development.
Azure AI Infrastructure
Microsoft’s Azure AI Foundry and its state-of-the-art data centers represent massive, strategic investments in creating an optimal environment for seamlessly running AI workloads on a global scale.
ND GB200 v6 Virtual Machines: They deliver up to 35 times more inference throughput compared with previous ND H100 v5 VMs, setting a completely new performance benchmark for AI workloads of all kinds.
Custom Server Designs: Custom server designs engineered to maximize performance and efficiency, allow the Blackwell GPUs to fully operate at their ultimate potential without any performance limitations.
NVIDIA Optimizations on Azure: Continuous software optimizations conducted across all NVIDIA architectures running on Azure serve to maximize developer productivity, dramatically lower the often-substantial total cost of ownership, and relentlessly accelerate all workloads, thereby enhancing overall efficiency per dollar and watt for all customers.
Generative AI on Personal Computers
The remarkable advancements in AI technology are also rapidly finding their way into personal computers, driving entirely new possibilities for revolutionary software applications and user experiences.
NVIDIA RTX AI PCs: Vastly simplified experimentation with generative AI and significant performance enhancement on Windows 11 are efficiently facilitated through the use of NVIDIA RTX AI PCs. They make state-of-the-art AI technology vastly more accessible to an increasingly large and diverse audience.
NVIDIA TensorRT: This state-of-the-art software development kit (SDK) has been meticulously optimized for use with RTX AI PCs. It combines high-performance with an eight-times smaller package size for far more seamless AI deployment. This simplification makes it significantly easier for developers to effortlessly integrate advanced AI features into their new and existing applications.
Windows ML Support: Native support for TensorRT directly in Windows ML further ensures broad hardware compatibility and state-of-the-art overall performance. This facilitates the totally seamless integration of AI directly into Windows applications.
New User Experiences: From remarkably lifelike digital humans to exceptionally insightful writing assistants, intelligent agents, and remarkably creative tools—generative AI is radically reshaping PC software and introducing entirely new captivating experiences. Users can now benefit extensively from far more interactive, intelligent, and endlessly creative applications.
The Strategic Vision
The highly collaborative partnership between NVIDIA and Microsoft is firmly built on a clear strategic vision, with the core aim being to lead the ongoing advancement of AI technology across a remarkably wide range of dynamic sectors. These synergistic collaborative efforts and the jointly developed technologies are all meticulously designed to accelerate the crucial adoption of AI across differing domains, significantly benefiting not only researchers and developers, but also end-users and pioneering organizations all across the globe.
Innovation: A continuous and relentless emphasis on driving innovation through strategic collaboration is dramatically accelerating transformative technological advancements. This helps to consistently maintain aposition of leadership in the highly and rapidly evolving field.
Accessibility: NVIDIA and Microsoft are effectively democratizing the use of AI by making it seamlessly accessible to a wide array of developers and end-users through highly optimized and easy to use tools, and through strategic integration with widely popular platforms, and through continuous improvements in cost-efficiency.
Performance & Efficiency: A highly focused and unwavering focus on dramatically enhancing both overall performance and crucial cost-efficiency guarantees that the remarkable benefits of AI technologies are readily and widely available to a highly diverse range of users, starting from highly individual researchers up to major enterprises with a global presence.
Real-World Applications: By consistently translating the often-arcane AI advancements into highly practical real-world solutions, NVIDIA and Microsoft are driving highly tangible benefits and fundamentally transforming numerous industries around the world, creating major and positive global influence.