While the whole world is struggling to get Nvidia GPUs, Microsoft, the world’s 2nd most valuable company by market cap, is struggling with another problem of getting sufficient power to use its Nvidia Hopper GPUs.
According to a recent news report from Omdia, Microsoft has bought over 485,000 Nvidia Hopper GPUs alone in 2024 and approximately contributed to 20% of NVIDIA’s revenue alone.
If we approximate the cost of one Nvidia Hopper GPU to be $30,000, this purchase will come to somewhere around a whopping $15 billion. I mean, it is huge.
Also Read: Nuclear Power is the Future of AI Energy Consumption Solutions
What Are Nvidia Hopper GPUs?
Named after computer programming pioneer Grace Hopper, Nvidia Hopper GPUs represent the 9th generation GPU architecture in Nvidia’s lineup.
It was introduced in March 2022, approximately two years after Nvidia’s Ampere architecture, which powered the famous Nvidia A100 GPUs.By the way, Nvidia Hopper GPUs came up with some awesome numbers.
Feature | Details |
|---|---|
Transistor Count | More than 80 billion |
Fabrication Process | TSMC 4N |
Flagship Models | H100 and H200 GPUs |
Additionally, they include five key innovations:
Transformer Engine for optimising large language models
NVLink, NVSwitch, and NVLink Switch System
NVIDIA Confidential Computing for secure data processing
Second-Generation MIG for Workload Partitioning
DPX Instructions
For more details, you can visit Nvidia’s official website.
How Will Microsoft Use These 485,000 Nvidia Hopper GPUs?
Microsoft will primarily use these Hopper GPUs to enhance their AI infrastructure across several key areas like:
1) Azure Cloud Services:
These GPUs power Azure’s AI capabilities, providing infrastructure for enterprises to train and deploy large AI models efficiently.Also, since Microsoft is aiming to attract businesses building AI applications by offering unparalleled performance for workloads like machine learning and data analytics. Who knows Microsoft might end up using a big chunk of these GPUs there.
2) Collaboration with OpenAI:
Microsoft has been a major investor in OpenAI, and OpenAI uses Hopper GPUs to support the development and scaling of advanced models like ChatGPT and DALL·E, as these GPUs facilitate faster training and inference of OpenAI’s models, enabling real-time improvements and deployment.
3) Enterprise AI Applications and Generative AI Expansion:
Recently Microsoft has been embedding generative AI tools across its ecosystem, like Microsoft Teams, Dynamics 365, and Edge browser AI features. Some of these GPUs might end up there, contributing to these AI capabilities, helping them to scale.
Also, Microsoft uses Hopper GPUs in Microsoft’s enterprise offerings, such as Copilot for Office 365, etc., so yes, it can pave its way to enterprise AI applications.
4) AI Research and Innovation:
Hopper GPUs had been a big helping hand to Microsoft’s research labs. Microsoft Research Labs generally work on:
- Reinforcement Learning
- Neural Architecture Search
- Multimodal AI Systems
How Many NVIDIA Hopper GPUs Did Other Companies Buy?
Based on data from Omdia, here’s a breakdown of Hopper GPUs purchased by tech giants:
Company | GPUs Purchased |
Microsoft | 485,000 |
Bytedance | 230,000 |
Tencent | 230,000 |
Meta | 224,000 |
xAI/Tesla | 200,000 |
Amazon | 196,000 |
169,000 |
Clearly, Microsoft bought roughly twice as many Nvidia Hopper GPUs as its competitors.
From this data, it’s evident that Chinese companies, such as Bytedance and Tencent, are also heavily investing in Nvidia GPUs.
Conclusion:
In recent times, Microsoft has been heavily investing in AI services, and these purchases from Microsoft just reinforce their commitment to AI services. Also, these purchases somewhere deny the rumours that AI growth has slowed down; instead, these numbers are saying just the opposite.
Discover more from WireUnwired
Subscribe to get the latest posts sent to your email.
