By Beth Mayer, Hammerspace Content Marketing Manager
It’s difficult to find a news story these days that doesn’t enthusiastically proclaim the seemingly endless potential of artificial intelligence (AI) to unlock innovation, efficiency, and productivity. Yet many fail to examine the real-life adoption of AI in the enterprise – how are organizations actually deploying and scaling their AI solutions?
Hammerspace released a new report this week, “State of the Next Data Cycle: How do you GPU?” that found adoption of enterprise AI largely remains in the exploratory phase. While executives speak often about AI’s ability to improve their businesses, most use cases lie months or years in the future — or are entirely hypothetical.
But don’t get discouraged. Despite this “stark contrast” between AI’s future promise and its current progress, the report critically finds organizations are discovering how to leverage their massive GPU investments beyond AI projects to activate their unstructured data and discover unexpected ROI in non-AI applications. According to the report, “The potential is real, and possibilities are just beginning to unfold.”
Hammerspace CEO David Flynn said, “Our research shows that the GPUs many enterprises originally purchased for AI projects are becoming the Swiss Army knife of data processing. This infrastructure is unlocking value in ways we never expected across various sectors.”
So what does this mean?
The exploratory phase for AI mirrors previous ventures by organizations to unlock value from big data, namely structured data living in drives. However, structured data only accounts for approximately 10 percent of enterprise data – the remaining 90 percent is unstructured data. This has fueled demand for tools to make unstructured data readily accessible to improve both existing business applications and AI.
“This trend underscores the critical need for flexible data orchestration in the modern enterprise. As GPUs evolve into versatile tools, companies must be able to efficiently move and process their data, regardless of where it resides or what hardware is being used. The businesses that can do this effectively will be the ones leading the charge in innovation,” explained Flynn.
A large percentage of Hammerspace customers using GPUs do so not just for AI, but also for existing projects requiring substantial compute power to complete. For example:
- Meta has significantly scaled its infrastructure to support large-scale training for its Llama 2 and 3 models by deploying more than 24,000 GPUs, enabling the company to achieve faster, more efficient training of AI models.
- Los Alamos National Laboratory (LANL)’s hybrid supercomputer integrates CPU and GPU processing to address critical national security projects such as pandemic preparedness and climate change mitigation.
- A leading streaming provider uses the combined power of its GPUs and CPUs to improve the speed and accuracy of its content recommendation system, ensuring personalized, data-driven suggestions for each viewer at scale.
In Conclusion
Any developing technology has an exploratory phase and in the case of AI, this has unlocked unexpected benefits from GPUs. Enterprise GPU investments have proven to benefit other applications and enable organizations to tackle their most pressing challenges while building a foundation for future innovation.
Want to learn more?
Download the report today, or get started with a self-guided tour, a demo with your local account team, or request a trial license to deploy Hammerspace in your own data center or on the cloud platform of your choice.
