Storage, not GPUs, is AI’s biggest challenge, says influential report — lack of bits and bytes is what can make or break your AI/ML project
A major report on the state of the AI industry has provided some key insights into potential issues for businesses and the sector as a whole as its growth continues to accelerate following the introduction of ChatGPT and other AI models.
WEKA recently completed its research on behalf of S&P Global Market Intelligence Global Trends in AI for 2024 report, which surveyed over 1,500 different tech executives and decision makers about underlying trends across a wide range of topics.
Some of the most interesting insights revolve around the array of AI-capable GPUs, nearly all of which are made by Nvidia, though companies like AMD and startups are also gaining ground.
The key takeaway, according to WEKA, is that AI applications “are now ubiquitous in the enterprise,” which is likely something most employees can already sense. However, scaling these AI apps is often difficult, largely due to legacy data architectures.
Generative AI has also sent organizations into a frenzy. The report notes that 88% of organizations are “now actively researching generative AI, outpacing other AI applications such as predictive modeling (61%), classification (51%), expert systems (39%), and robotics (30%).”
Nvidia, the business world needs you
High-end GPUs are of course a key focus of the report: having the raw power is fundamental to actually being able to do something with many AI applications.
“Hyperscaler public clouds are one path to GPUs, but many are also turning to specialized AI clouds,” it notes. “GPU clouds are emerging as a key location for both training – used by nearly a third, 32% of organizations – and inference, 31%.”
All this interest is only increasing orders for Nvidia chips.
Interestingly, WEKA highlights the issue of GPU scarcity in several major economies in the Asia-Pacific region, including India. “India, Taiwan, New Zealand and Australia are more likely to rank GPU availability as one of their top three challenges to bringing a model into production,” the report’s authors said.
“One of the most striking takeaways from our 2024 Trends In AI research is the astonishing pace of change that has occurred since the introduction of ChatGPT 3 and the first wave of generative AI models hit the market in early 2023,” Chief Research Analyst John Abbott told Blocks and files.
“In less than two years, the adoption of generative AI has eclipsed all other AI applications in the enterprise, creating a new breed of AI leaders and an emerging market of specialized AI and GPU cloud providers.”