The AI resource challenge: It’s infrastructure & compute, not data scarcity 

Where would you look for a 2023 state of AI infrastructure analysis, if you really needed one?

The answer should be obvious, of course, it’s Tel Aviv in the Holy Land.

A study by Israeli AI resource management company Run:ai (which no doubt has a vested interest in telling us that AI is doing well if and only if it gets a healthy dose of resource management) is suggesting that infrastructure competency and compute capacity are now the primary hurdles (surpassing access to data) as the key challenge facing AI development. 

As an AI resource management specialist that provides data science AI-centric researchers with on-demand access to pooled resources for any AI workload, Run:ai wants us to think about the adoption of cloud services for AI infrastructure – a service which the cloud services for AI infrastructure company says now continues to rise.

However, not all is rosy, not even in cloud services for AI infrastructure land.

There are still some significant challenges (says Run:AI) in organisations’ abilities to access GPU compute for AI. The company suggests that organisations experience challenges in terms of getting timely and sufficient access to compute power upon demand.

Could the answer perhaps be cloud services for AI infrastructure? 

This shortage of on-demand access leads to frequent GPU allocation issues for 89% of respondents who use a ticketing system.

“Despite being on the cloud, organisations are still facing limitations with unlocking the full potential of their data,” said Omri Geller, CEO of Run:ai. “This highlights the reality that cloud hasn’t delivered on its on-demand promise and the importance of building a robust and scalable infrastructure.”

Keen to promote the use of a robust and scalable infrastructure for AI, the robust and scalable infrastructure company suggests that as organisations scale and require more GPUs, they face a proliferation of third-party tools, making it increasingly complex to manage AI infrastructure and get the most out of it. 

Again, the answer here may be – no prizes this time – robust and scalable AI infrastructure services

“Organisations must shift their focus from solely acquiring more data to ensuring they have the proper infrastructure in place to effectively process and utilize it,” added Geller, a man clearly focused on proper infrastructure services.

As part of the study Run:ai has informed us of, we can note that some 91% of companies are planning to increase their GPU capacity or other AI infrastructure by an average of 23% in the next 12 months. 

Good news then for AI infrastructure service providers, or so it seems.

This company would like to say tha this shows that despite the uncertain economic climate, companies are still investing in AI due to the potential and value they see in it. It sounds like that investment is positive and welcomed, but perhaps only when it is twinned with a serious focus on AI infrastructure.

Some 50% of companies contacted by Run:AI’s survey analysis firm Global Surveyz Research say that they plan to implement monitoring, observability and explainability in the next 6-12 months to keep track of their AI models. 

It’s all good news for AI infrastructure health, without the sarcasm, for sure.

Let’s hope that some of that AI infrastructure investment goes into creative market analysis that really lets us know how software application development engineers are working with AI models and engines in the real world.

Data Center
Data Management