The Pressure Cooker That is Cloud Infrastructure
Let’s be real. The cloud’s everywhere. It’s everyone’s darling when it comes to training and running hefty AI applications. It’s supposedly the superhero we didn’t know we needed. But is it really? In reality, only a tiny bit of existing cloud infrastructure is set up to handle all that bravado. The rest is about as useful as a chocolate teapot.
The Big Three Scramble for a Solution
Sure, the bigwigs like Amazon Web Services, Microsoft Azure, and Google Cloud are scurrying to shift the odds. They’re trying to beef up their infrastructures to handle the ever-growing demand for AI, as other hardware providers eye a piece of the action. Generative AI models, designed to churn out original text and analysis, are making a feast of the cloud. As Ziad Asghar, senior vice president of product management at Qualcomm Technologies so eloquently put it, “There is insatiable demand.”
The Strain of The New Demand
Now the demand for AI isn’t just a whisper in the wind anymore. It’s an uproarious thunderclap in sectors like manufacturing and finance. This increased demand is putting unprecedented pressure on the limited computing capacity available. The big brands, Visa, Johnson & Johnson, Chevron are all on board with this trend, anticipating using cloud providers for their AI-related escapades.
But here’s the snag. Much of the cloud’s infrastructure wasn’t built for running such large and complex systems. The whole concept of cloud was to serve as a handy replacement for on-premise servers that could easily scale up and down capacity with a pay-as-you-go model.
The Cloudy Picture of The Future
In the near future, companies like AWS plan to deploy more AI-optimized server clusters. However, this doesn’t mean the company is necessarily moving away from the shared server—general-purpose computing—which is still valuable for companies.
Then there’s the likes of Dell Technologies expecting high cloud costs linked to heavy usage to push some companies to consider on-premise deployments. But before you think it’s all doom and gloom for the cloud, they argue that on-premises deployments could end up costing more in the long term because of the hefty investments needed for hardware upgrades.
Companies like Qualcomm and Hewlett Packard Enterprise are also exploring alternative avenues. Qualcomm suggests it might be cheaper and faster for companies to run models on individual devices, taking some pressure off the cloud. Meanwhile, HPE is rolling out its own public cloud service, powered by a supercomputer, available to enterprises looking to train generative AI models.
A Silver Lining?
At the end of the day, it’s clear as mud. The future could very well be a hybrid of computing on the cloud and individual devices. But, what’s certain is the fundamental raison d’être of the cloud is changing. As Amin Vahdat, vice president and general manager of machine learning, systems and Cloud AI at Google Cloud sums it up, “It’s really a phase change in terms of how we look at infrastructure, how we architected the structure, how we deliver the infrastructure.”
In short, the landscape is shifting under our feet. Let’s just hope we don’t fall into any sinkholes.
