Cloud and AI are often spoken about as though they go hand in hand. In fact, most cloud infrastructure is not able to support large-scale AI applications. Cloud providers are facing the challenge of bringing their offerings up to scale in order to reap the benefits of the AI gold rush, or risk losing out to the old guard of on-premises servers.
AI Needs More Than The Cloud Can Provide
AI models can be up to 100x bigger than previous kinds. Despite much of this technology being developed for the cloud, much of the cloud has not kept pace with this huge expansion and doesn’t have what it takes to support it.
With limited computing capacity and the global chip shortage, cloud providers are under pressure. Given the mass migration onto the cloud from leading businesses across a wide range of industries, it’s critical for their business that they bridge this gap.
The cloud has pitched itself as a perfect replacement for on-premise servers that could scale up or down and seamlessly step in with an efficient service. Despite its benefits, it turns out the cloud isn’t yet ready to perform at the scale that much of newest and increasingly popular technologies require. A lot of the cloud is run on general-purpose CPU chips, but very little runs on the kind of technology required for AI, that can facilitate bigger workloads.
At AWS for example, AI-optimised infrastructure makes up a small fraction of the company’s offering but they are working on growing this. AWS and their competitors are becoming wise to the urgent need for cloud providers to be ready for AI. Microsoft Azure and Google Cloud have also announced plans to improve their AI capabilities.
Chetan Kapoor, director of product management at Amazon Web Services’ Elastic Compute Cloud division told the Wall Street Journal…
“There’s a pretty big imbalance between demand and supply at the moment”
On-premises Here To Stay?
Some say that on-premises deployments might be necessary in order to fulfil the demand from AI. The cloud has appeared likely to ultimately replace on-premises servers for some time now. But perhaps AI, that once seemed like it’s natural accompaniment, will actually be its downfall.
John Roese, Dell’s Global Chief Technology Officer told the Wall Street Journal…
“The existing economic models of primarily the public cloud environment weren’t really optimized for the kind of demand and activity level that we’re going to see as people move into these AI systems”
The cloud has been growing exponentially in recent years, alongside AI. Now it’s becoming clear that AI may be outgrowing it. Given growing investment into the cloud and the widespread support for innovation within the industry, it is totally reasonable to expect that cloud providers are capable of giving AI what it needs. Do you think the cloud has what it takes?