Trend Forecast 2021: Artificial Intelligence Applied
AI means the possibilities are endless even if campus data centers are not
The idea of Artificial Intelligence (AI) dates back almost as far as modern man. Greek mythology incorporated the idea of an automated protector, and later, Mary Shelley endowed Dr. Frankenstein’s monster with it. More recently, in Stanley Kubrick’s 2001: A Space Odyssey, HAL 9000 thinks he knows best. Today, AI has moved beyond the arts and into the enterprise. Some 20 years after HAL 9000 was supposed to have put himself to his fullest possible use, we are only just beginning to see AI’s full potential.
So what does that mean for us in 2021?
AI (A)ppl(I)ed Even though modern AI and the infrastructure that it supports are still in their relative infancy, companies have quickly come to realize AI’s value. The problem is that most organizations don’t have the necessary compute power to take full advantage. In the not-too-distant past, people were using gaming GPUs for gaming and graphics rendering on video and photos. But forward-looking companies such as NVIDIA have evolved their product set to focus specifically on AI/ML/DL in the enterprise — for the smallest of startups to the largest of enterprises and federal agencies — and then built DGX, a GPU compute and ecosystem that sits on top of the hardware and that all work together like apps on an iPhone.
NVIDIA’s DGX platform is 12 to 18 months ahead of the market. They recognized early on that having a full-stack platform sitting on top of their hardware would be a key differentiation in bringing a holistic solution to the market. Cyxtera combines the security and control of dedicated infrastructure with the flexibility and speed of cloud. So, even as other GPU players in the market try to close NVIDIA’s lead, we think our new NVIDIA DGX-based solution will be a game changer, giving enterprises a flexible infrastructure model that leverages point-and-click provisioning via our data center services exchange.
Macro market movement Campus data centers are closing — AI in the cloud is expensive, and performance degrades at scale. Even if your campus DC stays open, running AI/DGX systems is extremely difficult due to the power and operating requirements. As a result, workloads have found their way to data centers naturally. Why? Because it’s hard — it’s hard to just deploy a single box, let alone at scale. It wasn’t that long ago when it was easy to crunch data in the campus data center because there just weren’t a lot of requirements. But now, the data sets needed to train more sophisticated algorithms require more power, more cooling, and, suddenly doing it on your own is difficult.
Scaling AI inside an enterprise-owned data center is so difficult that most people find their way to the public cloud pretty quickly. But once there, they run into two problems: First, it's expensive. Organizations pay a huge premium for virtualized GPU instances, and, the more they spin up, they find themselves with a case of diminishing performance returns — that’s a hard thing to sell to the CFO. Second, if you leave the public cloud, egress charges can be prohibitively expensive, and then you’re back to the question of whether you should go to a data center or try to do it on your own, which, as you already know, is hard … and expensive. At the end of the day, people want to run applications not infrastructure, so I think we’ll continue to see movement to hybrid architectures, where workloads run in a mix of public cloud and data center providers, where customers can fully leverage each provider’s platform’s respective strengths.
M(A)g(I)c 8 ball says ... When I to peer into my Magic 8 ball and ask about the future of AI, I’m certain the answer would be “Outlook good.” To paraphrase the Six Million Dollar Man: We have the technology. We have the capability … to build something better ... stronger ... faster (and cheaper). And, while it will be a little more of the same themes in terms of the evolution of hardware and the supporting platform, when it comes to what’s happening to the applications on top, I think we’ll see a telescopic evolution, as an example, with AI writing its own code while little pockets of innovation blossom around these large advancements in AI.
We continue to see increased use of AI in multiple sectors. In the federal government, for instance, especially among the secret “three letter” agencies. Financial services, too, especially where there’s high-frequency trading, will continue to utilize its power, along with the insurance sector for use in disaster modeling, for example. AI will continue to be enormously helpful within the health and life sciences — we’ve already seen success with Folding@Home, modeling proteins to cure certain diseases, and, there’s already large-scale adoption from the energy sector, where large oil and gas companies are using AI for everything from exploration to predictive analytics.
A(I)ding AI (A)dopt(I)on It’s incumbent on those of us in the data center space to not only provide the infrastructure that allows companies to run the interesting software but to ensure that everyone has access to it. When you consider that entry to a super computer is in the millions on the low end, that’s a pretty big barrier to entry for a startup. I think in the not-too-distant future, we will start to see data centers granting short-term access to AI infrastructure to democratize access for those who might not have been able to afford it otherwise. Evolving company offerings in the data center is the right answer, whether we’re talking about a large enterprise that can afford it, a company that’s CapEx constrained, or a young startup. Having a fully baked platform gives enterprises an alternative for those times when going to a public cloud entity isn’t the best (or desired) route. The future will be all about presenting options that allow people to execute their ideas.
This is the second in a series of trend pieces. Check back in the coming weeks to see what else we think is in store for data centers in 2021.
Views and opinions expressed in our blog posts are those of the employees who made them and do not necessarily reflect the views of the Company. A reader should not unduly rely on any statements made therein.