What’s Next in AI and HPC for IT Leaders in Digital Infrastructure?

This year, we’re witnessing a monumental evolution of new artificial intelligence (AI) technology coming to the market which will turbo-boost demand for enterprise AI. These new developments will radically change the way we think about enterprise IT environments.

Three of the top minds at Digital Realty answered the industry’s most pressing questions about what’s coming next for AI, HPC, and the data that drives them both. Below you’ll find their reactions to three IDC FutureScape: Worldwide. Future of Digital Infrastructure 2024 Predictions* that apply to AI this year.

Note: Interview responses have been edited for length and clarity.

Prediction 1: Interwoven IT infrastructures to increase

According to the IDC FutureScape, “With GenAI as a catalyst, by 2027, 40% of enterprises will rely on interwoven IT architectures across cloud, core, and edge to support dynamic, location-agnostic workflow priorities.”

Where are the pockets of opportunity for IT leaders to successfully manage greater demand for AI and HPC in enterprises?

Patrick Lastennet, Director, Enterprise
Inference, the stage where a trained AI model is deployed and used to make predictions or decisions on new, unseen data, will be more adopted in 2024. We'll see more generative AI (GenAI) inference clusters being deployed in colocation, interconnected to enterprise wide-area networks (WANs).

Larger enterprises will adopt GenAI inference use cases like fraud protection and price prediction, through GenAI-driven pattern recognition. GenAI inference assists with coding, document summarisation, and chatbots.

Gaining the benefits of GenAI inference rely on a strong WAN rather than the cloud. Leveraging GenAI inference creates an expanded collection of, and access to, a company’s intellectual property. This visibility helps enterprises identify and close gaps in understanding of their offerings. To make the most of this opportunity, enterprises will utilise accurate and clean data to leverage this type of AI.

Enterprises will start to deploy clusters of GPUs for AI inference in proximity to their colocated, network data hubs running network ops.

Daniel Ong – Director, Solutions Architect, APAC
The AI nirvana for enterprises? In 2024, we'll see enterprises build ChatGPT-like GenAI systems for their own internal information resources. Since many companies' data resides in silos, there is a real opportunity to manage AI demand, build AI expertise, and cross-functional department collaboration.

This access to data comes with an existential security risk that could strike at the heart of a company: intellectual property. That’s why in 2024, forward-thinking enterprises will use AI for robust data security and privacy measures to ensure intellectual property doesn’t get exposed on the public internet.

They will also shrink the threat landscape by honing in on internal security risks. This includes the development of internal regulations to ensure sensitive information isn't leaked to non-privileged internal groups and individuals.

It’s a tough fine line between openness and data security and privacy. Firms that walk that line well are best positioned to leverage voice assistants, knowledge-based chatbots, and cybersecurity threat detection.

In North America, what change have we seen from our clients and how are they prioritising architecting their infrastructure to handle the uptick in AI demand?

Steve Smith, Managing Director, Head of Americas
For AI and other emerging technologies, it's all about optimising and future-proofing your business.

We see many North American companies using AI as a tool and lever point to make better decisions, optimise how they do business, and better deliver their services and products.

While many of our customers are early in trying to understand how AI is optimised, early findings show promising results and capabilities that allow dramatic efficiencies across the board.

From identifying where your customers are on the front end, to how you build the supply chain, to optimising services, AI gives insights into how our customers are performing and enhancing their operations.

Customers that partner with Digital Realty are looking at how they can access these large learning capabilities for operational efficiency. Companies that succeed start simple and demand a higher level of service from their partners, including: an increase in their understanding of how the enterprise’s business works, and how the company can apply insights gleaned from AI.

Prediction 2: Put as-a-service for critical workloads front and centre

According to the IDC FutureScape, “By 2028, 80% of IT buyers will prioritise as-a-service consumption for key workloads that require flexibility to help optimise IT spending, augment ITOps skills, and attain key sustainability metrics.”

What are some of the best ways to manage the global roll-out of AI initiatives?

Patrick Lastennet, Director, Enterprise
The first step for enterprise leaders is to identify all critical workloads and determine which ones to manage initially. Data privacy (sovereignty issues) and user experience (latency) will dictate localisation of critical workloads.

Next, they'll want to consider working with and connecting to partners that can offer global orchestration of AI-support infrastructure, partners such as:

• Software development and operations (DevOps): The focus here is on development orchestration via a single pane of glass management.
• Network operations (NetOps): A strong NetOps stance ensures that workflows function even in a fragmented environment. Home in on global orchestration of infrastructure behind the scenes. (ServiceFabric™ is a great place to start when looking to access a broad AI partner network.)

Options for smart workload placement include:

AI in the cloud: using the cloud for initial steps in AI provides agility and elasticity. Once you start to train more and more, you lose elasticity, or the ability to grow or shrink capacity for CPU, storage, memory and input/output bandwidth, and total cost of ownership (TCO) starts to become very important.

To scale on-prem, turn to providers of managed solutions to tap into GPU-as-a-Service providers to manage GPU outside of the cloud that offer cost management and the same level of cloud tooling.

Daniel Ong – Director, Solutions Architect
At this early stage of AI initiatives, enterprises are dependent on technology providers and their partners to advise and support with the global roll-out of AI initiatives.

In Asia Pacific, it’s a race to build, deploy, and subsequently train the right AI clusters. Since a prime use case is cybersecurity threat detection, working with the respective cybersecurity technology providers is key.

Top-level enterprises partner with their existing compute systems providers using GPU add-ons. Lenovo, Dell, and HPE offer this assistance.

How are North American enterprises leveraging partnerships and Digital Realty’s open ecosystem?

Steve Smith, Managing Director, Head of Americas
Enterprises can leverage Digital Realty’s open ecosystem, which allows them to strategically connect to their priority AI-as-a-service provider and choose the right services that fit their needs.

Enterprises should also consider:

1. Who can support the demands of the high-compute environment of AI workloads?
2. How do AI enablement partners get access to our compute environment?
3. How do companies‌ start using the services?

Enterprises find the answers to these questions in highly connected data centre campuses, such as those owned and operated by Digital Realty. These campuses are the central points for channelling AI-workflow data through an array of partners, using ServiceFabric™.

Prediction 3: Greater adoption of at-ingest data classification

According to the IDC FutureScape, “By 2028, 60% of IT organisations will adopt at-ingest data classification engines to improve data logistics efforts for enhanced governance, Data Loss Prevention (DLP), and data analytics for competitive business leverage.”

How should IT leaders approach distributed data in preparation for greater demand for computing resources?

Patrick Lastennet, Director, Enterprise
A solid approach to distributed data is based on tensor gravity. To optimise data ingestion, IT leaders need to work more on the provenance of data. Think of where you're going to end up once you industrialise your AI workflow. Ask yourself where all the zones of ingestion are going to be, and then bring the compute to the data rather than the opposite.

Daniel Ong – Director, Solutions Architect
The location of data affects the efficiency and effectiveness of your AI workflows. It affects the accessibility, security, and latency throughout the AI pipeline.

A few benefits to moving compute to where data is gathered include:

1. Reduced latency and more accuracy: Data closer to AI clusters helps improve results. This enables faster predictions and decision-making.
2. Network and energy cost savings: Data location and the respective network costs will affect energy savings and ultimately, TCO.
3. Data privacy and security compliance: Global enterprises use compute for region-specific data gathering to follow an ever-evolving landscape of data security regulations and security control measures.

How are IT leaders in North America approaching distributed data for greater demand for computing resources? How is Digital Realty set up to meet these compute-intensive workloads in North America?

Steve Smith, Managing Director, Head of Americas
In the past, AI was about large blocks of compute, ingesting large amounts of data, which enterprises would run models from, while drawing conclusions to iterate from that data.

Now, enterprises will pull in distributed data closer to make real-time decisions. The differentiating factor in a hyper-competitive landscape will be which organisations can best secure, verify, and use data in a large language model or data block to make better decisions than the competition.

Partnerships will turn capabilities into value

Which insights can we draw from a broader look at the three predictions we’ve addressed here?

Steve Smith, Managing Director, Head of Americas
The race to enable AI efficiencies comes down to how enterprises leverage partners that can turn capabilities into value. If you look at Digital Realty’s scale and capabilities, we offer a superior competitive advantage. From our global reach to breadth of platform, to connectivity, and our key metros across the world, we bring together people, companies, and data to make things work more efficiently.

We are making tremendous strides in how we architect, how we cool, and how we provide interconnection solutions, to make sure we aren't just relevant today but to enable future-proof businesses for years to come.

To learn more about how to future-proof your AI strategy, download our latest whitepaper: AI for IT Leaders.

*IDC FutureScape: Worldwide Digital Infrastructure 2024 Predictions, doc #US50401023, October 2023