Kaladhar Voruganti, Senior Business Technologist at Equinix: Do companies have the skills and infrastructure to manage the impact of AI?
The advent of ChatGPT led to generative AI becoming mainstream. Thus prompting a great many organizations to focus onacceleration of AI initiatives to better serve customers, employees and partners while streamlining internal workflows.
The impact of AI: Information infrastructure not up to par
Nonetheless, it is also worth asking: Will these organizations have the skills, systems and infrastructure to manage the huge impact that AI will have on their operating models? The 2023 edition of Equinix’s Global Tech Trends Survey (GTTS) unearthed a clear answer. 42% of corporate IT leaders believe their IT infrastructure is not fully prepared to accommodate the growing adoption of AI. Additionally, 41% doubt their team’s ability to implement the technology.
Implementing AI in production introduces new challenges
IT teams starting to support the use of AI technologies in their organizations are faced with a number of challenges. Entirely new challenges in terms of cost, performance, data sharing, skills shortages and sustainability. Starting from the economic sphere, a source of worry for businesses, the result is that by the middle of this decade, most data will be generated outside the data center.
Some examples
For example, if the data is generated in the cloud, it makes sense to process it in the cloud. However, if the data is generated at the edge, it must be stored and processed at the edge. Therefore, centralized AI architectures are not scalable in terms of cost and performance. The cost of backhauling data generated at the edge and sent to the core can be prohibitive.
Equinix – The impact of AI
Organizations can also encounter barriers to achieving optimal performance for a variety of reasons. Such as difficulty accessing the latest GPU technology in the cloud, edge inference latency/throughput. And again the differences between system and implementation architecture, since while GPU vendors, OEMs, and clouds may all use the same type of GPU, there will be a difference in the overall performance of these implementations. And that’s because of the interconnect architecture of GPUs with networking, storage, and other GPUs in the cluster.
Don’t violate compliance rules
And yet, there are also challenges related to data sharing. Since, in many cases, organizations need to leverage external data (e.g. weather data, traffic data, etc.) to improve the accuracy of their AI models. Thus forcing themselves to know the path of the data and the external models they are using to ensure that they do not violate compliance regulations and to protect themselves from corrupted data manipulated by malicious agents.
Equinix – the main challenges: skills and sustainability
Yet, one of the biggest barriers is the difficulty to assume skilled workers in the field of AI. So much so that in the GTTS, 45% of IT leaders said the biggest skills challenge is the speed of transformation of the technology sector. Finally, sustainability.
Expertise and infrastructure to manage the impact of AI
AI training racks consume more than 30kVA per rack and air cooling becomes inefficient; higher kVA per rack requires liquid cooling; and most private (in-house) data centers are not equipped to handle these energy-intensive AI racks.
Are we ready for AI?
One thing is certain: building scalable AI solutions will require companies to be able to handle the ingestion, sharing, storage and processing of huge and diverse datasets. While taking into account sustainability. In other words, they will have to achieve production-level AI. And, from this point of view, participation in a digital ecosystem will be more and more fundamental. Helping IT infrastructure teams overcome the complexity of AI, manage massive volumes of data, and gain access to new technology partners with innovative solutions.