NicoElNino - Fotolia
Most organisations just don’t have the IT infrastructure to handle the compute, storage and energy requirements that artificial intelligence (AI) deployments bring. In fact, one-third think a total overhaul of their IT estate will be needed to be able to handle AI workloads.
Storage, compute and networking will all come up short, with data management and security also likely to be found wanting. On top of this, C-suites don’t really understand what’s needed to make a shift to AI-centric working.
That’s all according to a survey by storage supplier Pure Storage, which questioned 500 IT decision makers in the UK, US, France and Germany in companies of at least 500 employees.
In the survey – which Pure Storage carried out with partner Wakefield Research – 88% of those questioned found adoption of AI saw a big increase in compute power requirements. Nearly three-quarters (73%) found themselves unprepared for increased energy requirements and nearly half (47%) had to increase compute power by two times or more since deploying AI.
For significant numbers of respondents – getting on for half of them in each case – AI adoption had significant knock-on effects on other parts of the IT infrastructure.
On the hardware side, adoption of AI required upgrades to data storage for 46% of respondents, while 43% said they had needed to improve compute, and 44% needed to upgrade networking infrastructure.
Meanwhile, adoption of AI affected softer processes and capabilities. Nearly half (48%) had been obliged to improve data management tools as well as data management processes (46%). At the same time, security and privacy tools and processes had been upgraded by 44% of those questioned.
Read more on AI and storage
- Generative AI: Automating storage and backup management. We look at how generative artificial intelligence is being deployed to automate monitoring of repetitive tasks and report on enterprise storage and backup, but also the limits to its implementation.
- Generative AI: Data privacy, backup and compliance. We look at generative AI and the risks it poses to data privacy for the enterprise, implications for backup, and potentially dangerous impacts on compliance.
Sustainability is also a key issue, with pretty much all respondents (99%) reporting they face pressure to use providers that commit to sustainability in their products.
That translated into a perceived need to prepare the IT infrastructure to support AI, with 96% saying an upgrade to IT infrastructure is planned or underway and 29% believing a total overhaul will be required.
Among other consequences of not being prepared for the impact of AI on IT infrastructures were more pressure on IT departments to address the issues flagged by data teams, increased investment required to upgrade infrastructure to handle AI and even an inability to effectively use AI. These were consequences indicated by around half of respondents in each case (51%, 49% and 48%).
Getting past the C-suite
To smooth the path to infrastructure upgrades to accommodate AI workloads, IT leaders need to convince the board. Here, a number of difficulties were identified by respondents.
Chief among these is the expectation by C-level leaders – reported by 51% – that AI work would be done in the cloud. This is not always possible or desirable for reasons of cost, performance and availability.
Other reasons cited were that the board had a narrow view of the impacts of AI (50%), or perhaps conversely, a rush by the C-suite to adopt it (48%). Two-fifths (41%) said they thought the business leadership doesn’t understand the current IT infrastructure.
When it came to solutions to offset the increased energy usage of AI workloads, 61% said they would look to invest in energy-efficient datacentre hardware, while 49% said they would consider shutting down hardware when not in use.