Despite initial concerns over power consumption in the rush to adopt generative AI, research shows a growing emphasis on sustainable practices and energy-efficient strategies among businesses.
The emergence of generative AI (gen AI) had initially sent shockwaves across the business world, challenging long-held commitments to sustainability and efficient computing. This transformative technology necessitated the deployment of advanced computing infrastructures, seemingly putting traditional power conservation efforts on the back burner. However, emerging trends indicate a resurgence in sustainable practices, even amidst the ongoing AI revolution.
According to research conducted by Aberdeen in 2023, there was a slight decline in commitments to power-efficient computing as companies initially rushed to adopt powerful AI technologies. The scarcity of high-performance GPUs and the rising demand for powerful graphics chips underscored this shift, with some smaller companies resorting to repurposing gaming systems for AI tasks. This need for power-intensive infrastructure threatened to undermine efforts in sustainability and zero-carbon commitments.
However, recent shifts in how businesses approach AI deployment offer a more balanced path forward. Aberdeen’s latest survey reveals a pragmatic approach to gen AI is gaining ground, focusing on internal data usage and small custom language models. This transition indicates that building and deploying AI doesn’t necessarily require expansive, power-intensive systems. Surprisingly, some businesses have successfully developed AI models on a single engineer’s laptop, showcasing the potential for minimal power consumption.
The survey indicates that while 90% of businesses are using AI in some form, there’s an increasing preference for efficiency-focused strategies. Around 25% of organisations are making strategic investments in AI, a number predicted to rise by over 20% in the coming months. Despite the initial power consumption concerns, there is now a 10% increase in attention towards energy efficiency, suggesting that businesses recognise the dual benefit of leveraging AI while maintaining power conservation goals.
Notably, GPUs are no longer the primary technology focus for AI initiatives, dropping to fourth place in purchase preferences. Instead, increased storage capacity and hybrid cloud capabilities have become priorities, supporting the need to manage and store large datasets required for AI models. Hybrid cloud frameworks are proving valuable, combining the flexibility of cloud technology with the security and performance of on-premise solutions.
While concerns around power consumption remain for the major AI vendors like Amazon, Microsoft, and Google—who are securing power from new nuclear plants—there’s promising research underway in both the public and private sectors. Techniques such as Linear-Complexity Multiplication and Matrix Multiplication are being explored for their potential to significantly reduce generative AI’s power demands. In the interim, leading vendors are developing more energy-efficient servers and systems designed for AI tasks.
These developments suggest a future where generative AI and power efficiency are not mutually exclusive. As businesses continue to prioritise power savings and efficiencies, they demonstrate that it is indeed possible to embrace AI advancements without sacrificing sustainability ambitions. This emerging balance hints at an evolving landscape where technological innovation and environmental responsibility coalesce.
Source: Noah Wire Services
