The handprint and the footprint of Artificial Intelligence
By Andrea Osika, Senior Product Marketing Manager, Sustainability
The rise of AI marks a significant milestone in innovation that’s reshaping industries globally. Its impact is profound, bringing exceptional opportunities and considerable challenges, including the increased demand for energy and natural resources.
In coming years, finding new ways to harness the incredible potential of AI while minimizing environmental costs will greatly benefit organizations, communities and ultimately the planet.
The handprint
AI accelerates transformation, tackles complex problems. and enables positive impact. This is known as the handprint. It includes automation, data analytics, software development and cybersecurity that can cut a company’s costs and drive revenue.
In recent years, generative AI, also known as GenAI, has advanced rapidly and is now part of a modern organization's strategy to remain competitive. Tools like ChatGPT, GitHub Copilot and DALL-E have revolutionized how businesses operate by enhancing productivity, automating tasks and fostering creativity through advanced AI capabilities. Gartner predicts that by 2025, GenAI will be a workforce partner for 90% of companies worldwide.¹ Other analysts estimate that GenAI’s impact on productivity could add trillions of dollars to the global economy.
While AI comes with significant opportunities, it also comes with environmental consequences, including increased demand for energy and resources that generate a sizable carbon footprint.
The footprint
Datacenters and networks are the backbone for AI and contribute an estimated 2% of the planet's energy consumption. This is more than some countries generate in a year and demand is expected to increase.
In fact, a new report illustrates how datacenters could double their electrical consumption by 2026 due to increased demands in compute power, cooling AI workloads and related services. Another report corroborates this and predicts that increased AI usage will double emissions in the same time span. This trend has implications that work against the Paris Agreement and other global efforts to mitigate climate change and increased regulation which increasingly make sustainability an imperative for organizations worldwide.
The way forward
According to the 2024 Enterprise Cloud Index (ECI) report from Nutanix, 88% of respondents indicate that sustainability is a priority. The report also notes that implementing AI strategies is a top priority for CIOs. The way forward is a balancing act between incorporating the benefits of AI and keeping a watchful eye on sustainability goals.
While no one-size-fits-all strategy can solve this conundrum, identifying key focus areas and leveraging a hybrid multicloud infrastructure that enables efficiency, flexibility and choice can move you closer to achieving balance.
Here are some ways IT infrastructure modernization can play a crucial role in reducing the environmental impact of AI workloads:
- Shrink the footprint of AI workloads
Consolidating resources needed to support AI workloads through virtualization, hyperconverged infrastructure (HCI) and a hybrid multicloud approach that unifies on-premises, cloud and edge as a single unified entity, would be a significant first step on a path to reduced environmental impact.
Collectively, this approach requires less hardware to run workloads compared to traditional infrastructures that rely on separate devices for storage, compute and networking functions. This contributes to reducing costs, energy consumption and carbon footprint.
This is where Nutanix comes into play. On average, customers that shared their experiences using the Nutanix Cloud Infrastructure (NCI) solution resulted in a 70% decrease in physical footprint and a 50% reduction in energy consumption versus legacy systems.*
- Employ flexible scalability to eliminate waste from overprovisioning
Hybrid multicloud infrastructures dramatically simplify how and where you run applications and data across on-premises datacenters, edge locations and public/private clouds.
This enables unprecedented new levels of dynamic resource allocation so organizations can scale up or down based on demand. With exceptional flexibility, AI workloads can quickly access additional computing power from public clouds during peak times without over-provisioning on-premises resources.
- Deploy pretrained GenAI models and AI-ready infrastructure
GenAI and other large language models (LLMs) can use massive amounts of energy, especially while being developed or trained. For example, the electricity required to train GPT-3 was estimated to be equivalent to the amount of an average American household for 120 years.
There are already a significant number of LLMs and image models, with more being developed everyday. Using a pre-existing model that is off the shelf or an open source model can save big on energy while still enabling you to customize and tune that model with your data.
Deployment of AI on a traditional 3-tier infrastructure can make scaling AI resources difficult and costly. Even when not being used, pretrained GenAI models consume vast amounts of resources including infrastructure and accelerator memory.
Moving to HCI transforms the process of building out GenAI by enabling you to buy only what you need and then scaling out one server at a time. Compute, storage, networking and AI accelerators all work in unison in a smaller form factor with extraordinarily high efficiency.
The Nutanix GPT-in-a-Box turnkey solution brings all this to life. It unites your infrastructure with an open-source AI software stack – including kServe, PyTorch, Kubernetes®, and the Nutanix Unified Storage (NUS) platform for data management and security – to safeguard and scale your data for efficient inference and model tuning.
With Nutanix GPT-in-a-Box, GPU acceleration facilitates real-time inferencing and processing while storage hosts AI/ML models as well as additional context data, such as retrieval-augmented generation (RAG) models. Data stored on the Nutanix platform becomes part of a virtual datacenter where you can control and privatize anything generated by AI applications.
- Optimize workload placement to minimize energy consumption and emissions
In a recent report on the state of AI, 99% of respondents said they plan to upgrade their AI applications or infrastructure and more than half wanted seamless data mobility across core and edge environments.
Mobility matters in terms of performance, but where a workload is run can significantly influence the resulting amount of energy and emissions. By using a combination of on-premises and public cloud resources in hybrid fashion, organizations can optimize workload placement based on the energy efficiency of the available hardware or access to renewable energy. For instance, AI tasks can be allocated to different AI accelerators like GPUs or specialized CPUs that offer higher performance per watt.
The location of a workload can affect environmental impact too. Moving a workload to a geography that has access to renewable energy or with a lower carbon intensity can have a 20-fold impact on the emissions associated with it. While that might not be practical for every organization, migrating workloads to colocation facilities can be more efficient than private datacenters when it comes to saving on energy.
In some cases, extending workloads into the public cloud can help reduce emissions. Gartner said "public cloud providers can produce 70-90% fewer greenhouse gas (GHG) emissions than traditional server rooms, owned data centers, and midsize data center facilities.”² Why? Public clouds are incentivized to create ultra-efficient datacenters to optimize cost. Many of the hyperscalers have also committed to the use of renewable energy.
Technologies that help facilitate the portability of workloads between locations can provide practitioners with more flexibility to achieve their targeted outcomes. NCP enables identical cloud operations and enterprise data services across clouds so you can easily manage and move workloads across various clouds without refactoring applications.
Running on NCP, Nutanix Cloud Clusters (NC2) runs on bare-metal instances in hyperscalers like AWS and Azure. Nutanix has also committed to cloud-native designs that are under development to allow their software to run natively in public clouds based on customer needs.
Additionally, NC2 gives organizations the freedom to move applications anywhere and at any time between on-premises, edge and public/private cloud environments without ever having to refactor, replatform or re-architect.
- Accelerate the adoption of sustainable practices with visibility and control.
Having a distributed, siloed infrastructure for AI can be difficult to manage and control and things can get out of hand quickly. By implementing a single, unified platform with centralized visibility into consumption, IT teams can remain acutely aware of their cloud usage impact and improve the efficient use of resources.
Understanding patterns can help identify ways to optimize models for performance and cost while identifying inefficiencies and areas for improvement. Indispensable solutions like the Nutanix Cloud Manager (NCM) SaaS platform are crucial to managing and controlling your AI handprint and footprint so that costs, energy consumption and carbon emissions remain in check.
Harmony in the hybrid multicloud
Having a hybrid multicloud infrastructure empowers organizations with the freedom to choose their clouds, apps and technology stack without compromising on performance or cost. With this approach, AI workloads have the flexibility to run where it makes the most sense in terms of performance and security as well as efficiency and the flexibility to adapt to this rapidly evolving landscape.
* These space or energy savings claims are average results based on case studies of over 50 representative Nutanix customers as of 9/24/2024. Such case studies are publicly available on the Nutanix website. Because potential customer outcomes depend on a variety of factors including their use case, individual requirements, and operating environments, these accounts should not be construed to be a promise or obligation to deliver specific outcomes. We invite you to contact Nutanix here to discuss how we may be able to provide an optimal solution for your specific circumstances.
¹Gartner® Press Release, “Gartner Says AI Ambition and AI-Ready Scenarios Must Be a Top Priority for CIOs for Next 12-24 Months,”6 November 20234
²Gartner® Press Release, “Gartner Says CIOs Must Balance the Environmental Promises and Risks of AI,” 7 November 2023
GARTNER is a registered trademark and service mark of Gartner, Inc. and/or its affiliates in the U.S. and internationally and is used herein with permission. All rights reserved.
Learn more about the factors that can impact energy and emissions using Nutanix Validated Designs:
Tap into additional resources on sustainable IT solutions:
www.nutanix.com/solutions/sustainability-and-it
www.nutanix.dev/?s=sustainability+
©2024 Nutanix, Inc. All rights reserved. Nutanix, the Nutanix logo and all Nutanix product and service names mentioned herein are registered trademarks or trademarks of Nutanix, Inc. in the United States and other countries. All other brand names mentioned herein are for identification purposes only and may be the trademarks of their respective holder(s). Certain information contained in this publication may relate to, or be based on, studies, publications, surveys and other data obtained from third-party sources and our own internal estimates and research. While we believe these third-party studies, publications, surveys and other data are reliable as of the date of this paper, they have not independently verified unless specifically stated, and we make no representation as to the adequacy, fairness, accuracy, or completeness of any information obtained from third-party sources.