The tech industry has a growing power usage and environmental impact problem.
Between the rise of cloud and compute-intensive applications such as cryptomining and artificial intelligence, power demand has expanded at an alarming rate. The Bitcoin network uses approximately the same amount of energy that Washington State does yearly. Mark Papermaster, chief technology officer for Advanced Micro Devices Inc., has warned that computer energy consumption for machine learning systems is on pace to surpass the world’s energy production supply within the next two to three decades.
The rapid emergence of generative AI and the hype surrounding OpenAI LP’s ChatGPT are a particular source of alarm for a number of climate activists and tech industry leaders. According to a recent Bloomberg story, it’s estimated that the training of a single AI model can now consume more electricity than what is needed to power 100 U.S. homes for an entire year.
“AI has a dirty secret: It’s dirty,” Leslie Miley, technical adviser to the CTO at Microsoft Corp., said in keynote remarks during the QCon Conference in London this week. “Generative AI is amazingly intensive. It’s booming, but so is its carbon footprint. Data centers are being built everywhere.”
Need for different infrastructure
Miley (below), who noted in his remarks that the views expressed were his and not his employer’s, was familiar with data center growth, having previously served in leadership roles with major tech firms such as Google LLC, Apple Inc., Twitter Inc. and Slack Inc. He was also CTO for former President Barack Obama’s foundation and has been outspoken about the need for sustainability in the tech sector.
“We create these technologies that push more CO2 into the atmosphere and then we say we’re going to fix it later,” Miley said. “The problem is that generative AI is going to need a different type of infrastructure. This is a problem that’s not going to go away.”
Part of the problem is that existing infrastructure has been built to support a massive number of servers, whether or not those machines are serving a useful purpose. In a separate QCon presentation, Red Hat Inc.’s Holly Cummins described the proliferation of “cloud zombies,” cloud-connected servers that continue to run while providing no useful work.
According to a “State of Tech Spend” report by Flexera, overprovisioning and always-on resources led to more than $26 billion in public cloud waste in 2021 alone. And this figure is on top of an additional number of servers that are not being fully utilized at or near capacity.
“When you add the zombies and the underutilized servers together, you have a really big problem,” Cummins said. “It’s so much money, it’s so much carbon, it’s so wasteful.”
Keys and device pollution
In addition to AI and zombie cloud servers, other areas of the tech world are contributing to climate pollution as well. In a QCon keynote, Radia Perlman, a fellow at Dell EMC, reiterated her concerns about the technology behind cryptographic keys.
The keys, which are widely used to encrypt information using a data string, fail to meet an acceptable standard for accuracy and environmental sustainability, according to Perlman. “Humans are incapable of securely storing high-quality cryptographic keys, and they have unacceptable speed and accuracy when performing cryptographic operations,” Perlman said. “They are also large, expensive to maintain, difficult to manage, and they pollute the environment.”
Devices on which much of the world’s data is stored are also raising concern. The global solid state drive or SSD market is expected to grow from $9.68 billion in 2022 to $38.81 billion by 2030, according to one industry estimate. However, SSDs are coming under scrutiny for how they are made.
SSDs are composed of integrated circuits that use far less power than hard drives, which store information on a rotating magnetic disc. Yet they’re significantly more carbon-intensive to manufacture because chip fabrication facilities must operate at extreme temperatures that require intensive energy resources to maintain.
“SSDs have a huge carbon content to make them,” Adrian Cockcroft, a former executive with Amazon Web Services Inc. and Netflix Inc., said in a presentation. “The carbon footprint of an SSD is all in the manufacturing.”
The tech industry is turning to the innovation playbook to find new solutions for climate impact concerns. One of these can be seen in the Green Software Foundation, a nonprofit organization with a mission to make software an enabler of climate solutions.
“Many of us especially in the software industry have an outsized opportunity at work,” Sara Bergman, senior software engineer at Microsoft and an individual contributor to the foundation’s work, said during a session at QCon. “Software has a role to play here, and that role is up to us.”
To that end, one of the foundation’s core principles involves carbon-aware computing using software tools to help lower emissions. Its projects include a software development kit to create applications that do more when the power comes from clean, low-carbon sources and less when it does not.
In addition, the foundation provides a Green Software Patterns Catalog for helping to delete unused storage resources and matching utilization requirements with preconfigured servers.
Other solutions within the industry include Cloud Carbon Footprint, a free and open-source tool for monitoring and reducing carbon emissions during cloud use. There’s also a host of carbon calculator tools now available, including an analytics-based solution that can be integrated with SAP SE’s ERP system, and the Climatiq API, which can be deployed in data center environments.
Speakers throughout the three-day conference noted that the tech industry has a remarkable ability to pursue solutions combining efficiency with innovation, without pausing to consider the overall impact. The rising prominence of AI and expanding growth of the cloud have carried the technology world to a juncture where it may soon have to make a hard choice between an existing carbon-spewing model or one less damaging to the environment.
Efficiency, speed and agility have become the mantra for tech companies seeking to gain competitive advantage. Environmental realities may soon force businesses to make a different set of choices. As the business management philosopher Peter Drucker once put it succinctly, “There is nothing so useless as doing efficiently that which should not be done at all.”