Best of LinkedIn: Sustainability & Green ICT CW 34/ 35

00:00:00: Welcome to the Deep Dive.

00:00:01: Today, we're cutting through the noise on sustainability and green ICT.

00:00:05: We want to bring you the sharpest, maybe even the most surprising insights we've seen recently on LinkedIn.

00:00:10: That's right.

00:00:11: This Deep Dive is powered by Thomas Allgaier and Frennis.

00:00:14: And it's based on the most relevant posts about sustainability and green ICT over the last couple of weeks, calendar weeks, thirty-four and thirty-five specifically.

00:00:22: And just quickly, for those who might not know, Frenis is a B to B market research company.

00:00:27: They help enterprises optimize their campaigns with account and executive insights that really go far beyond what AI alone can deliver.

00:00:36: So our mission today is pretty straightforward.

00:00:39: distill the top themes, the key movements, and hopefully some genuinely actionable takeaways from this really vibrant online community.

00:00:46: Think of it as a shortcut maybe to understanding the cutting edge of sustainable tech right now.

00:00:50: Yeah, exactly.

00:00:51: We'll cover everything from AI's true environmental footprint, which is still evolving to some really practical green coding strategies.

00:00:59: Got a lot to explore.

00:01:00: OK, let's jump right in.

00:01:01: then, our first big theme.

00:01:03: It seems to be all about the industry's push for more transparency and crucially measurable efficiency in AI.

00:01:10: Right.

00:01:10: There's this real drive, it feels like, to understand AI's environmental impact much more clearly.

00:01:16: And some pretty eye-opening data just dropped recently.

00:01:19: Giving us a peek behind the curtain, so to speak.

00:01:21: Kind of, yeah.

00:01:22: What's really fascinating is Google's detailed report on its Gemini apps.

00:01:26: Anna Lerner-Nesbitt highlighted this one.

00:01:28: It gives some incredibly granular insights.

00:01:30: Okay, it's like wet.

00:01:30: Well, they found that a median prompt uses just point two four watt hours of energy.

00:01:35: Tiny, right?

00:01:36: Wow, point two four.

00:01:38: Yeah, that sounds minuscule.

00:01:39: And about point two six milliliters of water.

00:01:41: But here's the kicker.

00:01:42: And Alex Sinevoys pointed this out.

00:01:44: Google is claiming a huge reduction year over year, like a thirty three x energy reduction and a forty four x emissions reduction.

00:01:51: for per prompt usage.

00:01:53: In

00:01:53: just one year, that's impressive.

00:01:54: It is.

00:01:55: And the report even breaks down that tiny point two four watt hours.

00:01:58: Fifty eight percent is the AI chip, twenty five percent the host machine CPU and memory, ten percent backup gear, and eight percent is just data center overhead.

00:02:06: OK, those individual numbers, they sound incredibly small, almost negligible, like you said.

00:02:11: Alexinovals even put it in perspective.

00:02:13: Less energy than watching nine seconds of TV, apparently.

00:02:15: Right.

00:02:16: But.

00:02:18: And this is the big but, isn't it?

00:02:20: When you connect this to the bigger picture, you know, the sheer volume of AI queries happening every single day, does that tiny number really hold up?

00:02:29: The scale is just mind-boggling.

00:02:31: That's exactly the point many people are making.

00:02:34: Elaine Parr reflected on this.

00:02:35: Sure, a single query is tiny, but billions of prompts a day.

00:02:39: That quickly adds up.

00:02:40: So what kind of scale are we talking?

00:02:42: Well, potentially the electricity needs of tens of thousands of homes.

00:02:46: and hundreds of thousands of gallons of water every single day.

00:02:50: Wow.

00:02:51: Yeah.

00:02:51: Dr.

00:02:51: Satyalushyani and Naveen Balani both really underscored this cumulative impact.

00:02:55: That point two four watt hours.

00:02:58: multiplied by Google's daily query volume, it becomes industrial and magnitude.

00:03:01: So not negligible at all when you look at the whole system?

00:03:04: Not at all.

00:03:05: Naveen Balani put some numbers on it.

00:03:07: If AI prompts replace traditional search at a rate of say one billion prompts a day, that's two hundred forty megawatt hours of energy, thirty tons of CO two and two hundred sixty thousand liters of water daily.

00:03:22: That's huge.

00:03:22: A really powerful counterpoint to the per prompt efficiency story.

00:03:26: It

00:03:26: is.

00:03:27: But then there's another layer.

00:03:28: Peter Slattery shares some interesting research offering a slightly different angle on AI's long term path.

00:03:34: Oh.

00:03:34: What was that?

00:03:35: He mentioned this idea of a green AI kuznets curve.

00:03:38: The theory suggests that, yes, AI initially increases energy consumption.

00:03:42: That's the hump.

00:03:42: Okay.

00:03:43: But over time, as the tech matures and gets more efficient and hopefully integrated with more renewables, it could actually lead to emission reduction.

00:03:50: Huh.

00:03:51: So an initial surge then potentially a decline.

00:03:53: Potentially,

00:03:54: yeah.

00:03:54: Especially once a certain AI market penetration per capita is reached.

00:03:58: He mentioned we might already be seeing this in places like Singapore and the US.

00:04:01: So maybe there's a peak environmental impact and then things could improve.

00:04:05: That's a more optimistic long-term view, I guess.

00:04:09: But it highlights the complexity, doesn't it?

00:04:11: Tiny per prompt use.

00:04:13: Massive cumulative impact.

00:04:15: Maybe a future greening.

00:04:16: Exactly.

00:04:18: which leads to the next logical point.

00:04:20: How do we even measure this stuff consistently?

00:04:22: Right.

00:04:23: What does this all mean for the broader industry?

00:04:26: It really sounds like we desperately need some common ground for measurement, especially if we want to compare different AI providers

00:04:32: fairly.

00:04:33: That's becoming a critical issue.

00:04:35: Liam Distant pointed out there really isn't a single view or consensus on AI's impact.

00:04:40: And a big reason is the lack of common frameworks.

00:04:43: So everyone's measuring differently.

00:04:44: Pretty much.

00:04:45: Coordinator Korn actually questioned if comparing vendors right now is like comparing apples with oranges without these standards.

00:04:51: Makes sense.

00:04:52: And Boris Kamizachikov echoed that.

00:04:54: calling for standardized frameworks.

00:04:56: He mentioned an emerging AI energy score concept, trying to make data genuinely comparable.

00:05:01: Are big players pushing for this too?

00:05:03: Seems like it.

00:05:04: Salesforce's AI sustainability outlook, which Sunya Norman shared, really advocates for trusted, reliable, and sustainable AI practices as basically a prerequisite for adoption.

00:05:16: They're focusing on smart demand, efficiency, and clean supply.

00:05:20: So yeah, the push for transparency and standards is definitely growing.

00:05:24: Okay, so beyond the sort of computational side of AI queries, let's talk about the physical stuff, the infrastructure, data centers.

00:05:31: Their footprint is massive, right,

00:05:33: and growing.

00:05:33: Oh, absolutely, booming.

00:05:35: Karen van der Zanden shared a pretty stark statistic.

00:05:37: In the

00:05:38: U.S.,

00:05:38: more money is now spent building data centers than office buildings.

00:05:41: Wow.

00:05:42: That really signals the AI era, as she put

00:05:44: it.

00:05:45: It really does, and the consumption is huge.

00:05:47: Shakun Shan noted that a single AI data center can use the same electricity as a hundred thousand homes.

00:05:52: and the ones being planned, even bigger.

00:05:55: This growth is genuinely starting to outstrip our current energy supply in some areas.

00:05:59: And it's not just energy, right?

00:06:00: Water is another big factor, and one that seems to cause some confusion.

00:06:04: I saw Simhussein clarifying this.

00:06:06: Yes, he did a great job explaining the water dilemma, as he called it.

00:06:10: There's a key difference between site-based water measurement that's the water used directly for cooling at the data center.

00:06:17: like the point two six l m l per gemini.

00:06:20: prompt google reported okay.

00:06:22: and then there's embodied water measurement.

00:06:25: this includes the water consumed by the power plants that supply the electricity to the data center and that figure can be way higher maybe four times higher.

00:06:33: ah so that explains the different numbers we hear are both valid.

00:06:37: He argues, yes, they are.

00:06:38: They just serve different purposes.

00:06:40: The site-based number is more for operational metrics, for engineers improving cooling efficiency.

00:06:46: The embodied number is more for policy metrics, for regional planning, understanding the total water stress.

00:06:52: Got it.

00:06:53: So different lenses for different decisions, that's helpful.

00:06:56: Definitely.

00:06:56: It's not about right or wrong, but understanding the context.

00:06:59: And besides energy and water, are there other environmental impacts we should be thinking about, especially as these centers get built close to where people live.

00:07:07: Yeah, that's a growing concern.

00:07:09: Mexica all good.

00:07:10: specifically highlighted noise pollution.

00:07:12: Noise from data centers.

00:07:15: Yep, especially from the cooling systems, which are often on the roof.

00:07:17: Yeah.

00:07:18: She pointed out they can be incredibly loud for nearby residential areas.

00:07:22: It's an often overlooked externality, but it definitely impacts quality of life.

00:07:26: Hmm, something to consider.

00:07:28: But it's not all doom and gloom on the infrastructure front, is it?

00:07:32: We're seeing some positive developments too, like efforts towards greener data centers.

00:07:36: Absolutely.

00:07:37: There are some really exciting initiatives.

00:07:39: Andy Davis shared news about Vesper infrastructure partners investing in a company called TerraCraft in Norway.

00:07:46: Norway.

00:07:47: Lots of hydropower there, right?

00:07:48: Exactly.

00:07:49: They're building a green AI data center powered by one hundred percent renewable hydroelectric energy.

00:07:55: Plus, they're using lake water and liquid cooling

00:07:58: for

00:07:59: ultra-efficient performance.

00:08:01: That sounds promising.

00:08:02: Any other examples?

00:08:03: Well, Karen van der Zanden also mentioned how Europe... particularly places like the Amsterdam region, is focusing on integrating data centers better with local environments, like using their waste heat to warm nearby homes.

00:08:15: Oh,

00:08:15: that's clever, circular economy thinking.

00:08:17: Right.

00:08:18: And Rick Schoenmecker detailed how the ports at Barcelona modernized its whole IT setup with IBM Linux Oni.

00:08:24: They consolidated workloads, significantly cut energy use, and boosted security and scalability all at once.

00:08:32: So real world progress is happening.

00:08:34: Okay, good to hear.

00:08:34: Let's pivot now.

00:08:36: Moving from the big infrastructure down to what professionals, developers, ops teams can actually do right now.

00:08:43: It feels like green software and DevOps are becoming really critical.

00:08:46: Yeah, definitely.

00:08:47: Donna Jaiber of Frame DevOps is a true driver of sustainable IT.

00:08:51: He even coined the term green ops revolution.

00:08:54: Green ops revolution,

00:08:54: I like that.

00:08:55: Me too.

00:08:56: It's all about consciously reducing waste, optimizing resources and building eco-friendly systems.

00:09:01: Trying to tackle that two, four percent share of global CO-II emissions.

00:09:05: industry is responsible for.

00:09:07: Okay, so for people listening who are in those roles, what are some concrete ways green ops translates into code or daily operations?

00:09:13: What actions can teams take?

00:09:15: Don and Jaya Burra laid out five really practical strategies.

00:09:18: First, cloud cost optimization.

00:09:20: Which also saves money, presumably.

00:09:22: Exactly.

00:09:23: Using tools like AWS Compute Optimizer or Kubernetes Vertical Pod Auto Scalar.

00:09:28: And simple things like scheduling lawn production environments to automatically switch off when no one's using them.

00:09:34: Smart.

00:09:34: What's next?

00:09:35: Second, green coding practices.

00:09:37: This means actively adopting energy-aware development techniques.

00:09:41: Maybe even using AI-powered linter's code.

00:09:43: carbon was mentioned.

00:09:44: They can flag inefficient energy-hungry code before it even gets deployed.

00:09:50: Like a spellchecker, but for energy use in code.

00:09:52: Sort of, yeah.

00:09:53: Third, serverless and edge computing.

00:09:56: The idea is to eliminate idle servers just sitting there consuming power.

00:09:59: Makes sense.

00:10:01: Fourth, observability for sustainability.

00:10:03: basically monitoring your carbon footprint as a key metric, just like you monitor performance or cost.

00:10:09: Using tools like Microsoft Cloud for sustainability or Scafandre.

00:10:13: Making it visible.

00:10:13: Right.

00:10:14: And finally, fifth, culture and advocacy.

00:10:16: This is about invading sustainability thinking into the team's processes, maybe adding sustainability to the definition of done for tasks, training teams on these greenhouse principles.

00:10:27: So it becomes part of the team's DNA.

00:10:29: Exactly.

00:10:30: And Sahel Bhatia made a great point too, emphasizing that sustainable coding is often just good coding.

00:10:35: It encourages optimization, efficiency, thoughtful design.

00:10:39: It's not necessarily extra work, just better work.

00:10:41: That's a powerful framing.

00:10:43: And these green ops practices, they sound good for the planet, obviously, but.

00:10:47: Are there tangible business benefits too?

00:10:49: Are companies seeing a real ROI?

00:10:52: Oh, absolutely.

00:10:53: The business case is getting stronger all the time.

00:10:56: Will Nordberg highlighted one case study where just refactoring inefficient code cut energy use by twenty nine percent per user per month.

00:11:03: Twenty

00:11:04: nine percent.

00:11:04: That's significant.

00:11:05: Huge.

00:11:06: And he noted practical coding choices can sometimes cut energy costs by up to thirty percent.

00:11:11: Green cloud architectures, serverless, they directly translate to lower cloud bills.

00:11:15: It's becoming undeniable.

00:11:16: So it's hitting the bottom line.

00:11:18: Definitely.

00:11:19: And Kid Ardeo, whose insights were shared by Sujit Patanj, stressed the sustainable coding isn't really optional anymore.

00:11:25: He talked about its pillars and how a genetic AI-AI designed to act autonomously could actually multiply these green efforts.

00:11:32: Interesting.

00:11:32: Okay, moving from strategies to specific tools.

00:11:36: There is news about a new browser extension.

00:11:39: something to give everyday users more visibility.

00:11:41: Yeah, this sounds really neat.

00:11:43: Emanuel Bartolisi just launched a browser extension called the Green Software Indicator.

00:11:47: What does it do?

00:11:48: It basically evaluates the website you're visiting against green software development practices.

00:11:54: Then it gives you a simple color-coded badge, green for excellent, yellow for good, red for needs improvement.

00:12:00: Ah,

00:12:00: like an energy rating for websites.

00:12:02: Kind of.

00:12:03: And it gives you a breakdown too, looking at things like image optimization, whether JavaScript and CSS are minified, even energy efficient design choices, like having a dark mode option.

00:12:13: Making

00:12:13: sustainability visible right in the browser.

00:12:15: Cool.

00:12:16: Yeah, helps raise awareness and maybe nudge developers.

00:12:19: And we're also seeing these green AI principles apply in some maybe unexpected areas beyond just optimizing software.

00:12:26: Definitely.

00:12:27: Daniel Chen shared a fantastic example.

00:12:29: NVIDIA and a company called Carbon Robotics are using computers computer vision AI to spot and then laser eliminate weeds in farm fields in real time.

00:12:39: Wow, laser zapping weeds.

00:12:41: Yeah.

00:12:41: He called it truly green AI in action.

00:12:44: It reduces the need for toxic herbicides and cuts down on wasted resources in agriculture.

00:12:50: A really tangible environmental benefit from

00:12:52: AI.

00:12:52: That's a great example of AI as a solution, not just a problem.

00:12:56: Okay, so we talked software operations, data centers.

00:13:00: What about the hardware itself?

00:13:02: The physical chips and cards?

00:13:04: They have a footprint too, right?

00:13:05: A very significant one, and often kind of hidden.

00:13:08: Understanding the full lifecycle is crucial, and there's some groundbreaking research emerging here.

00:13:12: Like what?

00:13:13: David Eckshauser presented results from a super comprehensive lifecycle analysis, or LCA, of an NVIDIA A- one hundred GPU.

00:13:21: The

00:13:21: big AI chip.

00:13:22: Exactly.

00:13:23: This might be the most in-depth study of its kind for this type of hardware.

00:13:26: They went to extreme lengths, apparently even pulverizing the graphics card to analyze all its chemical elements.

00:13:31: Whoa, serious stuff.

00:13:34: Yeah, real rigor.

00:13:35: This kind of research really accelerates the conversation about hardware level impacts.

00:13:40: It highlights why we need more transparency right down to the component level so organizations can make smarter procurement choices and manufacturers can design better.

00:13:48: Absolutely.

00:13:49: Okay, bringing this all together, this whole deep dive really underscores that sustainability in tech isn't something one company or one team can solve alone.

00:13:59: It's a massive collaborative effort.

00:14:01: Totally.

00:14:01: So

00:14:02: what

00:14:02: signals are we seeing from the wider global community?

00:14:06: And what about governance, future direction?

00:14:08: Well, things like the JISC Digital Sustainability Newsletter, which Kalinas curates, are great for keeping track of global policy shifts, new practices, and tech developments.

00:14:18: Keeping people informed.

00:14:19: Right.

00:14:20: And there are calls for bigger action.

00:14:21: So Mia Balaraman made a strong case for a global AI and environment accord.

00:14:25: She sees green AI as a potential long-term competitive advantage for nations and companies that get it right.

00:14:31: A global agreement.

00:14:33: And Liam Diskind again emphasized that urgent need for global reporting standards.

00:14:39: He mentioned groups like the Green Software Foundation are making real headway there, developing practical frameworks.

00:14:45: So the momentum for standards is building globally too.

00:14:48: And how are individual organizations tracking their own progress?

00:14:52: And maybe just as importantly, how are they fostering this culture of sustainable tech internally?

00:14:58: We're seeing new metrics emerge.

00:15:00: Sanjay Patur introduced Accenture's SAIQ metric that stands for Sustainability AI Quotient.

00:15:05: It tries to quantify AI's return on sustainability.

00:15:09: How does it do that?

00:15:10: It tracks how efficiently AI converts inputs, money, energy, emissions into actual performance outputs.

00:15:17: So, measuring the green bang for your buck, essentially.

00:15:19: Okay,

00:15:19: measuring the efficiency of the sustainability effort itself.

00:15:22: Exactly.

00:15:22: And beyond metrics, community building is huge.

00:15:25: Wilco Bergraf shared about an ASML meetup focused specifically on embedding sustainability into software engineering, bringing different disciplines together.

00:15:33: Connecting practitioners.

00:15:34: Yep.

00:15:35: And Green I.O.

00:15:36: London, highlighted by Sandra Seedow, Gael, the UEZ, and Ann Curry, they announced a program focused on responsible technology, offering hands-on use cases and sector-specific green ops insights.

00:15:47: Practical

00:15:47: learning.

00:15:48: Definitely.

00:15:49: Training is popping up, too.

00:15:50: James Martin is leading a frugal AI course, promoting responsible AI practices.

00:15:55: And Amber G. and Kavanaugh D. Rajan highlighted the importance of accessible e-learning, mentioning courses from Position Green, Microsoft, LinkedIn, making this knowledge available to a much broader audience.

00:16:07: So it's a mix of metrics, community, training, education, a multi-pronged approach.

00:16:12: It has to be really.

00:16:13: So as we wrap up, we've journeyed through this really fast-moving landscape of sustainability and green ICT from those hyper-efficient AI queries right up to the massive data centers powering them.

00:16:24: What really stands out, I think, is this deal nature of AI.

00:16:27: It's clearly a challenge environmentally, but also potentially a really powerful solution.

00:16:31: Right, it cuts both ways.

00:16:32: Exactly.

00:16:33: Considering that, maybe we can leave you, our listener, with this thought.

00:16:37: If individual AI queries are becoming almost negligible in energy use, as that Google report suggests.

00:16:43: And if AI itself holds promise for optimizing systems and accelerating decarbonization, should our main focus maybe shift?

00:16:52: Shift from what to what?

00:16:53: Well, maybe shift away from just focusing purely on individual query consumption and start asking a deeper question.

00:16:59: Like, how do we ensure that the exponential growth of AI and all the data centers needed to support it isn't just efficient but is intentionally aligned with global sustainability goals.

00:17:11: Moving beyond less bad to net positive.

00:17:14: Precisely.

00:17:15: How do we make AI a true net positive force?

00:17:18: What does genuine earth alignment for AI actually look like and how do we start building systems with that principle baked in right from the core?

00:17:25: Especially when it means maybe prioritizing societal and environmental benefit over

00:17:29: raw

00:17:30: computational power or speed.

00:17:32: That's a big question, a really important one for the future of the field.

00:17:35: Something to mull over.

00:17:36: If you found this deep dive valuable, remember new ones drop every two weeks.

00:17:41: We also encourage you to check out our other editions covering cloud, digital products and services, artificial intelligence, and ICT and tech insights.

00:17:49: Thank you for joining us on this exploration.

00:17:51: Yeah, thanks for listening.

00:17:53: And we hope you'll subscribe so you don't miss our next deep dive into the fascinating world of tech and its impact.