Best of LinkedIn: Sustainability & Green ICT CW 40/ 41

00:00:00: This episode is provided by Thomas Allgeier and Freeness, based on the most relevant LinkedIn posts about sustainability and green ICT in CW-Forty and Forty-One.

00:00:09: Freeness supports ICT enterprises with market and competitive intelligence, decoding green software developments, benchmarking emerging standards, tracking regulatory shifts, and analyzing competitor strategies.

00:00:22: Welcome back to the deep dive.

00:00:24: Our mission for you today, well... It's pretty clear.

00:00:27: We're distilling the absolute top operational shifts and trends in sustainability and green ICT stuff that really dominated LinkedIn over the last couple of weeks.

00:00:37: This deep dive is all about getting you past the buzzwords straight into the actionable insights, you know, drawn directly from our sources.

00:00:43: And what's really striking, I think, is seeing how fast the conversation has shifted.

00:00:48: It's not just should we do this anymore.

00:00:49: Right.

00:00:50: It's, okay, how do we get... measurable cost effective efficiency.

00:00:54: The sources show this like laser focus on optimization that actually hits the bottom line.

00:00:58: It covers everything from, you know, how we write AI prompts all the way up to global data center strategies.

00:01:03: That's a great point.

00:01:04: And to make it all digestible, we've kind of broken down this flood of information into four main themes, interconnected ones.

00:01:11: We'll kick off with the human and code layer thing.

00:01:14: AI efficiency and governance.

00:01:16: Then we'll dive into the physical stuff data center infrastructure.

00:01:19: The hardware.

00:01:20: Exactly.

00:01:21: Third, we'll look at some really practical green software levers, the granular stuff.

00:01:26: And finally, we zoom out for the strategic big picture, alignment and sovereignty.

00:01:32: Makes sense.

00:01:33: So let's start with AI then.

00:01:34: It's probably the hottest topic.

00:01:37: everywhere right now.

00:01:38: There's this really rising urgency to get a handle on the massive energy footprint of these advanced models.

00:01:44: It's

00:01:44: a huge footprint.

00:01:45: Yeah, and you see the hyperscalers now actively positioning greener AI data centers, not just as like a compliance thing, but as key levers for their own decarbonization efforts.

00:01:54: Right.

00:01:55: The challenge is pretty straightforward, really.

00:01:57: How do we stop this exponential AI innovation from automatically jacking up our collective carbon footprint?

00:02:02: Okay, so let's unpack that at the most fundamental level, like the code itself.

00:02:07: Wilco Bergraf flagged a really interesting study on something called green prompt engineering.

00:02:13: Ah, yes.

00:02:14: It showed that the way you actually design your prompt, specifically how readable it is, how long it is, that directly impacts energy use and the quality of the output.

00:02:25: It's kind of counterintuitive, isn't it?

00:02:26: Usually we obsess over the model's complexity, not the text we feed it.

00:02:29: Yeah, exactly.

00:02:31: If you think about the actual computation needed to process a really dense, badly written prompt versus something concise and clean, well, the difference in energy use starts to make sense.

00:02:41: Totally.

00:02:42: The study basically concluded that, yeah, choosing an efficient model is number one, but optimizing the prompts.

00:02:48: That's the next crucial step.

00:02:50: Right.

00:02:50: The takeaway seems to be tune for readability.

00:02:54: Actively avoid that sort of.

00:02:56: ornamental complexity.

00:02:57: If you overcomplicate the prompt with jargon or just too many words, the energy use goes up, but you don't get a better answer.

00:03:03: You're literally burning watts for nothing useful.

00:03:06: Burning watts for complexity.

00:03:07: that doesn't pay back.

00:03:09: And that leads nicely into the financial side.

00:03:12: Shankarama Swami warns, quite rightly, that unchecked AI innovation will increase the global carbon footprint.

00:03:18: No doubt.

00:03:19: but the market is responding.

00:03:21: We saw this great example.

00:03:23: A cloud provider managed to cut their emissions by twenty one percent and save two point four million dollars annually.

00:03:31: And that was just by focusing on model efficiency and using energy-aware training schedules.

00:03:36: That's the kind of ROI that really grabs attention.

00:03:39: Yeah,

00:03:39: absolutely.

00:03:40: That's real money.

00:03:41: And here's where things get even more connected.

00:03:43: The ethical and environmental sides, they have to merge.

00:03:46: Dr.

00:03:47: Sasha Luchione and Giada Pastilli really stressed this point.

00:03:50: Yeah, they did.

00:03:50: Ethics and sustainability, they argue, must evolve together for AI to be truly responsible.

00:03:56: Yeah.

00:03:56: You can't really claim an AI is ethical if it's wrecking the planet in the process.

00:04:01: No, exactly.

00:04:01: Their argument is we have to move beyond just looking at accuracy metrics.

00:04:05: We need to measure the full environmental and social costs.

00:04:08: And that's why tools like the AI energy score are becoming so important.

00:04:12: It helps developers actually quantify the environmental impact of running their models, makes those hidden costs carbon resources visible and measurable.

00:04:21: It forces accountability.

00:04:22: And accountability is definitely needed.

00:04:24: Montgomery Singman raised a flag about a potential AI bubble brewing.

00:04:29: Oh, interesting.

00:04:30: Yeah, he notes this kind of disconnect between the really bold high level promises of AI and, well, it's practical, sustainable adoption in actual vertical markets.

00:04:41: The reality check.

00:04:42: Pretty much.

00:04:43: He suggests longevity won't come from hype.

00:04:45: It'll come from companies demonstrating real world value and sustainable integration, focusing on applications that actually pass both the environmental and the economic

00:04:54: test.

00:04:55: Makes sense.

00:04:55: Focus on the substance.

00:04:56: And speaking of practical applications, we are seeing tools emerge to help with this.

00:05:01: Alexis Normand announced Greenleys EcoPilot.

00:05:03: Ah,

00:05:04: the AI Copilot.

00:05:05: That's the one designed specifically to automate carbon accounting and generate actual reduction plans.

00:05:11: It tries to take those measurement complexities that can really bog companies down.

00:05:15: Yeah, they can be a nightmare.

00:05:16: And push them directly towards taking action.

00:05:19: Okay, so let's shift gears now.

00:05:21: from the let's say invisible complexity of code to the very visible demands of the physical stack.

00:05:28: data centers

00:05:29: for structure.

00:05:29: Right.

00:05:30: If AI needs more power, where is it coming from and how do we deliver it efficiently?

00:05:34: Exactly.

00:05:34: And this is critical because, I mean, the numbers are stark.

00:05:38: Global electricity demand from data centers is projected to double by twenty thirty.

00:05:42: Double.

00:05:43: That's huge.

00:05:44: It

00:05:44: is.

00:05:45: So we need fundamental shifts in power distribution.

00:05:48: Morten, we already mentioned the collaboration between ABB and NVIDIA.

00:05:51: They're working to accelerate high efficiency power tech, specifically focusing on next-gen, eight hundred volt DC data centers.

00:05:58: Okay,

00:05:58: wait, eight hundred VDC.

00:06:00: Why is that specific voltage such a big deal right

00:06:02: now?

00:06:03: Good question.

00:06:04: So traditionally, data centers run on AC power, right?

00:06:07: But that involves multiple conversion steps, AC grid to DC for servers, then often stepped down again.

00:06:13: Each conversion loses energy.

00:06:14: Okay, inefficiencies.

00:06:16: Exactly.

00:06:16: Direct current or DC, especially at a higher voltage like eight hundred V, allows for fewer conversions.

00:06:23: That dramatically cuts power loss and actually increases system resilience, too.

00:06:28: It's a foundational change needed if we're going to handle that doubling demand efficiently.

00:06:32: Got it.

00:06:33: Fewer steps, less waste.

00:06:34: Makes sense.

00:06:35: But doubling power also means more heat, right?

00:06:39: So intensifying scrutiny on cooling and water use.

00:06:42: Absolutely.

00:06:43: Sanisa Nicolik from Lenovo.

00:06:45: highlighted their Neptune liquid cooling tech.

00:06:48: Apparently it can cut the power used just for cooling by up to forty percent.

00:06:52: Forty

00:06:52: percent.

00:06:53: That's significant.

00:06:54: And it lets you pack servers more densely too.

00:06:56: We're definitely seeing liquid cooling become a key differentiator in new builds.

00:07:00: Joe Bell noted net mountains VL one data center in Germany.

00:07:03: They're using direct liquid cooling DLC.

00:07:05: Yep, DLC, which is a big step because it targets the heat right at the source, the CPUs and GPUs, which is way more efficient than just blasting cold air through the whole room.

00:07:15: Right, much more targeted.

00:07:16: However, we probably need a bit of caution with all these green labels.

00:07:19: Nigel Green made a really valid point.

00:07:21: Given how much power they inherently consume, no data center is truly green in an absolute sense.

00:07:27: The realistic goal has to be incremental, measurable improvements, smarter cooling, optimized compute, all aimed at cutting operational costs.

00:07:36: Yeah, it's about greener, not perfectly green.

00:07:39: And that focus on operational costs leads straight to resource management.

00:07:44: especially water.

00:07:45: Douglas Mootin reported that Microsoft's Mount Pleasant Campus, just as one example, could use up to eight point four million gallons of water a year.

00:07:54: Eight million

00:07:54: gallons.

00:07:55: That puts huge pressure on local resources, particularly if you're building in water stressed areas.

00:08:00: Huge pressure, which really underscores the strategic importance of where you build.

00:08:04: Lewis Vaughn detailed how it north is specifically leveraging the cooler climates and abundant renewable energy in the Nordics for their expansion.

00:08:12: Smart move.

00:08:13: Yeah, and it mirrors what Solano is doing with their Hunderbilla W campus in Finland.

00:08:17: Location becomes the first layer of sustainability really.

00:08:20: It cuts cooling needs and reduces reliance on fossil fuels right off the bat.

00:08:23: And that sustainability thinking needs to extend to the hardware life cycle too, right?

00:08:28: Absolutely.

00:08:29: Maddie Backer and shared some fascinating stuff on Rack Renew.

00:08:32: They remanufacture OCP IT racks.

00:08:36: OCP, that's the open compute project.

00:08:38: Standardized racks.

00:08:39: Exactly.

00:08:40: Standardized open source designs.

00:08:42: Bakren highlighted the sheer volume of these racks being scrapped prematurely.

00:08:46: He basically said remanufacturing them is a massive opportunity.

00:08:50: It's cost effective and sustainable, really pushing towards that circular economy idea for data center hardware.

00:08:56: That's great.

00:08:56: to see a concrete example of circularity in action.

00:08:59: Okay, so moving up the stack again, let's talk about practical efficiency gains right at the application and developer level.

00:09:05: green software.

00:09:06: Yeah, the small, powerful levers that developers actually control day to day.

00:09:10: Precisely.

00:09:11: And sometimes the biggest wins come from, well, seemingly tiny code changes.

00:09:15: Yeah.

00:09:15: R.W.L.O.N.

00:09:16: offered a specific SAP green core tip.

00:09:18: Oh, yeah.

00:09:19: What was it?

00:09:19: It's about SQL select statements, specifically in column-oriented databases like SAP ANA.

00:09:25: The tip.

00:09:26: Stop using select add.

00:09:27: Huh.

00:09:27: The wild card.

00:09:28: Why is that such a resource hog in ANA?

00:09:30: Well,

00:09:30: because HA is column-oriented, Using forces the system to do a lot of unnecessary CPU and IO work.

00:09:37: It has to read and process every single column header in definition, even if you only actually need data from two columns.

00:09:43: Ah, I see.

00:09:44: So it reads everything, even if unused.

00:09:46: Exactly.

00:09:47: By just specifying the exact column names you need, you drastically cut down that churn.

00:09:52: Less CPU, lower emissions.

00:09:55: It's tangible efficiency baked right into the query.

00:09:58: Simple, but effective.

00:10:00: I love those kinds of deep in the weeds, practical tips.

00:10:03: On a much broader scale though, we are finally seeing standards start to solidify for the web.

00:10:08: Tim Frick highlighted the new W-three-C web sustainability guidelines draft.

00:10:12: Yeah, that's a big deal.

00:10:13: It's got what, ninety-two guidelines?

00:10:14: Something

00:10:14: like that.

00:10:15: Aimed at standardizing low-impact design choices for web products.

00:10:19: Covers everything from network requests to image handling.

00:10:22: These standards are vital because they give developers a clear roadmap, like a baseline for sustainable design.

00:10:27: And standards help drive governance, which brings us to green ops.

00:10:30: Kevin Leslie noted its growing importance, putting it right alongside fine ops.

00:10:34: Makes sense.

00:10:35: Cost and carbon are linked.

00:10:37: Exactly.

00:10:38: The hidden infrastructure costs of inefficient software are just getting too big to ignore anymore.

00:10:44: And enabling that green ops thinking is carbon aware computing, which really cool.

00:10:48: here is the rise of practical tools.

00:10:50: Aiden Mir Mohamedi discussed the carbon aware SDK.

00:10:53: The

00:10:53: SDK.

00:10:54: What does that let developers do?

00:10:56: It lets them basically time shift compute heavy jobs.

00:11:00: Time shifting.

00:11:01: You mean running them when the grid energy is cleaner?

00:11:04: Precisely.

00:11:04: The SDK can check the real-time carbon intensity of the local electricity grid.

00:11:10: So developers can programmatically delay non-urgent tasks until, say, there's more wind or solar power available.

00:11:17: Oh, wow.

00:11:18: Yeah, so you could potentially shift a job from running when the grid intensity is say, seven hundred grams of CO two per kilowatt hour down to a time when it's only fifty grams.

00:11:27: That's a massive difference just from timing.

00:11:29: Huge reduction, purely driven by intelligence scheduling based on real grid data.

00:11:33: That is powerful.

00:11:35: And ultimately, as Crystal Love and Kaia pointed out, green software isn't just a technical nice to have anymore.

00:11:40: It's becoming a key strategic advantage.

00:11:43: It compounds efficiency, builds resilience, and it's now standing right alongside speed and security as a core requirement in RFPs and public contracts.

00:11:52: If your software isn't green, you might not even get considered.

00:11:55: Yeah, the bar is definitely rising.

00:11:58: And that strategic relevance, it brings us neatly to our final theme, the bigger picture.

00:12:03: geopolitical strategy, technological sovereignty.

00:12:07: Sustainability is now being explicitly used as a competitive edge by, well, entire economic blocks.

00:12:14: Right.

00:12:14: T. Ganbold detailed the EU's strategic vision.

00:12:17: They're aiming to build green, cyber-secure AI infrastructure.

00:12:21: And this isn't just about, you know, being good global citizen.

00:12:24: Oh, it's

00:12:24: hard strategy.

00:12:25: It absolutely is.

00:12:26: It's aimed squarely at achieving technological sovereignty.

00:12:28: And

00:12:29: how are they planning to pull that off?

00:12:30: Yeah.

00:12:30: By leveraging what they already have, right, Europe's leadership in renewable energy, Germany, for instance, already hitting a sixty-two percent renewable mix in their grid.

00:12:38: Wow, sixty-two percent.

00:12:39: Yeah.

00:12:40: So the EU is essentially turning sustainability requirements, like demanding low-carbon IT, into a competitive advantage for their own digital industry.

00:12:49: They're creating a market for highly efficient tech that others then have to scramble to meet.

00:12:54: That's

00:12:54: a clever strategic move.

00:12:56: using constraints to drive innovation, which brings us, I think, to the core philosophical challenge that Barney Race-Jones laid out.

00:13:03: He frames this whole thing as an evolutionary test.

00:13:07: An evolutionary test.

00:13:08: That sounds serious.

00:13:09: It

00:13:09: is.

00:13:10: He argues that AI's ultimate impact on the planet hinges entirely on one thing.

00:13:15: Can we decouple intelligence growth from energy growth?

00:13:18: Right.

00:13:19: We can't just keep throwing more and more power at these models indefinitely.

00:13:22: Exactly.

00:13:22: So what's the answer?

00:13:23: If AI is going to consume all this energy, it has to, as he put it, earn its keep.

00:13:28: Earn its keep.

00:13:29: By pointing AI's capabilities at the world's biggest energy hogs, the electrical grid itself, steel production, cement manufacturing, massive logistics operations.

00:13:38: Okay, so use AI to optimize the biggest emitters.

00:13:42: Precisely.

00:13:43: Ensure that the efficiency gains AI delivers in those sectors dramatically outweigh its own energy consumption costs.

00:13:49: The goal has to be a net positive environmental impact.

00:13:54: AI needs to solve more problems than it creates, environmentally speaking.

00:13:58: Okay, so to sort of synthesize the key takeaway here for you, the listener, the shift we're seeing right now.

00:14:05: It's fundamentally about systemic change.

00:14:07: It's moving way beyond just isolated projects.

00:14:10: It's about integrating efficiency from the absolute micro level, like prompt design, those specific SQL queries we talked about, all the way through the infrastructure level, eight hundred VVC power, liquid cooling and right up to governance like phenops and green ops and even global strategy.

00:14:26: It's efficiency woven through the entire stack.

00:14:28: That really is the crux of it, isn't it?

00:14:30: Companies that manage to anchor the sustainable thinking early on, right in the design and planning phase, they don't just increase efficiency and save money today.

00:14:38: They also make their systems genuinely fit for the future.

00:14:41: They ensure their innovation actually passes that evolutionary test, Barney Rice-Jones mentioned, and stays competitive in the long run.

00:14:47: Will put.

00:14:48: If you enjoyed this deep dive, new episodes drop every two weeks.

00:14:51: Also check out our other editions on cloud, digital products and services, artificial intelligence and ICT and tech insights, health tech, defense tech.

00:14:59: Thank you for joining us and remember to subscribe.