Best of LinkedIn: Sustainability & Green ICT CW 38/ 39

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about sustainability and green ICT in calendar weeks thirty-eight and thirty-nine.

00:00:10: Frennus 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: Okay, let's unpack this.

00:00:25: Our mission today is a deep dive into the, well, the practical levers and strategic shifts that are really dominating the conversation around sustainability and green ICT globally.

00:00:36: Yeah, and we're drawing specifically from the most valuable professional insights shared just over the last two calendar weeks on LinkedIn.

00:00:42: Exactly.

00:00:43: So this isn't just about... distant roadmaps or vague corporate promises.

00:00:47: No, it's about actionable, right now, insights for professionals like you, folks making procurement, design, and strategic decisions today.

00:00:54: So

00:00:54: we're going to cover quite a bit.

00:00:56: The paradox of AI is growing footprint energy in water versus its potential.

00:01:00: Also, the necessary evolution of data center infrastructure, definitely.

00:01:03: And the rise of green software as a real commercial differentiator now.

00:01:07: And finally, that urgent push for accountability basically through better transparency.

00:01:12: So what does this all mean for you?

00:01:14: What means quickly getting visibility into the trends shaping eye-keep procurement, architectural choices, and frankly, competitive strategy?

00:01:23: We've clustered these insights into four key areas that really stood out in the professional dialogue across the green ICT space these past weeks.

00:01:30: Okay, where do we start?

00:01:31: Let's jump in with a topic that consumes everyone's attention, literally.

00:01:35: Theme one, AI's energy paradox and the efficiency frontier.

00:01:40: Ah, AI, yeah.

00:01:42: It always starts there.

00:01:43: Right.

00:01:43: It's the engine of digital transformation, sure.

00:01:46: But we have to reckon with the, well, the brutal realities of its energy consumption.

00:01:50: And the scale here is just, it's fascinating with being reported.

00:01:54: Stelman Q shared some numbers.

00:01:56: They're staggering.

00:01:57: Going.

00:01:57: By twenty thirty, AI workloads are projected to consume one hundred and fifty-six gigawatts globally.

00:02:02: One hundred

00:02:03: and fifty-six gigawatts.

00:02:05: Yeah.

00:02:05: And to put that in perspective, that's apparently seventy-one percent of total data center power demand.

00:02:10: That's a massive, like, three point five x increase in AI power used from today.

00:02:15: Wow.

00:02:16: That kind of trajectory.

00:02:18: It places immense, maybe impossible pressure on global grids.

00:02:22: Absolutely.

00:02:23: You hear seventy one percent and you have to ask, is that even viable?

00:02:27: Is the grid going to support it?

00:02:29: Or are we going to see regulators stepping in and restricting deployments?

00:02:32: It's a real question.

00:02:33: And beyond energy, Dagmar Eisenbach shared insights via Nadine Clark, pointing out the quote, invisible thirst of AI.

00:02:41: The water consumption, right.

00:02:42: Exactly.

00:02:43: A single hyperscale data center can use up to five hundred thousand gallons of water per day.

00:02:48: for cooling

00:02:49: half a million gallons a day.

00:02:50: that kind of water stress.

00:02:51: it's becoming a huge factor in where you even build these things.

00:02:54: it absolutely is which you know raises the next question if the operational cost is that high how do we mitigate?

00:03:00: right?

00:03:00: and the answer according to insights shared by folks like Angela Sorte and Elise Altanoff seems to lie less and just bigger infrastructure

00:03:07: and more and smarter architectures.

00:03:09: precisely Professionals are actively talking about needing smaller task-specific models, not relying on these giant foundational models for every single job.

00:03:20: Okay, but here's where it gets really interesting for me.

00:03:23: It's not just the running costs, the operational side.

00:03:26: You're talking embodied carbon.

00:03:28: Exactly.

00:03:28: The footprint created before the model even starts training.

00:03:32: Mark Butcher referenced a critical study on the NVIDIA A- one hundred GPU.

00:03:37: Yes, I saw that.

00:03:38: The manufacturing impact.

00:03:39: It's

00:03:39: huge.

00:03:40: Manufacturing accounts for a massive eighty one point eight percent of the total climate impact and eighty percent of fossil resource depletion for that GPU.

00:03:48: Eighty percent.

00:03:49: before it runs anything.

00:03:51: Right.

00:03:51: So focusing only on operational efficiency while important isn't nearly enough.

00:03:56: Yeah, if we're just checking out hardware after two years, we've basically locked in most of its negative environmental impact way before its useful life is over.

00:04:04: It

00:04:04: makes you rethink the whole life cycle.

00:04:05: So the solution isn't just scaling down models,

00:04:08: maybe.

00:04:08: It's also about engineering better models from the get-go.

00:04:11: Yeah.

00:04:11: Boris Gamazaichikov highlighted Alibaba's Quinn-III next model.

00:04:15: What was special about that one?

00:04:16: It achieved high performance.

00:04:18: but with only three billion active parameters.

00:04:20: That's a significant push towards a better performance efficiency frontier, you know, balancing quality with the actual resource cost.

00:04:29: Makes sense.

00:04:30: And we also saw a big focus on kind of democratizing the sufficiency, bringing it down to the individual user level.

00:04:36: Yeah, Julian Natalie Koenig was urging teams towards mindful sustainability practices.

00:04:42: Like what specifically?

00:04:43: Things like more efficient AI prompting.

00:04:46: Summarizing long email threads instead of feeding them whole Basically reducing the waste generated by us, the users.

00:04:53: And then on a larger scale, Oliver Cronk explored the potential, well, the potential game changer of on-device AI.

00:04:59: Like the rumored Apple Veritas.

00:05:01: Exactly.

00:05:02: If we distribute that influence work across devices already out there, your phone, your laptop, instead of concentrating it all on new power-hungry hyperscale

00:05:09: centers.

00:05:10: That fundamentally changes the infrastructure economics.

00:05:12: Yeah.

00:05:13: It pushes the load onto devices that are already powered, often in existing green zones potentially.

00:05:17: Totally

00:05:18: different model.

00:05:18: which takes us pretty neatly into our second theme.

00:05:21: Data centers, infrastructure, and crucially, grid alignment.

00:05:26: If AI is demanding all this capacity, the underlying digital infrastructure has to adapt fast.

00:05:32: And the core message here seems to be that sustainability is getting baked into the design now, not just bolted on later.

00:05:38: Yeah, we're seeing that in location strategy.

00:05:40: Joe Bell noted the expansion of the Swiss operator Green Data Center into Germany, specifically Frankfurt.

00:05:46: Right, a key hub.

00:05:48: And their strategy is very focused, providing energy efficient, high availability capacity, specifically to capture that booming AI growth.

00:05:55: And location is also critical for no carbon accountability.

00:05:59: Anna Horner and Bill Mandra discussed placing options technologies new AI data center way up in Iceland.

00:06:05: Iceland.

00:06:06: So capitalizing on that cheap, abundant green power, geothermal hydro.

00:06:10: Exactly.

00:06:11: But isn't

00:06:11: there a trade-off?

00:06:12: Doesn't a remote location like Iceland create like significant latency compared to a major hub like Frankfurt?

00:06:19: Which factor wins out in bids?

00:06:21: That's the key question, isn't it?

00:06:22: And it depends.

00:06:23: For general cloud use, yeah, latency is still king.

00:06:26: Right.

00:06:26: But for certain compute-heavy jobs, especially AI training that isn't real-time, the cost savings and the verifiable green credentials of Iceland?

00:06:35: Well, they're starting to outweigh that latency penalty for some users.

00:06:39: So the demand for actual green power is strong enough now to justify slightly slower speeds in some cases?

00:06:44: It

00:06:45: seems so.

00:06:45: For specific sectors, yes.

00:06:47: It shows a shift in priorities.

00:06:48: That makes sense.

00:06:49: Okay, what about cooling?

00:06:51: Still central, especially with the heat density of AI gear.

00:06:54: Oh,

00:06:55: absolutely.

00:06:55: Melanie Nakagawa highlighted significant investment going into microfluidics based.

00:07:00: Get this.

00:07:02: Zero water cooling systems.

00:07:03: Zero

00:07:03: water.

00:07:04: That tackles that half million gallon a day thirst problem directly.

00:07:07: Exactly.

00:07:08: And Andrew Bus shared this fantastic real world example.

00:07:12: Norway's new supercomputer, Olivia.

00:07:14: It's at left.

00:07:15: all mine data centers.

00:07:16: In

00:07:16: an old mine.

00:07:17: Yeah.

00:07:17: And it uses a hundred percent direct to chip liquid cooling, plus it draws cold water directly from the adjacent fjord.

00:07:24: That's resilience and huge efficiency combined.

00:07:26: Wow.

00:07:27: Beyond just the cooling tech itself, there's also grid flexibility, right?

00:07:30: Yes.

00:07:31: Douglas Mutant pointed this out.

00:07:32: Data centers agreeing to dial back their load during peak grid hours.

00:07:36: Google's shown this works, haven't they?

00:07:38: They have.

00:07:39: And it protects consumers, stabilizes the grid, and crucially avoids firing up expensive dirty fossil fuel peaking plants.

00:07:47: The data center becomes less of a constant drain, more of a grid partner.

00:07:51: Interesting dynamic shift.

00:07:53: And we're also seeing this surprising convergence.

00:07:55: Energy and compute right at the edge.

00:07:58: Corey Bloom announced a partnership, Electra and EG for electronics.

00:08:01: What are they doing?

00:08:03: Delivering battery systems integrated directly with AI micro data centers.

00:08:07: So the battery isn't just backup power anymore.

00:08:09: It's like

00:08:09: a gateway.

00:08:10: Exactly.

00:08:11: A gateway into the global AI economy built right into the energy storage.

00:08:15: This level of integration.

00:08:17: It just hammers home a point we saw across the sources.

00:08:21: Buyers, you need to interrogate providers' operational metrics much more rigorously now.

00:08:25: Absolutely.

00:08:26: Forget the glossy pledges.

00:08:28: Ask for the data.

00:08:29: Which leads us perfectly into theme three.

00:08:31: green software engineering and IT operations.

00:08:34: Right.

00:08:34: Moving from the hardware to the code and the ops.

00:08:37: And this topic, green software, it's clearly moving past being just, you know, a nice to have.

00:08:42: Oh yeah.

00:08:43: Crystal Levinkaya highlighted how it's becoming a genuine strategic advantage.

00:08:47: It directly ties cost reduction and reliability improvements to a competitive edge in RFPs and tenders.

00:08:54: So if you want to win business, you need to show your efficient.

00:08:56: Pretty much.

00:08:57: And the key takeaway here seems to be integrating sustainability early, like really early.

00:09:03: Philip Kirsting emphasized that green IT starts right back in requirements engineering, or RE.

00:09:08: Yeah, for listeners, maybe not in dev, RE is that initial design stage, defining what the system actually needs to do.

00:09:15: So the idea is, if the requirements don't demand massive compute power,

00:09:19: then don't provision massive compute power.

00:09:21: Be fundamentally lean from the very first line of the spec.

00:09:24: Don't just try to optimize bloated code later on.

00:09:27: And what's impressive is how these small early choices or even simple operational changes translate into measurable, often massive impact.

00:09:35: Keith Kavanaugh shared a fantastic real-life case.

00:09:38: A team simply snooze two Azure development VMs.

00:09:42: Turn them off overnight and on weekends.

00:09:44: Okay, standard practice for saving money, maybe.

00:09:47: Right, but look at the numbers.

00:09:48: Powering them off for a hundred and thirteen hours a week.

00:09:51: Massively reduced their CO-II emissions.

00:09:53: Went from ninety-eight kilograms down to thirty-two kilograms CO-II annually.

00:09:58: And it cut their cloudsband by roughly sixty-six percent a year.

00:10:02: Sixty-six percent cost saving.

00:10:04: Just by turning off two dev VMs you weren't using at two AM on a Saturday?

00:10:08: It's the ultimate low-hanging fruit.

00:10:10: It perfectly shows how cost reduction and sustainability are basically two sides of the same coin in IT ops.

00:10:17: No kidding, that's a powerful example.

00:10:19: And for teams, maybe in organizations where green thinking isn't quite mature yet, Anita Schutler offered some practical advice.

00:10:25: Like what?

00:10:26: Start by measuring from outside in.

00:10:28: Don't dive straight into complex code level metrics if the culture isn't ready.

00:10:32: So start simpler.

00:10:33: Yeah, maybe just count your virtual machines, track basic utilization, focus on practical levers that scale without messing up the user experience, things like workload rights, sizing efficient data pipelines, get some quick wins.

00:10:44: Okay, that makes sense.

00:10:45: Build momentum.

00:10:46: Now, shifting focus.

00:10:49: How do we actually hold all these providers infrastructure software accountable for the sustainability gains they claim?

00:10:56: The crucial question.

00:10:58: This takes us to theme four.

00:10:59: Transparency, policy, and accountability.

00:11:02: Yeah,

00:11:02: if we want these efficiency gains to be real and scalable, governance is absolutely critical.

00:11:07: And transparency.

00:11:09: Oh, it's a major sticking point right now.

00:11:11: A real pain point.

00:11:12: Oh,

00:11:12: so.

00:11:12: Reporting gaps.

00:11:14: Sicariate Drift mentioned reviewing a major cloud providers report and noted a classic issue.

00:11:19: They highlighted a decrease in carbon intensity.

00:11:22: Which

00:11:22: sounds good.

00:11:23: But their total emissions actually went up by six percent.

00:11:25: Oof.

00:11:26: That's the definition of green-watching risk, isn't it?

00:11:28: If your intensity improves but your overall footprint grows because you're just doing more, the net impact is still negative.

00:11:35: Exactly.

00:11:36: Dr.

00:11:36: Sasha Lucioni and others were calling for standardized, transparent reporting specifically on AI's environmental impacts.

00:11:43: We need to get past some marketing claims and enable actual fair comparisons.

00:11:48: We need consistent standards.

00:11:49: Which is why policy and coalitions are becoming so important.

00:11:52: Right.

00:11:53: Sunina Aitin believes that leveraging regulations like the EU AI Act, can actually drive sustainable AI innovation.

00:12:01: How?

00:12:02: By forcing models to meet safety and security standards, which inherently requires better documentation, efficiency, and transparency.

00:12:10: Interesting angle.

00:12:11: And we're seeing national coalitions step up, too.

00:12:14: Yes, the National Coalty Durzime Digitalizering, or NCDD, in the Netherlands, is very active.

00:12:20: Jus van Lier presented the first version of their enterprise architecture principles for sustainability.

00:12:26: So practical guidelines for companies.

00:12:28: Exactly.

00:12:28: Guiding organizations on how to embed sustainability is an integral starting point for their entire digital strategy, not an add-on.

00:12:34: So the operational shift is towards much more granular governance then.

00:12:38: It seems so.

00:12:39: Leaders are pushing beyond those high-level annualized claims.

00:12:42: They want life-cycle-based KPIs.

00:12:44: Like what Robert Kias and others are working on.

00:12:46: Disclosing metrics down at the PROMS level or the workload level.

00:12:49: Precisely.

00:12:50: That's the kind of granularity you need for accurate, maybe even real-time, benchmarking.

00:12:54: It lets procurement teams actually verify vendor claims in a meaningful way.

00:12:58: Okay, we've covered a lot of crucial ground today from the frankly staggering scale of AI's energy and water demands.

00:13:06: And that huge embodied carbon in things like an A-one hundred GPU.

00:13:10: To

00:13:10: the real competitive advantage now being delivered by green software.

00:13:15: And the absolute need for new digital architecture principles and much better governance.

00:13:20: Yeah, what really stands out to me looking across all these posts is that the focus has clearly shifted.

00:13:24: It's not if we need to optimize anymore.

00:13:27: No, it's definitely How?

00:13:28: How do we implement concrete solutions?

00:13:30: And it requires, well... deep collaboration, developers, info architects, procurement teams, policymakers, everyone needs to be involved.

00:13:38: Aligning that IT strategy with genuine environmental outcomes.

00:13:42: Exactly.

00:13:42: Well, if you enjoyed this deep dive, new deep dives drop every two weeks.

00:13:47: Also check out our other editions covering cloud, digital products and services, artificial intelligence and ICT and tech insights, health tech and defense

00:13:55: tech.

00:13:55: Thank you for joining us for this deep dive into the latest trends shaping the green ICT landscape.

00:14:01: Subscribe so you don't miss our next deep dive.