Best of LinkedIn: Sustainability & Green ICT CW 48/ 49

00:00:00: This episode is provided by Thomas Alguyer and Frennis, based on the most relevant LinkedIn posts about sustainability in green ICT in CW-Forty-Eight and Forty-Nine.

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

00:00:21: Welcome to the deep dive.

00:00:23: So our core mission today really is to get into the practical shifts happening right now in sustainability and green ICT.

00:00:29: I think we've moved past the phase of just setting these abstract far-off

00:00:34: goals.

00:00:35: Oh, absolutely.

00:00:36: The conversation now is all about concrete, measurable execution.

00:00:39: You can feel the pressure building.

00:00:40: It

00:00:41: is.

00:00:42: Looking at the insights from the last few weeks, we really stood out.

00:00:45: We're four critical areas where the industry is moving from potential to actual precision.

00:00:50: We're going to look at the convergence of efficiency practices, so phenops and green ops.

00:00:54: Then there's the immense and truly radical innovation happening in data centers.

00:00:59: After that, the urgent challenge of measuring and dealing with the massive AI power surge.

00:01:05: And finally, how policy and system design are really proving to be the ultimate drivers here.

00:01:12: Let's start with that convergence.

00:01:13: It's a great place to begin because it's the low hanging fruit, but with a huge impact.

00:01:19: We're talking about this mandate to integrate fine ops and green ops.

00:01:22: Yes.

00:01:22: And what's fascinating is that at their core, both of them have the exact same goal, eliminating waste.

00:01:28: And the scale of that waste when you see the numbers is just startling.

00:01:32: I saw a post from Kamalash Kumar, for instance, who put a clear number

00:01:35: on it.

00:01:35: What was it?

00:01:36: Roughly, thirty percent of cloud spending is just wasted on idle or mismanaged resources.

00:01:42: And every dollar wasted there is compute cycles wasted.

00:01:45: Which translates directly into needless carbon emissions.

00:01:48: Exactly.

00:01:49: So the logic is powerful.

00:01:50: If you make the system more financially efficient, you automatically make it greener.

00:01:54: That's the point Bjorn Stengel and Philippa Vinara from IDC were making.

00:01:58: They see these ops practices as delivering truly efficient cloud solutions because they break down the silos.

00:02:04: The silos between finance, product, engineering.

00:02:08: All

00:02:08: of them.

00:02:08: It drives collaboration and data-driven decisions.

00:02:11: But isn't the real challenge getting engineers you know, whose main focus is performance and delivery to actually prioritize the planet alongside the budget, that's been a struggle for years.

00:02:21: It has,

00:02:22: but that's where the implementation of green ops becomes so important.

00:02:26: Kevin Leslie from AWS re-invent, he noted that the key is linking the budget directly to the planet.

00:02:32: How so?

00:02:32: You have to give engineers cost recommendations together with the environmental metrics.

00:02:37: So here's the cost, and here's the greenhouse gas, the kilowatt hours, the water use.

00:02:41: So it makes it a mandated priority, not just a nice-to-have suggestion.

00:02:45: Instantly.

00:02:46: And that immediately pushes the focus right into software architecture.

00:02:50: Anita Schutler highlighted this.

00:02:51: She said future-proof IT systems depend on smart, flexible design.

00:02:56: So modular architecture, being conscious about the tech you select.

00:02:59: And interoperability.

00:03:00: That flexibility is everything for green software.

00:03:03: It lets your systems react to the energy grid in real time.

00:03:06: Your computing loads become movable.

00:03:08: So you can pause long processes, adapt service requirements, and

00:03:12: chase renewable energy when it's available instead of just running nonstop on whatever the grid gives you, which might be fossil fuels,

00:03:18: which brings us to the code itself.

00:03:20: Then if architecture is the canvas, the code is the paint.

00:03:24: We need green paint.

00:03:25: Exactly.

00:03:26: But you know, the push for true green coding standards is it's hitting some hurdles.

00:03:31: Johnny Cusso pointed out that while Phenops is pretty established.

00:03:35: We can measure dollars easily.

00:03:36: Right.

00:03:37: But measuring the carbon impact of green coding is still really complex.

00:03:41: It's hard to measure the precise impact of one line of code.

00:03:44: Too many variables.

00:03:45: Hardware, the grid, resource intensity.

00:03:49: It's a challenge, but efforts are gaining traction.

00:03:52: Cutsa did mention some open source tools that Creedango that are starting to pop up.

00:03:56: And what do they do?

00:03:57: They extend existing systems like SonarCube with eco design rules.

00:04:00: So it brings that environmental feedback right into the developer's workflow.

00:04:04: Okay, so let's move from the abstract code to the very physical infrastructure.

00:04:08: Theme two, data centers.

00:04:11: This is where the challenge feels most immediate.

00:04:13: Oh, for sure.

00:04:14: I saw a post from Daniel Amor noting the scale we're heading for.

00:04:18: Global AI energy demand could hit one One thousand terawatt hours by twenty twenty six.

00:04:24: One thousand terawatt hours.

00:04:26: That's that's the consumption of an entire industrial country.

00:04:28: Our grids just aren't ready for that kind of sudden demand.

00:04:31: So the response has to be intense innovation.

00:04:34: And we're seeing a heavy focus on circularity, on turning waste into value.

00:04:39: Like

00:04:39: waste heat.

00:04:40: Exactly.

00:04:41: Peter Chanoff and Tina David Sainin highlighted the huge global movement toward waste heat reuse.

00:04:47: Look at the Microsoft and Fordham project in Finland.

00:04:50: That's the world's largest heat project, isn't it?

00:04:52: It is.

00:04:53: It's set to provide up to forty percent of the district heating for multiple cities.

00:04:56: That's a fundamental paradigm shift.

00:04:58: That's massive.

00:04:59: It changes the data center from just being a resource drain to being a core energy provider for a whole community.

00:05:05: And policy is starting to lock this in.

00:05:08: Muffit Yilmaz Gokman noted that Germany is making this mandatory.

00:05:12: Starting in twenty twenty six, data centers will have to reuse ten percent of their waste heat.

00:05:17: That's a powerful accelerator.

00:05:19: We're also seeing big shifts in power resilience technology away from just diesel generators.

00:05:24: Thank

00:05:24: goodness.

00:05:25: Joaquin Rodriguez, Antibon, and Anita Falls were pointing to the rise of hydrogen fuel cells.

00:05:30: Relco is using Toyota fuel cells, for instance, as a clean, zero-mission backup.

00:05:35: That's a huge move, and it forces a new way of thinking about long-term design.

00:05:40: Forrest Halualani of Compass Data Centers talks about designing for the next hundred years.

00:05:46: the mindset away from being a drain to being a positive grid partner.

00:05:50: Yes, and the tech is getting there.

00:05:52: Pierre Queer also mentioned that adopting silicon carbide, or six-leash technology, is contributing to this new level of efficiency in the power infrastructure itself.

00:06:01: Okay, hold on.

00:06:02: Here's where it gets really interesting.

00:06:03: I mean, it borders on science fiction.

00:06:05: Jonathan Mattis-Glender reported on a truly radical solution.

00:06:08: Which is?

00:06:09: Orbital data centers.

00:06:10: In space.

00:06:11: That sounds like something from a novel, but it's actually happening, isn't it?

00:06:15: It is.

00:06:15: SpaceX confirmed the launch of Starlink V-III satellites in twenty-twenty-six, and they're promising massive computing power in

00:06:22: orbit.

00:06:22: Wow.

00:06:23: But beyond the novelty, what are the real environmental advantages?

00:06:27: Why do this?

00:06:28: Well, it tackles the three biggest bottlenecks on the ground.

00:06:30: Yeah.

00:06:31: Land, power and water.

00:06:34: Okay, so in orbit, you get endless solar energy, zero land footprint and zero water for cooling.

00:06:39: That's a big one.

00:06:40: A huge one.

00:06:41: Heat just radiates into space.

00:06:43: It completely bypasses our terrestrial resource constraints.

00:06:47: That really puts the scale of the crisis into perspective.

00:06:50: which moves us to the software driving all that demand, the AI carbon tsunami.

00:06:54: Right.

00:06:55: We know AI can be a powerful enabler.

00:06:57: Microsoft's Sparrow project for wildlife tracking is a great example.

00:07:01: But the environmental impact is huge and often very opaque.

00:07:05: And this is where we see the strategic cost.

00:07:07: Rianne Riemanns noted that tech companies are quietly readjusting their sustainability goals because of AI's energy impact.

00:07:13: Really?

00:07:14: Yes.

00:07:15: The discussion is shifting away from climate pledges to the tactical need to just secure energy, even if that means using fossil fuels.

00:07:21: It suggests a kind of operational desperation.

00:07:24: Well, it's hard to secure energy when you can't even measure what you're using.

00:07:27: Paul Johnston pointed out that any of those energy-per-prompt claims you see.

00:07:33: They're pure guesswork.

00:07:34: Because no one is publishing verifiable data.

00:07:37: No one.

00:07:38: And Will Nordberg added that even the low numbers we do see, they only measure the spark, the initial process.

00:07:44: They ignore the whole ecosystem, the cooling, the networks, the orchestration.

00:07:48: It's like measuring the apple while ignoring the entire orchard.

00:07:52: It's why standardization is so urgent.

00:07:54: How do you compare one LLM against another?

00:07:56: That's

00:07:56: the exact problem Matthew Francois is tackling with the one token model, or OTM.

00:08:00: The OTM, okay, explain that.

00:08:02: Think of it as a bridge.

00:08:03: It lets generative AI usage be calculated using the existing software carbon intensity or SCI framework, and it uses a normalized unit called the Antarctica token.

00:08:14: The

00:08:14: Antarctica token, I like that, gives us a standard unit of measurement.

00:08:17: which we desperately need because the new AI energy score V-II data that Boris Gumbazichkov and Dr.

00:08:23: Soshil Chuyoni presented, it's a massive reality check.

00:08:27: What's the key takeaway from it?

00:08:28: It's that the reasoning models, you know, the ones that are truly analyzing and synthesizing information, they consume hundreds of times more energy than the non-reasoning versions.

00:08:38: Hundreds of times, wow.

00:08:40: An efficiency is less predictable.

00:08:42: We can't just assume that a newer model is automatically a more efficient one anymore.

00:08:46: And that gets even worse when you look at massive video models.

00:08:49: I saw Nathaniel Barola's analysis of the potential consumption of Sora II.

00:08:53: He projected it could require four hundred and eight megawatts of power at scale.

00:08:58: That is just an astronomical number.

00:09:00: It is.

00:09:01: And his analysis suggested that could generate one point nine million tons of kuyo annually.

00:09:07: Whether that's the maximum or typical usage, it proves the environmental conversation has to be a primary design constraint now.

00:09:13: It can't be an afterthought.

00:09:14: Which brings it all back to conscious usage.

00:09:17: Cynthia Jebor's observation was so spot-on.

00:09:19: Sustainable AI means using it where it creates real value, not just because you can.

00:09:24: Optimize the process first before you automate it with a power-hungry model.

00:09:28: Exactly.

00:09:29: And this leads us perfectly into our final theme.

00:09:31: policy, procurement, and what we're calling systemic coherence.

00:09:35: Right, because you can invent all the great tech in the world.

00:09:38: But if the system doesn't reward its adoption, it just won't scale.

00:09:42: Precisely.

00:09:42: There was a fantastic study on this by Remy Pakhu about AI optimized microgrids.

00:09:48: It's a really concrete example.

00:09:49: The AI achieved an eighteen percent cost reduction.

00:09:52: So great economic performance.

00:09:54: Great.

00:09:55: But the environmental benefits only aligned where the price signals reflected carbon intensity.

00:10:01: In South Australia, for instance, where energy price tracks how green the grid is, the AI chose the green path.

00:10:08: But not elsewhere.

00:10:09: In places like California, where price and carbon are decoupled, the AI could actually induce environmental costs.

00:10:16: It just optimized for the cheapest dollar, regardless of the source.

00:10:20: The AI does exactly what we tell it to.

00:10:22: It's a faithful servant.

00:10:23: That just reinforces this need for systemic change.

00:10:26: Julio C. Serrano, who was the lead author on the ERA report, argued we have to stop seeing data centers as just passive consumers.

00:10:33: They

00:10:33: need to be integrated energy assets that provide flexibility to the grid.

00:10:37: Not just demand from it.

00:10:39: And that structural change needs top-down buy-in.

00:10:42: Sanjay Potter and Kerry O'Donnell stress that embedding sustainability into the DNA of digital transformation leads to faster innovation.

00:10:50: It makes it a business imperative, not a line item in a CSR report.

00:10:54: And ultimately, the stick is reinforcing the carrot.

00:10:57: Dr.

00:10:57: Katharina Grimmie and Muffet Yilmaz Gokman both noted the rising regulatory pressure, especially the EU's CSRD.

00:11:05: The Corporate Sustainability Reporting Directive.

00:11:07: Yes, and that demands transparency and standardized reporting on ICT's footprint.

00:11:12: And that mandatory reporting is the structural lock-in.

00:11:16: It makes this data, whether it's carbon per token or heat reuse, an auditable business risk, a core financial concern.

00:11:23: So if you synthesize everything we've covered, It feels like the whole industry is moving from talking about potential to demanding precision.

00:11:30: I think that's a perfect summary.

00:11:31: We have the tools now, like the one token model green ops frameworks, and we have the technology from hydrogen fuel cells to orbital compute.

00:11:38: So this brings us to the real structural takeaway.

00:11:41: The posts we've seen show clearly that the biggest barrier to scaling green ICT is an invention anymore.

00:11:46: It's governance.

00:11:48: For true sustainability, we have to ensure that the economic systems we use, procurement, energy pricing, financial reporting, are all structurally coherent with our environmental goals.

00:11:59: So that doing was financially efficient.

00:12:01: Always aligns with doing what is environmentally responsible.

00:12:04: If we can get that coherence, that's the next frontier.

00:12:07: If you enjoy this deep dive, new episodes drop every two weeks.

00:12:11: Also check out our other editions on cloud, digital products and services, artificial intelligence, and ICT and Tech Insights, HealthTech, DefenseTech.

00:12:20: Thank you for tuning in for this deep dive.

00:12:22: Make sure to follow us to stay well informed.