Best of LinkedIn: Sustainability & Green ICT CW 02/ 03

00:00:00: This episode is provided by Thomas Allgeier and Frennis, based on the most relevant LinkedIn posts about sustainability and green ICT in CWO two and O three.

00:00:10: Frennis supports ICT enterprises with market and competitive intelligence, decoding green software developments, benchmarking emerging standards, tracking regulatory shifts and analyzing competitor strategies.

00:00:22: So welcome back to the deep dive.

00:00:25: Today, we're digging into calendar weeks two and three of twenty twenty six.

00:00:30: And I have to say, looking through the posts, the whole tone, it just feels different.

00:00:35: Like the industry is waking up to a really loud alarm clock.

00:00:38: That is a perfect way to put it.

00:00:39: Yeah.

00:00:39: You know, for so long, sustainability in tech felt like, like a separate department, right?

00:00:45: Quotely siloed.

00:00:45: Exactly.

00:00:46: You had your core strategy and then, you know, once a year, you put out the sustainability report.

00:00:50: But what we're seeing now.

00:00:51: It feels like those two worlds have just slammed into each other.

00:00:54: Yeah,

00:00:54: it doesn't feel like we're just talking about compliance anymore.

00:00:57: Not at all.

00:00:57: I mean, this isn't about recycling bins or, you know, printing double-sided.

00:01:01: We are talking about the fundamental architecture of our digital world.

00:01:05: And the physical survival of it.

00:01:06: Yes.

00:01:06: Yeah.

00:01:07: It's a head-on collision between this massive exponential energy appetite of AI and the very real, very physical limits of our power grids.

00:01:18: and our water supplies.

00:01:19: It's become an engineering crisis as much as a climate one.

00:01:22: We have a lot to get through today.

00:01:23: I mean, from this energy blind spot of AI to some really surprising things like wooden data centers.

00:01:30: Oh, that financial case study is going to turn some heads too.

00:01:33: But yeah, we have to start with a big picture first.

00:01:35: Let's

00:01:35: do it.

00:01:35: So, theme one, the AI energy paradox in, well... all the risks that come with it.

00:01:41: We saw a really striking post from Azra Alam that frames this perfectly.

00:01:45: He calls it a massive energy blind spot.

00:01:48: This is

00:01:48: such a critical concept.

00:01:50: Alam points out this paradox that I think catches a lot of people off guard.

00:01:54: Right.

00:01:55: If you look at the technical progress in just three years, the cost of AI inference, that's the compute you need to actually use a model, has dropped by a factor of a thousand.

00:02:04: It's become incredibly efficient.

00:02:05: And in almost any other industry, if you make something a thousand times more efficient, your resource use goes way down.

00:02:11: It

00:02:12: goes down.

00:02:12: Exactly.

00:02:13: But here we have the Jevons paradox.

00:02:15: When a resource like Compute gets cheaper, demand doesn't just stay flat.

00:02:20: It explodes.

00:02:21: All on highlights that because it's so cheap to use AI, we're putting it everywhere.

00:02:25: So while the unit cost is down, the total energy consumption is just.

00:02:29: skyrocketing.

00:02:30: He had some numbers that were frankly hard to process.

00:02:32: He said global data center electricity is set to more than double to nine hundred and forty five terawatt hours by twenty

00:02:40: thirty.

00:02:41: Just to put that number in perspective nine hundred and forty five terawatt hours is it's roughly the annual electricity use of a major industrial country like Japan.

00:02:49: So we're adding a new Japan to the grid just for servers.

00:02:52: We basically are.

00:02:54: And he drilled down on the U.S.

00:02:55: saying data centers could go from four percent of national electricity to maybe twelve percent.

00:02:59: It's not a small increase.

00:03:00: That is a structural change to the entire grid.

00:03:02: A huge change.

00:03:04: And John Beger added another layer to this.

00:03:06: He noted that we focus on inference, the user side, but the training of these models is an industrial beast.

00:03:12: Right.

00:03:13: The initial build.

00:03:14: Training a frontier model now uses electricity comparable to you know, a national infrastructure project.

00:03:20: This isn't light software anymore.

00:03:22: It's heavy industry that just happens to be digital.

00:03:24: Heavy industry with a very sleek UI.

00:03:27: But underneath, it's all heat, metal, and electricity.

00:03:31: And speaking of heat, that brings us to the other resource Masoud Barakzai brought up, water.

00:03:36: Yeah, this is the wet cost of AI.

00:03:39: It gets overlooked all the time.

00:03:41: Data centers are incredibly hot.

00:03:43: And to cool them, many systems evaporate huge volumes of fresh water.

00:03:48: Barakzai said the consumption now rivals the entire global bottled water industry.

00:03:52: That's

00:03:52: a staggering visual.

00:03:54: So every time I'm chatting with an AI, somewhere a tiny invisible bottle of water is being poured out to keep the servers cool.

00:04:00: That's essentially it.

00:04:01: Yes.

00:04:01: And in places already facing drought, I mean, this becomes a huge political issue.

00:04:05: You've got data centers competing with farms for water.

00:04:07: Which explains why the tone from the financial side is shifting.

00:04:10: We saw Inel again talking about investor sentiment.

00:04:14: The vibe has definitely changed in the boardrooms.

00:04:16: Lagenz says investors are now demanding to know the footprint per use case.

00:04:21: So it's a license to operate issue.

00:04:23: It is.

00:04:23: You can't just pitch a cool AI tool anymore.

00:04:27: If the metabolic cost of that tool is too high, the capital is just going to walk away.

00:04:32: Yeah.

00:04:32: Because it's a risk.

00:04:33: That's

00:04:33: huge risk.

00:04:34: Precisely.

00:04:35: And Nolan Gadard took that risk assessment even further.

00:04:38: She argues we're facing a compound threat.

00:04:41: It's not just climate risk.

00:04:42: It's climate plus cyber plus AI risk.

00:04:45: all tangled together.

00:04:46: So how does that work?

00:04:47: Like a chain reaction?

00:04:48: Yeah,

00:04:49: imagine a cyber attack on a power grid that's already strained by AI demand, all happening during a climate change driven heat wave.

00:04:56: It's a cascading failure.

00:04:57: Godard's point is that this creates an insurance gap.

00:05:00: Insurers are looking at this mess and saying, This is becoming uninsurable.

00:05:04: It's a pressure from all sides.

00:05:06: The grid, the water, the insurers.

00:05:07: We can't just keep building the same way, which, I mean, is a perfect segue to our second theme.

00:05:12: Green infrastructure.

00:05:14: Yes, this is a crucial pivot.

00:05:16: PS Lee introduced a concept I think we're going to hear a lot more about.

00:05:18: He says, we need to stop viewing data centers as AI factories.

00:05:22: A factory that doesn't apply, you know, a one-way street.

00:05:26: Raw materials in, product doubt, waste out the back.

00:05:29: Exactly.

00:05:30: Lee argues we should treat them as climate assets.

00:05:33: Okay, that sounds like a great rebrand, but what does that mean in practice?

00:05:36: How does a building full of servers become a climate asset?

00:05:40: It means the facility is designed to solve energy problems, not just create them.

00:05:45: For one, designing for two hundred and forty seven carbon free energy.

00:05:49: But the really big one is treating waste heat as a resource.

00:05:52: Great capturing all that heat.

00:05:54: Yeah,

00:05:54: if you can capture that heat and pipe it into a district heating system for homes or use it in greenhouses, you turn a waste product into a revenue stream.

00:06:02: You actually integrate the data center into the community.

00:06:05: That's

00:06:05: the ideal.

00:06:06: But the transition is messy.

00:06:07: Anna Leonard-Nesbitt shared a reality check from Phoenix, Arizona that really stuck with me.

00:06:11: The Phoenix story is fascinating.

00:06:13: It's a perfect example of what she calls the decarbonization paradox.

00:06:17: So laid out for us, Phoenix is a desert, tons of sun, so lots of solar potential, but a huge data center boom.

00:06:25: A massive boom.

00:06:26: They have thirty thousand megawatts of data center requests in the queue.

00:06:31: That's four times their current peak load.

00:06:33: Wow.

00:06:34: And the paradox is, despite all that new solar power, emissions are actually going up.

00:06:39: How

00:06:39: is that even

00:06:40: possible?

00:06:40: It's the timing.

00:06:42: Solar is intermittent.

00:06:43: It works when the sun shines, but AI data centers are always on.

00:06:47: They have a flat high load, two forty seven.

00:06:50: So when the sun goes down, they still need massive power and the utilities have to fire up gas plans to cover that night shift.

00:06:57: So adding renewables doesn't automatically clean up the grid if the demand doesn't match when the power is being made.

00:07:03: Exactly.

00:07:03: It's a grid jam.

00:07:04: But while the energy side is tricky... We did see some really cool innovation on the construction side, less about power, more about materials.

00:07:12: Carl Rabe was posting about wooden data centers.

00:07:15: I have to admit, wooden data centers sounds like a fire hazard from the eighteen hundreds, but this is high tech, right?

00:07:20: Yeah, we aren't talking about log cabin.

00:07:22: This is engineered timber, like cross laminated timber.

00:07:25: It's incredibly strong and fire resistant.

00:07:27: But the key is what's called embodied carbon.

00:07:30: That's the carbon emitted just to make the building materials in the first place.

00:07:33: Correct.

00:07:34: Concrete and steel are incredibly carbon intensive.

00:07:38: Wood stores carbon.

00:07:40: RABE even calls using Timber a status symbol.

00:07:43: It signals you're thinking about the entire life cycle, not just operations.

00:07:47: And Victor Patek added to this with the trend of prefabricated data centers.

00:07:51: Oh, prefab is the unsung hero here.

00:07:53: Instead of a messy construction site, you build modular units in a factory.

00:07:57: Patek notes that prefab generates over fifty percent less embodied carbon per square meter than a traditional concrete build.

00:08:04: It's faster, cleaner, just much less wasteful.

00:08:08: But sometimes it's not how you build but where.

00:08:11: Alice Edge shared a great example of vivo power buying a hydropower data center in Norway.

00:08:16: This

00:08:16: goes right back to the climate asset idea.

00:08:18: It's about geography.

00:08:20: If you're training a huge AI model, it doesn't need to be in downtown Phoenix.

00:08:23: You don't need low latency for training.

00:08:25: So you move the compute to the power.

00:08:27: And Norway has tons of hydropower and natural cooling.

00:08:30: Exactly.

00:08:31: Instead of bringing the power to the data center, you take the data center to the power.

00:08:35: You create a sustainable ecosystem.

00:08:37: that just isn't possible in a water-stressed desert.

00:08:40: Okay, so we've covered the shell, the timber buildings, the Norwegian hydro plants.

00:08:45: Now let's look inside.

00:08:47: Theme three, green software and architecture.

00:08:51: This is where the rubber really meets the road for developers.

00:08:55: You can have the greenest building in the world, but if your code is inefficient, you're still wasting energy.

00:09:00: Anita Schutler made a compelling case that software architecture is a climate issue.

00:09:05: She used this comparison of monolithic systems versus verticalization.

00:09:09: Can you break that down?

00:09:10: Sure.

00:09:11: Think of a monolithic system like a giant factory, where everything, all the lights, all the machines is on a single on-off switch.

00:09:17: So even if you just need one little machine, you have to power up the whole building.

00:09:21: You got it.

00:09:21: Incredibly wasteful.

00:09:23: But that's how a lot of legacy software works.

00:09:25: Verticalizing means you break it down into smaller components, so you only power up the specific resources you need for a specific task.

00:09:32: And this isn't just about saving the planet.

00:09:34: This is where the money comes back in.

00:09:37: Arwell Owen shared what might be the stat of the week about SAP systems.

00:09:41: This

00:09:41: was a massive aha moment.

00:09:43: His team managed to cut five million euros from a client's SAP on AWS spend.

00:09:49: Wait.

00:09:50: Five million euros?

00:09:51: That is not

00:09:51: pocket change.

00:09:52: Not at all.

00:09:53: And they did it just by applying fine ops and green computing principles.

00:09:56: It just proves that green IT is in a charity project.

00:09:59: It's a profitability strategy.

00:10:01: When you optimize your code, you are directly cutting your OPEX.

00:10:05: That's the argument that gets the CFO on board.

00:10:07: You tell them it's green, they smile, you tell them it saves five million euros, they sign the check.

00:10:12: Exactly.

00:10:13: But how do developers actually do this?

00:10:16: Hannah Herbst emphasized green coding.

00:10:18: Is that like a specific language or a methodology?

00:10:20: It's a methodology.

00:10:22: It's about optimizing algorithms to lower CPU load, just writing cleaner, tighter code.

00:10:27: But Nicholas Sundberg added another layer, moving from code to operations.

00:10:31: He talks about workload placement.

00:10:33: That sounds like the digital version of the Norway example.

00:10:36: It is.

00:10:37: Sundberg says we need to be smart about when and where we run big jobs.

00:10:41: If it's not time sensitive, don't run it at six p.m.

00:10:44: when the grid is dirty.

00:10:45: run it at two a.m.

00:10:46: when the wind is blowing, or move the job to a region with a cleaner grid.

00:10:50: To do that, you need to know which regions are green.

00:10:53: Kevin Leslie from GreenPixie shared a tool for this, the cloud region scorecard.

00:10:57: Transparency is everything here.

00:10:59: You can't manage what you don't measure.

00:11:01: This scorecard rates cloud regions from A plus down to D. It's like the energy sticker on your fridge.

00:11:08: So a developer can look and say, OK, US East is a D, but this region in Canada is an A plus E. I'll deploy there.

00:11:14: It empowers engineers to make those micro decisions that add up, and we're seeing this get more formal too.

00:11:20: David Kot mentioned the KDI open source framework received the Blue Angel certification.

00:11:24: The Blue Angel is a big deal in Europe.

00:11:26: It's a serious eco label.

00:11:28: Seeing software get that kind of certification shows we're moving past the Wild West era.

00:11:33: These are becoming standardized, verifiable metrics.

00:11:35: Okay, so we've covered the huge infrastructure and the invisible code, but we can't ignore the stuff we can actually touch.

00:11:43: Theme four brings us to circularity and the devices on our desks.

00:11:47: Yeah, this is often the forgotten piece.

00:11:49: We talk about the cloud, but we access it through plastic, glass, and rare earth minerals.

00:11:53: Rudd Neal had a great post about circular computing.

00:11:56: He was talking about remanufactured laptops, and he made a really important distinction between remanufactured and just, you know, used.

00:12:04: That distinction is critical.

00:12:06: A CTO isn't going to give their team beat up laptops.

00:12:10: Neil argues that true re-manufacturing restores the device to like new or even better than new performance.

00:12:16: It removes the trade off.

00:12:18: So you get a top tier machine, but you haven't had to dig up a mine for new cobalt?

00:12:22: Exactly.

00:12:23: And Nabil K reinforced this with a concept he calls digital sobriety.

00:12:27: I love

00:12:27: that

00:12:28: term.

00:12:28: It's a simple but powerful idea.

00:12:30: The most sustainable device is the one you already own.

00:12:33: We have this culture of constant refresh cycles.

00:12:36: Digital sobriety just challenges that.

00:12:38: Do you actually need the upgrade or do you just want it?

00:12:40: Speaking of habits, Graham Donnelly offered a really interesting checklist for remote businesses.

00:12:46: And he dropped a fact about pensions that honestly blew

00:12:49: my mind.

00:12:49: The twenty-one X stat.

00:12:51: Yes.

00:12:51: He said, moving your pension to a sustainable fund can have twenty-one times the impact of giving up flying.

00:12:57: It's all about leverage, right?

00:12:58: We focus on our visible actions recycling.

00:13:00: skipping a flight.

00:13:02: But the capital markets are the engine room.

00:13:04: If your pension fund is invested in fossil fuels, your money is working against your goals while you sleep.

00:13:10: Redirecting that capital is huge.

00:13:12: Donnelly also mentioned decarbonizing your data, basically deleting old emails and cloud junk.

00:13:19: Digital hygiene.

00:13:19: I think we're all guilty of this.

00:13:21: Retreat cloud storage like it's infinite.

00:13:24: But every gigabyte of useless data on a server is drawing power every second of every day.

00:13:28: My inbox is basically a digital archaeological site.

00:13:31: I definitely need to do some cleaning.

00:13:33: It all ends up.

00:13:33: So let's try to pull all of this together.

00:13:35: We've looked at the AI energy crisis, the shift to climate asset infrastructure, the profitability of green software, and the circular economy.

00:13:44: What's the big takeaway?

00:13:45: I think the best way to wrap this is with a thought from Lubamla Jordanova, who analyzed the WEF global risks report.

00:13:53: She talks about something called the horizon gap.

00:13:55: The horizon gap sounds ominous.

00:13:58: It is, but it's also a call to action.

00:14:01: The horizon gap is this disconnect between what companies are doing right now and what the data says they need to be doing for the next decade.

00:14:08: So companies are reacting to the noise right in front of them.

00:14:11: Trade wars, the latest AI hype.

00:14:14: hitting quarterly numbers.

00:14:16: Exactly.

00:14:17: They're optimizing for short-term volatility.

00:14:19: But if you look at the ten-year outlook in that report, environmental risks completely dominate.

00:14:24: Jordan Nova's point is that our short-term reactions are feeling to align with long-term reality.

00:14:29: That's a heavy thought.

00:14:30: It implies we might be efficient today, but we aren't building resilience for twenty thirty-five.

00:14:35: That's

00:14:35: the question I'd leave listeners with.

00:14:37: Are your current tech efficiencies actually building resilience?

00:14:40: Or are you just optimizing a system that won't survive the regulatory and physical climate of the next decade?

00:14:46: It pushes the conversation way beyond just being green.

00:14:50: It's about being future-proof.

00:14:52: If your entire AI strategy relies on cheap infinite energy and water, you might be building on sand.

00:14:58: Precisely.

00:14:59: The winners in the next decade won't just be the fastest.

00:15:02: They'll be the ones who close that horizon gap.

00:15:04: Well, on that note, we have plenty to think about before we upgrade our next laptop or deploy our next AI model.

00:15:10: Indeed.

00:15:11: It's a challenging time, but seeing all this innovation, it's also a very exciting

00:15:15: one.

00:15:16: If you enjoyed this episode, new episodes drop every two weeks.

00:15:19: Also, check out our other editions on cloud.

00:15:21: digital products and services, artificial intelligence, and ICT and tech insights, health tech, defense tech.

00:15:28: Thanks for listening everyone.

00:15:29: Don't

00:15:29: forget to subscribe.

00:15:31: See you next time.