Best of LinkedIn: Sustainability & Green ICT CW 06/ 07

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about sustainability in green ICT in CW six-and seven.

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

00:00:22: Hi everyone, it's really great to be back looking at another set of insights with you.

00:00:25: Yeah

00:00:26: glad we're doing this deep dive today!

00:00:27: So were looking at the trends across LinkedIn for calendar weeks six and seven.

00:00:33: I have to say, the mood in The Feed has really shifted lately.

00:00:36: It definitely is

00:00:37: Like for last year or so.

00:00:39: we've basically been this massive honeymoon phase with AI and digital transformation Lots of excitement lots talking about art of the possible.

00:00:47: Now it feels like the bill finally arrived

00:00:49: Exactly!

00:00:49: The posts that were curated by you today aren't about cool new prompts or these vague software promises anymore.

00:00:57: They're about really concrete, heavy physical problems.

00:01:01: Yeah, I mean we're talking massive gas turbines fundamentally broken measurement tools and legislation that is actually starting to bite into the industry.

00:01:11: It's a wake-up call

00:01:12: It is.

00:01:12: If I had to pin a single theme on this whole collection of insights for you today, it would just be accountability.

00:01:18: Accountability?

00:01:19: I like that!

00:01:19: Yeah because we're seeing this massive collision between the physical reality of AI—just the sheer mass of metal and energy it needs —and the sudden realization that our accounting methods aren't quite ready for it.

00:01:33: neither are the financial ones nor the carbon ones...

00:01:36: ...It IS messy but i think it makes for a super fascinating deep dive today Absolutely So.

00:01:40: we've structured this into three main clusters to keep it organized for you.

00:01:45: First, We're going look at AI and infrastructure.

00:01:47: the physical reality of all The hardware war is the infrastructure crisis?

00:01:51: The geography of compute

00:01:52: And then after that will move in two green software and cloud.

00:01:55: so the code reality looking At why our measurement tools might actually be lying To us and more importantly how the industry's trying to fix them

00:02:06: How policy digital sovereignty and activism are basically rewriting the whole map.

00:02:12: Let's just dive right into that physical reality then because The numbers coming out of the energy sector right now or frankly staggering

00:02:19: Staggering is putting it really mildly.

00:02:21: I was reading through the data shared by Boris Gamazichikov And Michael Thomas.

00:02:25: Oh, yeah That thread

00:02:27: had actually stopped me in my tracks.

00:02:28: they were looking at the pipeline for new power generation.

00:02:32: It's being built specifically to feed Data centers here in the US

00:02:36: and just to set the stage for you listening The industry narrative up to this point has heavily been that renewables are just gonna handle this load, right?

00:02:44: Right.

00:02:44: The idea was we'll Just build enough wind and solar to power the AI boom And everything will be fine.

00:02:49: exactly.

00:02:50: but Boris and Michael report That there are two hundred and fifty-two gigawatts of gas fired plants currently in development specifically To support their status.

00:02:58: enter demand.

00:02:58: Two

00:02:59: hundred and Fifty-two Gigawatts.

00:03:00: it's insane.

00:03:01: It's

00:03:02: a completely preposterous number.

00:03:04: I mean, to put that into perspective for you... That is roughly five times the total electricity demand of New York City.

00:03:10: just for data centers and this is gas we're talking about.

00:03:14: We aren't talking about green hydrogen or giant battery arrays.

00:03:17: We are talking about burning fossil fuels Just to keep the servers

00:03:21: humming.

00:03:21: And Michael Tom has added detail there that really highlights The sheer desperation in situation.

00:03:27: He noted.

00:03:27: datacenter developers Are actually planning To build fifty-six gigawatts of their own power plants.

00:03:33: They're not even waiting for the utility companies anymore?

00:03:36: Exactly,

00:03:36: that completely contradicts the idea that the grid is just going to seamlessly absorb this AI demand.

00:03:42: The grid is full and Q is too long so these tech companies are basically becoming power companies themselves.

00:03:48: And if you look at Boris's analysis on this If all that gas infrastructure actually gets built it could increase US energy in industry emissions by twenty seven point five percent.

00:03:59: That almost a third.

00:04:00: It wipes out years of hard-fought climate progress in one go.

00:04:05: Its not some marginal increase, it is a systemic shift.

00:04:09: We're effectively deciding as society that AI uptime Is more important than our climate targets.

00:04:14: So the strategy right now?

00:04:15: just panic and build?

00:04:17: For some players yeah definitely You are seeing literal gas turbines being strapped to semi trucks.

00:04:21: Just get temporary capacity online quickly.

00:04:24: That's wild.

00:04:25: But, and this is where the nuance comes in.

00:04:27: Not everyone agrees that cranking up the supply Is The Only Lever We Have.

00:04:31: Ellen Wu provided a really strong counter argument In The Sources This Week.

00:04:35: I really liked her take on it.

00:04:36: She argues That It Isn't Just A Supply Problem Its Actually A System Design Problem.

00:04:41: Precisely!

00:04:42: She's Stepping Back And Asking Why Do we Even Need This Much Power In First Place?

00:04:46: And she Outlines Six Specific Levers To Reduce Energy Use Before We Think About Pouring Concrete For Our New Power Plant.

00:04:53: Right.

00:04:54: Things like software efficiency, waste heat reuse and liquid cooling.

00:04:58: She called it treating energy as a first-class design constraint

00:05:01: Which is huge mindset shift for software engineers.

00:05:05: Usually developers just assume power's infinite in cheap.

00:05:07: You write the code and hit run

00:05:09: Write.

00:05:09: that cloud is magic

00:05:10: Exactly.

00:05:11: But Alan is saying Power Is The Actual Bottle Neck.

00:05:14: Now If you design around that constraint from day one... ...you might not actually need that gas

00:05:20: plant.

00:05:21: And that mindset shift seems to be bleeding directly into the hardware wars too.

00:05:25: Normally we talk about the chip war, you know, Nvidia versus AMD vs Intel is just a race for pure speed.

00:05:32: Who can do most floating point operations per second?

00:05:35: Right But Riccardo Moreno suggests that Metric is completely dead.

00:05:39: now

00:05:39: He really does!

00:05:40: he argues.

00:05:41: The new metric of the industry Is intelligence per megawatt

00:05:44: Which makes perfect sense when think it.

00:05:47: If u have fastest chip in world but requires power supply You literally cannot procure from the grid.

00:05:53: Or it runs so hot, and melts the server rack...

00:05:55: Then its completely useless!

00:05:57: Power is the primary constraint.

00:05:59: now The winner of the chip war won't be whoever ships most raw speed.

00:06:03: It'll be who ever delivers the most capability without blowing a breaker.

00:06:07: And this constraint isn't just changing hardware.

00:06:10: Its changing where we physically build these things.

00:06:13: Because if you can get power in places like Northern Virginia or Frankfurt anymore ...you have to go into what electrons are.

00:06:20: Right.

00:06:20: We saw two very different geographic visions emerging in the feed for this, there really was a tale of two climates.

00:06:26: Oh yeah!

00:06:26: The location debate.

00:06:28: so on one side you had Ahmed Samir El-Bermbali making a really strong case for the Mine region... ...the

00:06:34: Middle East and North Africa

00:06:36: shifting from being a global gas station to a green engine as he put it?

00:06:40: The logic there is pretty straightforward just go where the sun is

00:06:43: right.

00:06:44: countries like Saudi Arabia and the UAE have some.

00:06:46: The old twentieth century model was extract the oil, put it on a physical ship send to west and burn

00:06:56: there.

00:06:57: And

00:06:58: capture solar energy right there.

00:06:59: locally.

00:07:00: process massive AI workloads.

00:07:03: just export results.

00:07:04: you don't move heavy electrons but light weight bits

00:07:07: But then have complete flip side of that coin.

00:07:10: James Varga and Joel Cohen were pitching Scotland as ultimate green AI hub.

00:07:16: The cool option?

00:07:17: Literally, cool.

00:07:19: Their argument is that while MENA has the abundant solar power Scotland has low carbon electricity combined with cooling

00:07:26: Which a massive factor.

00:07:28: People forget that cooling can easily account for forty percent of a data center's total energy bill

00:07:34: Forty percent just to run AC.

00:07:36: Right

00:07:37: and Scotland have great windpower sure but crucially it has climate where you could use free air cooling almost entire year.

00:07:45: You don't need these energy-hungry industrial chillers, you basically just open the vents and let the ambient air do.

00:07:59: I

00:08:15: love this part of the deep dive.

00:08:17: It was probably most surprising contrast in research for me, it perfectly captures how split industry is right now.

00:08:24: so one corner we have Carl Rabe advocating for wooden data centers.

00:08:29: Now be honest when you first read that did immediately picture a little rustic log cabin with a bunch servers humming inside?

00:08:36: A little bit yeah.

00:08:37: Yeah I was picturing a highly secure barn.

00:08:40: But it's actually this very high-tech engineered mass timber.

00:08:45: And the point he is making isn't about the aesthetics at all, It's about embodied carbon

00:08:50: Which is crucial to break down.

00:08:52: We usually only talk about operational carbon.

00:08:54: You know The gas you burn To generate electricity To run servers Right but building the actual physical facility Pouring thousands of tons of concrete Smelting structural steel That creates a massive carbon depth Before we even turn single light on Concrete is just a nightmare for emissions.

00:09:10: And what's the opposite of that?

00:09:12: Exactly, wood physically stores carbon!

00:09:15: So Rabe is arguing if you build this structure out of engineered timber... The building itself acts as a carbon sink.

00:09:21: You are literally negative on carbon before even start operations.

00:09:25: Okay

00:09:26: so thats' the highly terrestrial earthy solution But then in other corner we have space

00:09:32: Project Suncatcher.

00:09:33: Yeah

00:09:34: Michael Rockwell and Annette Yultink were discussing this.

00:09:37: Data center is in orbit, it sounds like pure science fiction!

00:09:41: What's the actual business case for putting a server into space?

00:09:44: It all about availability of water... Think about it, in space you have twenty-four seven solar power.

00:09:50: There's no night cycle.

00:09:51: there are no clouds to block the panels and use absolutely zero water for cooling because just radiate heat directly out into vacuum of space.

00:10:00: So operationally its pristine?

00:10:02: Completely

00:10:02: pristine!

00:10:04: That is a massive catch.

00:10:05: The upfront carbon debt launching heavy lift rocket burns tremendous amount fuel that falls under scope.

00:10:11: three emissions.

00:10:12: It was huge trade off.

00:10:13: It Is...the debate essentially becomes Do you accept this massive carbon cost up front with the rocket launch just to secure zero emissions during your operations for the next decade?

00:10:25: It's a wild trade-off, do even have to consider.

00:10:27: But I think it really shows how desperate the industry is becoming... ...for solutions that don't rely on the overstressed terrestrial power grid!

00:10:34: It really

00:10:35: does.

00:10:35: but at the end of day whether its in a satellite and space In wooden structure or middle desert we eventually actually run code on these machines

00:10:45: perfectly brings us to our second theme for today.

00:10:48: Green software and cloud because optimizing the hardware is really only half of battle here,

00:10:53: And honestly?

00:10:54: The insights in this section were a bit disturbing.

00:10:56: they work.

00:10:57: We generally operate on it as confident assumption that we know exactly how much power Our software uses.

00:11:03: dr.

00:11:03: Nealadri Choudhury Really hammered home That old saying This week you can't manage what You don't measure

00:11:08: right.

00:11:09: But the insights shared by Renee Bostic suggest we might actually be flying completely blind.

00:11:14: She highlighted some recent research on Kepler.

00:11:17: for you listening who may not be deep in the weeds on this, Kepler is this standard open source tool used in Kubernetes environments to estimate energy usage.

00:11:28: It's basically the speedometer for your cloud infrastructure

00:11:32: and it turns out The speedometer has completely broken really

00:11:35: badly broken.

00:11:36: The research Rene shared found that Kepler can actually overestimate power consumption by up to fifteen times in some specific cases.

00:11:44: Fifteen times?

00:11:45: That is a massive margin of error!

00:11:48: It is, it apparently really struggles with things like CPU dynamic scaling and multi-tenant interference which is just the technical way saying its gets confused when multiple applications are sharing exact same physical chip at this time.

00:12:01: But why's such big deal?

00:12:03: isn't better accidentally Overestimate your carbon footprint than underestimate it.

00:12:08: Keep yourself honest

00:12:09: not necessarily if you're data is off by a factor of fifteen.

00:12:12: You literally cannot optimize anything

00:12:14: because you're chasing ghosts.

00:12:15: Exactly, you might spend weeks of expensive developer time Rewriting a piece of code to fix supposed power hog that actually running totally efficiently or conversely You might ignore a very real problem because the data is just too noisy to see it.

00:12:32: If the measurement is garbage, your management's gonna be garbage.

00:12:35: We desperately need a real source of truth.

00:12:37: we do but fortunately It wasn't all bad news.

00:12:40: on The Tooling Fund this week Chris Adams shouted out some really cool new solutions, things like Root & Branch and Green Coding Solutions.

00:12:48: These are trying to measure things differently right?

00:12:50: They aren't just guessing anymore!

00:12:51: Right they focus on direct energy reporting for websites right on the front end...they're not just modeling it based on broad server assumptions but they try get down into actual ground truth of what the end user's device is physically doing.

00:13:04: I also saw Theano D'Alo mention Carmen.

00:13:06: That's another open source measurement engine from Amadeus that just joined the green software foundation.

00:13:12: It really feels like the whole industry is suddenly scrambling to build better rulers because they realize old ones are completely warped.

00:13:19: And interesting thing, transparency has finally started trickling down directly into consumer too!

00:13:25: Robert Cuse launched something called GreenPT for iOS.

00:13:29: I saw this one... it an AI chat interface but explicitly shows you the bill

00:13:34: The carbon bill?

00:13:34: Yeah Every single time you type a prompt, it tells exactly the energy and CO₂ used for that specific query.

00:13:41: I think that psychological link is just vital right now because when you ask an AI to write a poem or generate a funny image of cat...it feels free!

00:13:51: It feels totally weightless.

00:13:52: Exactly…the cloud hides cost.

00:13:55: but if app stops and tells you hey that query costs five grams of carbon then fundamentally changes your behavior.

00:14:00: takes this completely invisible cost.

00:14:03: It

00:14:03: totally reminds me of the research Dominika Garsinska shared about clinical AI.

00:14:08: She noted that a single GPT-IV query uses about zero point four three watt

00:14:13: hours, which doesn't sound like much at all

00:14:16: until you actually do The math on the scale of it.

00:14:19: she pointed out That in a large clinical setting with say seven hundred million queries just to process routine medical notes that quickly adds up to the annual energy consumption Of thousands of residential households

00:14:33: Just for doing paperwork.

00:14:34: Exactly!

00:14:35: So optimization at the code level isn't just a nice PR thing to have anymore, it's operational survival And

00:14:41: its financial survival too.

00:14:43: Valeria Salas brought up this really important convergence happening between fine ops and green ops

00:14:49: Where money meets emissions

00:14:50: Right.

00:14:51: Fine Ops is the established practice of saving money on your cloud bills.

00:14:55: GreenOps is the newer practice of saving carbon.

00:14:58: Valerio points out that these are no longer two separate conversations happening in different departments

00:15:02: because the cloud providers basically charge you directly for your compute usage.

00:15:06: exactly if You have a good fainops team and they identify as zombie server, A server that's spinning drawing power but not actually serving any traffic And you turn it off instantly.

00:15:16: save cash and you save carbon.

00:15:18: The incentives are completely aligned now.

00:15:21: Erica Jeffers added to the point as well.

00:15:22: She noted that while we have all these measurement frameworks and dashboards now, The real shift has to be cultural.

00:15:29: it has to become embedded in the day-to-day operations of the engineering teams.

00:15:33: It

00:15:33: can't just be some glossy sustainability report That the marketing team files once a year?

00:15:50: Because when corporate culture and voluntary actions just aren't enough to move the needle, The regulator starts stepping in.

00:15:56: And

00:15:56: they are definitely bringing laws with them.

00:15:57: now Let's

00:15:58: talk strategy policy and regulation.

00:16:01: To kick this off Karen van der Zanden introduced a concept that I genuinely think is going to define geopolitics for the next decade.

00:16:09: She said no digital sovereignty without energy sovereignty.

00:16:13: It's a really profound point, especially when you look at Europe right now.

00:16:16: Digital sovereignty is this big buzzword that usually just means hey we want our citizens data physically stored on servers located inside our own borders!

00:16:27: We don't want it sitting in a data center.

00:16:30: But Karen is pointing out that localizing the data isn't enough anymore.

00:16:34: Right, because if your sovereign data is sitting securely in Paris but the electricity required to keep those servers running relies entirely on imported gas or proprietary technology you don't control...you aren't actually sovereign!

00:16:47: You're

00:16:47: still totally dependent on an external supply chain?

00:16:50: Exactly so.

00:16:51: for Europe to truly be digitally independent it has to physically link its digital infrastructure directly.

00:16:59: massive offshore wind farms in the North

00:17:01: Sea.

00:17:02: Energy policy basically is tech policy.

00:17:04: now, you really can't separate the two anymore?

00:17:06: You can't.

00:17:07: but even if a country secures that sovereign energy we still have to ask the harder question what does all this AI actually being used for?

00:17:14: Modifyod brought up a philosophical point I thought was just incredibly critical

00:17:18: The impact first mindset.

00:17:20: Yes

00:17:21: He argues that we shouldn't just be talking about green AI, which just focuses on raw efficiency.

00:17:25: We should be shifting the conversation entirely to impact first

00:17:29: AI.".

00:17:30: It's The Ultimate.

00:17:30: So What's Factor.

00:17:32: Majed asks does the AI we are powering actually solve a real human problem like discovering a new antibiotic?

00:17:40: Or does it generate endless slightly different selfies?

00:17:43: And if it's just generating selfies, making the underlying code ten percent more green doesn't suddenly make it worth the massive energy expenditure.

00:17:50: Right you can hook a slot machine up to a solar panel.

00:17:53: It technically runs on Green Energy

00:17:55: But its still just a slot-machine.

00:17:56: Its fundamental question of societal utility versus environmental cost

00:18:01: And governments are definitely starting to look closely at that utility and waste aspect.

00:18:05: Dr.

00:18:05: Ahmad Sabirin Arshad highlighted a really landmark new law over in France specifically regarding anti-obsolescence.

00:18:13: France

00:18:13: is really leading the charge globally on this, it's now completely illegal there.

00:18:18: design products intentionally fragile or designed break after certain period.

00:18:23: they've made repairability index legally mandatory for electronics.

00:18:29: This hits the hardware side of the ICT sector incredibly hard.

00:18:32: You can't just churn out these glued together disposable devices

00:18:37: anymore.".

00:18:38: It legally

00:18:39: forces manufacturers to think about the full end-to-end life cycle.

00:18:43: if i build this new smartphone Can a user actually open it up and replace the battery?

00:18:47: If The answer is no,I literally cannot sell in that market.

00:18:51: That fundamentally changes the initial design process.

00:18:53: However, while France is actively tightening the screws on hardware Anita Schittler pointed out a massive, massive loophole in how companies currently report their overall impact.

00:19:04: It's a critical flaw in the GHG protocol.

00:19:06: Oh this is the classic shell game of corporate carbon accounting?

00:19:10: Walk us through this one because it really seems like a place where a lot of companies are hiding a lot dirt from their balance sheets.

00:19:15: Okay so In Carbon Accounting you have Scope II emissions.

00:19:19: That represents electricity a company buys directly.

00:19:22: So if a major bank runs its own legacy data center right there in the basement of it's headquarters, The power bill for that server room is their scope to omission.

00:19:32: Got It!

00:19:32: Direct Power Purchase.

00:19:33: But If That Same Bank Decides To Modernize And Moves All Of Those Workloads Off-Premise and Into The Cloud Say over to AWS or Azure, those emissions legally move from their scope two ledger over to their scope three ledger.

00:19:48: Scope Three being indirect supply chain

00:19:51: emissions.".

00:19:51: So the bank can put out a press release saying look our direct scope-two emissions just dropped to zero.

00:19:56: we achieved

00:19:58: Exactly.

00:19:59: It looks absolutely fantastic on their annual ESG report, but the physical real-world emissions haven't changed by a single gram?

00:20:06: They've just been contractually shifted over to Amazon or Microsoft's ledger!

00:20:11: It is and it allows these huge enterprises to publicly claim a climate victory without actually reducing overall global load

00:20:20: growing frustration with this kind of corporate maneuvering is really starting to boil over in the professional community.

00:20:26: We saw a very fiery, direct post from Mark Butcher touching on it.

00:20:32: Yeah, he really didn't mince any words.

00:20:34: He is actively calling for US tech companies to explicitly and publicly state their stance on climate change right now.

00:20:42: And this isn't direct response to the recent political shifts in the?

00:20:45: u.s.

00:20:45: Right?

00:20:46: Correct

00:20:46: We're seeing a significant rollback of federal environmental regulations and Butcher's arguing that in this context Corporate silence is basically complicity.

00:20:56: If the government removes the legal guardrails, are these tech giants going to voluntarily keep their historical climate commitments or they just gonna take the deregulation as a free pass to pollute because it's cheaper?

00:21:09: He's demanding moral clarity in a market that usually prefers to operate on gray areas

00:21:14: and That has very dangerous demand for public company.

00:21:16: but he argues its absolutely necessary

00:21:19: right now.

00:21:20: Looking at all these posts together, it really feels like all of the separate threads are finally tying into one big knot.

00:21:26: Well just look at her from a high level!

00:21:28: We have the physical reality these massive new gas plants and shifting chip wars pushing right up against The hard physical limits of the power grid.

00:21:35: then we had the software Reality where were actively struggling to even measure.

00:21:40: The mess, we've made with our code.

00:21:42: And finally we have the regulatory reality governments an activist trying To impose some kind of order on all this chaos through New laws in demands for sovereignty.

00:21:51: it is a complete collision course.

00:21:53: What we're witnessing are the very real growing pains of a digital civilization that is just now realizing it has strict physical limits.

00:22:01: So as we wrap up this deep dive into week six and seven, what does single main take away for you?

00:22:06: For me its'that the era invisible compute.

00:22:09: officially over For twenty years, we've all treated the internet and cloud as this magical weightless thing.

00:22:15: But whether it's a two hundred megawatt gas plant being built in your backyard... ...a strict law in France preventing you from buying disposable phone or an iOS app telling exactly how much CO² your AI pump just cost.

00:22:27: The physical heavy costs of our digital life is finally becoming visible

00:22:31: And tools to handle are getting better.

00:22:34: We're spotting those fifteen X reporting errors.

00:22:38: We are actively debating the merits of wooden data centers versus space-based solar ones.

00:22:44: The sophistication in this conversation is definitely increasing!

00:22:47: It IS, but I think that ultimate question remains the one Majd Fayyad asked... Even if we fix the tools… Are we actually using a massive unprecedented energy expenditure for things which truly add value to humanity?

00:23:00: Right Just because we can build a massive data center and just becausely write slightly more efficient code on it should?

00:23:07: Here is a final provocative thought for you to chew on before our next deep dive.

00:23:11: Building that idea of visibility, what happens when energy becomes so constrained?

00:23:16: we see the introduction of literal carbon compute tax?

00:23:19: Imagine if every packet of data you requested carried visible fluctuating carbon price tag... ...that YOU-the end user had to pay directly based upon the grid's carbon intensity at that exact second!

00:23:29: If downloading a movie cost more than at night then it noon when sun is shining.

00:23:32: That would fundamentally change our relationship with technology overnight.

00:23:37: It really

00:23:37: would, it puts the accountability right in users' hands.

00:23:40: If you enjoyed this episode new episodes drop every two weeks.

00:23:44: Also check out other editions on cloud digital products and services artificial intelligence and ICT and tech insights health tech defense tech.

00:23:53: Thanks for tuning-in today

00:23:54: And don't forget to subscribe.

00:23:56: see you next deep dive.