Best of LinkedIn: Sustainability & Green ICT CW 12/ 13

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-II and XIII.

00:00:09: Frennes supports ICT enterprises in the form of delivering precise ICT market and pricing intelligence that analyst subscriptions and existing databases cannot provide.

00:00:18: you can find more info in the description

00:00:21: so welcome to The Deep Dive everyone.

00:00:22: today we are unpacking some sustainability in green ICT trends seen across LinkedIn over these past couple of weeks.

00:00:31: Yeah, and the mission here is to really cut through the fluff.

00:00:34: we are looking at what it actually takes To balance this.

00:00:37: you know explosive digital growth We're seeing with the physical limits of our planet because while we treat The cloud like It's just this magic invisible thing

00:00:45: right?

00:00:45: but its Really?

00:00:45: Just massive Physical warehouses burning Through unbelievable amounts Of power And water Which leads us Right into Our first Theme Today data centers, cloud and resource efficient infrastructure.

00:00:58: Exactly!

00:00:58: And Ryan Shulin and Boris Gamazichikov shared some pretty eye-opening insights from recent data center conferences.

00:01:06: Basically the only thing industry cares about right now is speed to power.

00:01:09: Oh a hundred percent.

00:01:10: because well Gamazitchikov points out that AI companies expect something like a fifty six to six hundred twenty five times multiple on their energy costs.

00:01:18: I mean A single GPU can generate over one hundred bucks in value Every single hour it runs.

00:01:24: Wait,

00:01:24: really?

00:01:24: A hundred dollars an hour

00:01:25: per trip!

00:01:26: Yeah, per chip.

00:01:27: So what's fascinating here is that the energy cost isn't a bottleneck anymore.

00:01:30: It's practically a rounding error.

00:01:32: They just want to facility online as fast and physically possible

00:01:36: Right.

00:01:36: But this where I get interesting.

00:01:38: All of that.

00:01:39: localized power creates massive physical footprint.

00:01:42: Nathaniel Barola recently highlighted research on the data heat island effect.

00:01:46: Oh

00:01:46: yeah Where HVAC systems exhaust all that process or heat locally

00:01:50: Exactly.

00:01:51: These AI data centers are actually raising local land surface temperatures by an average of two degrees Celsius, and that warming effect can stretch up to ten kilometers away from the

00:02:00: facility.".

00:02:01: That is wild...ten kilometers!

00:02:03: And of course dealing with all this extreme heat brings us straight into a water crisis.

00:02:08: Ali Ahmad noted datacenters globally are consuming between five hundred and seven-hundred billion liters annually

00:02:15: Because they rely so heavily on those massive evaporative cooling towers,

00:02:19: right?

00:02:19: Right.

00:02:19: Exactly.

00:02:20: But we are seeing some counter moves.

00:02:22: Alexis Bateman shared that AWS managed to hit a water usage effectiveness or WE of point one five liters per kilowatt hour.

00:02:30: Oh wow!

00:02:31: That's incredibly efficient

00:02:32: It really is.

00:02:33: And Vrushali God reported that Google Is actively replenishing sixty four percent Of its freshwater use globally right now.

00:02:40: Plus Adam Elman posted something Really cool about how They're interacting with the power grid too.

00:02:45: Google is using one gigawatt of data center capacity for grid demand response.

00:02:49: Wait, so they are acting like a shock absorber?

00:02:51: For the power grid rather than just draining it

00:02:53: exactly They can literally shift or pause heavy workloads when the local grid is stressed.

00:02:59: But you know if data centers are The engine running hot ai's definitely the heavy cargo forcing them to overwork.

00:03:07: Which brings us to theme two AI sustainability governance and the Jevons Paradox.

00:03:13: Yeah, Eli Tofty Anderson shared a really striking observation about this.

00:03:17: Google actually made each individual AI query thirty three times more energy efficient.

00:03:23: Okay but wait if they made it thirty-three times more efficient why did their overall carbon footprint still grow by forty eight percent?

00:03:31: Right that is The Jevon's Paradox in action.

00:03:33: basically making the technology cheaper and more efficient just drove up.

00:03:37: So if making AI more efficient just makes us use it way more, aren't we kind of accelerating towards the cliff here?

00:03:43: Basically yes.

00:03:44: And Mary Stevens had a really stark warning about this.

00:03:47: she predicts that companies might start commingling energy guzzling generative AI with you know highly-efficient traditional AI to hide their massive emissions.

00:03:55: Oh like bundling those toxic mortgages during the two thousand eight financial crisis.

00:03:59: so literally Just Like That they just slap a generic AI label on The Whole Package To Mask The Environmental Cost.

00:04:07: But Dr.

00:04:08: Sasha Lucioni points out that AI is really a spectrum, she actually compared it to onions and ogres from Shrek.

00:04:14: Ah!

00:04:15: It has layers?

00:04:15: Right.

00:04:16: so binary AI free labels aren't enough.

00:04:19: we need structural solutions like what Naveen Balani shared regarding Google's Gemma Four.

00:04:24: Oh those are the smaller effective parameter models.

00:04:27: right Exactly

00:04:28: they run on device creating these hybrid architectures that seriously reduce the need for massive cloud compute in first place.

00:04:35: Which is a perfect transition into theme three, green software engineering and measurement tools.

00:04:41: because whether it's AI or just traditional Cloud infrastructure The ultimate control lever is the software architecture itself.

00:04:48: For sure Green Software really moving from theory to daily engineering practice.

00:04:52: now.

00:04:53: Yeah, Arwell Owen from Velastar shared a crazy metric about this.

00:04:57: He helped the client evaluate their proposed SAP S-IV ANA infrastructure and they managed to cut their proposed emissions by an incredible ninety percent.

00:05:04: Wait...a ninety percent reduction?

00:05:07: Just from the software architecture.

00:05:08: Yeah, they took it from seven point two tons down to point three tons just through smarter upfront infrastructure decisions and ruthless right sizing.

00:05:17: that

00:05:17: is massive.

00:05:18: yeah And you know developers are finally getting the visibility They need to actually execute on this.

00:05:23: Chris Adams showcased open code Which is an open source AI tool?

00:05:28: That displays directly measured energy consumption live while you code.

00:05:33: Oh, that's super practical.

00:05:34: It is!

00:05:35: An ego-emise check on shared a new code carbon extension specifically for VS Code That tracks live emissions right there in the editor.

00:05:42: Okay let's unpack this for a second.

00:05:44: Measuring your emissions in real time Is well.

00:05:47: it's like staring at the dashboard In your car tells you You're wasting gas but still have to actually change how you drive.

00:05:52: Yeah and thats exactly The point.

00:05:54: Louise Cruz And Jeremy Mcconnell argue Green software has To move beyond just basic energy measurement Toward lasting design & cultural Change.

00:06:02: Right And Philip Christing shared a super practical three-step method to optimize existing software.

00:06:08: He basically proves that doing this immediately cuts both costs and emissions without causing a developer rebellion, but

00:06:14: you know You can't change the engineering culture if the executives don't have accurate data To justify the investment.

00:06:22: Which brings us to theme four sustainability data reporting and decision intelligence.

00:06:27: Oh, this is where things get really messy.

00:06:30: Shrine of Us Patil warns that spend-based carbon data is just deeply flawed

00:06:35: because inflation artificially inflates your carbon footprint right?

00:06:38: Even if you're actual server usage stays flat

00:06:40: exactly!

00:06:41: If your cloud vendor raises prices by twenty percent Your foot print suddenly looks twenty percent worse.

00:06:46: patil says companies have to switch to activity based engineering.

00:06:49: truth

00:06:50: an agritree show points out This major aggregation problem.

00:06:53: with ESG scores too.

00:06:55: Rating agencies compress totally different metrics, like carbon emissions and supply chain ethics into one single number.

00:07:01: Which completely masks the real underlying risks?

00:07:04: That sounds like an absolute nightmare for finance teams!

00:07:07: It is.

00:07:08: Tatjana Munch pointed out that Finance & Sustainability Teams absolutely must align using CFO grade data to drive any real business value.

00:07:17: So how are they actually solving this data gap?

00:07:19: Well, Professor Dr.

00:07:20: Andreas Brunert and Jamie O'Cowen explain how the ISO software Carbon Intensity Standard actually provides the auditable numbers needed for European CSRD disclosures.

00:07:31: Oh so there is a real standardized translation layer now.

00:07:34: Yeah

00:07:35: And Peter DeWenz shared great success story using Microsoft's Sustainability Manager at Azilis to hardwire this reporting directly into their digital landscape.

00:07:44: Plus, Nolan Goddard notes that AI is now automating a lot of this reporting grunt work so teams can shift from fighting with spreadsheets to executing actual continuous decarbonization.

00:07:54: Which is the ultimate goal?

00:07:55: but you know we've optimized our software and cleaned up the reporting data But eventually physical servers running all these stuff have to die.

00:08:01: Where do they?

00:08:02: Right!

00:08:02: Theme five Circular IT & Lifecycle Value Creation.

00:08:06: Yeah Robert Little revealed some staggering internal numbers.

00:08:09: Google's circular economy approach harvested eight point eight million individual server components in

00:08:18: twenty-twenty four alone,

00:08:20: and they completely wiped and reused three million hard drives treating decommissioned servers as a high yield repository of commodities.

00:08:28: Wow, and you don't even have to be a trillion dollar hyperscaler do this.

00:08:32: Remco van Roijen and Sophie Rybroek from the Circular IT Group helped ABN AMRO run a digital cleanup week.

00:08:40: Oh where employees brought in their old personal devices?

00:08:42: Yeah exactly!

00:08:43: To securely wipe and refurbish those devices for charitable pre-use.

00:08:47: it's brilliant culture builder

00:08:49: It really is.

00:08:49: And if we connect us with the bigger picture Anita Shetler perfectly summarizes over arching theme of these two weeks.

00:08:56: Which

00:08:56: what exactly

00:08:57: that?

00:08:57: reaching sustainability skeptics requires framing green goals around pure cost reduction, risk management and system modernization.

00:09:05: Ah!

00:09:05: So when sustainability equals operational efficiency everyone wins?

00:09:08: Exactly it stops being just a PR exercise.

00:09:11: If you enjoyed this episode new episodes drop every two weeks.

00:09:14: also check out our other editions on cloud digital products & services artificial intelligence an ICTN Tech Insights health tech defense tech.

00:09:22: Thank You so much for joining us On This Deep Dive.

00:09:25: make sure to hit subscribe.

00:09:27: And before you go, I want to leave you with one final thought to mull over.

00:09:30: Sparks by a post from Wilco Bergerath.

00:09:33: We obsess over these mega data centers and massive cloud models but could the future actually be hyper-local?

00:09:40: Developers are already building personal AI powered coding environments running entirely on open models in their own rooftop solar panels.

00:09:47: So ask yourself what if the most sustainable tech stack of the future isn't in the Cloud at all But sitting right on your desk?