Best of LinkedIn: ICT & Tech Insights CW 23/ 24

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about ICT and tech insights from CW-Twenty Three and Twenty Four.

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

00:00:20: you can find more info

00:00:22: Yeah, and we have a really fantastic deep dive lined up for you today.

00:00:26: We absolutely do!

00:00:27: I mean if you are listening to this... You're probably dealing with some level of network architecture or enterprise data right now?

00:00:33: Right And we really want to set the stage here.

00:00:36: We are cutting through all the surface-level noise today To bring you the top ICT in tech trends that we've curated directly from LinkedIn across calendar weeks.

00:00:44: twenty three and twenty four of twenty twenty six

00:00:46: Exactly because there's just so much hype out there You know?

00:00:49: But we want to look at the raw mechanics of where the industry is actually heading.

00:00:52: Yeah, The real stuff!

00:00:54: So here's the roadmap for today.

00:00:55: We're going break this down into four main clusters.

00:00:58: First were looking at AI agents and actual data they run on

00:01:01: Which a huge bottleneck right now.

00:01:03: Oh absolutely They are moving in physical reality of AI.

00:01:06: so were talking massive infrastructure And edge compute realities.

00:01:10: Then when you push everything to the Edge.

00:01:13: You have talk about cyber security In this new hyper fast era.

00:01:16: Terrifying stuff there.

00:01:19: And finally, we will wrap up with quantum readiness and why digital sovereignty has suddenly become a massive boardroom priority.

00:01:28: I can't

00:01:28: wait.

00:01:29: Let's just jump right into that.

00:01:30: first thing because like we said everyone on your feed is talking about AI.

00:01:34: for

00:01:34: sure It's inescapable

00:01:36: but Sandra Veller had this post That really stood out.

00:01:39: she pointed out this fundamental disconnect.

00:01:42: She basically said the most enterprises out there Right now they don't actually have an ai problem.

00:01:47: Wait Really?

00:01:48: where they have them.

00:01:49: They have a clean data and undocumented process

00:01:51: problem.

00:01:52: oh wow yeah that makes total sense right?

00:01:54: it's like the ultimate reality check.

00:01:55: you see these companies, they're buying up this massive enterprise licenses for advanced language models but their internal data is just lots of disaster.

00:02:04: so

00:02:04: the information is totally siloed permissions are probably all over

00:02:08: Exactly, and the processes are entirely undocumented.

00:02:12: So if you just slap a highly advanced algorithmic engine on top of that mess...

00:02:16: It doesn't actually fix anything!

00:02:18: No it just executes your operational chaos but at an unprecedented scale?

00:02:23: Its so funny to say this because I mean giving in autonomous AI access to un-documented corporate data.

00:02:30: its not even like adding jet engines into a horse drawn buggy.

00:02:33: Yeah..its putting high speed automated drill In pitch black room.

00:02:37: Oh man That is a great analogy.

00:02:39: You know, he's going to find a structural pipe in the wall and it's gonna burst at way faster than any human ever could.

00:02:45: just totally automated mistakes.

00:02:46: Yeah you're just making faster mistakes.

00:02:48: And that actually sets up on much larger issue that Dr.

00:02:51: Inshawn May-John brought up In his post.

00:02:53: Okay what was this take?

00:02:54: Well He's looking at evolution of these tools.

00:02:57: We are crossing critical threshold right now Where AI moving from being helpful assistant you know, like a chat bot drafting an email.

00:03:06: Right the passive stuff.

00:03:07: yeah passive but now it's becoming an autonomous agent meaning

00:03:10: it actually executes actions completely on its own

00:03:13: exactly.

00:03:14: so we're talking about agents that can autonomously move corporate funds or grant and revoke system access.

00:03:21: they can even push unreviewed code directly into production.

00:03:24: That is terrifying

00:03:25: It Is!

00:03:26: And Dr Mayjohn asks this critical question When that Agent inevitably makes a mistake who actually answers for

00:03:34: it.

00:03:34: Oh, the liability aspect?

00:03:35: Yeah!

00:03:36: Like if an autonomous system incorrectly revokes access to your entire customer database during peak hours, whose name goes on that incident report?

00:03:46: I mean i would assume it's the IT director right or whoever actually deployed the agent.

00:03:50: You'd think so but Meijan is arguing that we have to define these accountability operating models before the fallout happens not during the post mortem when everyone is pointing fingers.

00:03:58: But let me push back on this for a second because if the underlying data structure of autonomous action is this high, shouldn't companies just hit the brakes?

00:04:10: You'd think that would be.

00:04:18: The

00:04:26: competition is moving too fast.

00:04:28: Exactly!

00:04:29: You can't just pause your entire deployment because... ...your data house isn't perfectly clean, Your competitors definitely aren't waiting around for you.

00:04:37: So what's the solution then?

00:04:38: If you can't stop?

00:04:40: But going too fast is dangerous.

00:04:42: Shabel says it comes down to boardroom level strategic decisions about data partitioning.

00:04:47: Okay, so deciding what the AI has actually allowed to see?

00:04:50: Yeah You have to draw hard borders.

00:04:53: Leadership has to define exactly What proprietary data Is safe to feed into the models To get that efficiency boost and what core IP absolutely Has to stay locked in The vault protect company.

00:05:03: That makes a lot of sense.

00:05:04: And he also emphasized that you need genuine buy-in from the humans who are actually using these systems, because if your engineering team or your finance team doesn't trust the AI agents or they feel threatened by them... ...the whole deployment just stalls out anyway!

00:05:17: It always comes back to people, doesn't it?

00:05:19: It

00:05:19: really does.

00:05:20: But we're talking about these agents running millions of automated tasks.

00:05:24: but that intelligence isn't floating in the ether.

00:05:27: Oh no definitely not.

00:05:28: It is boiling water into data centers somewhere.

00:05:32: And this is a great transition to our second theme, the physical infrastructure.

00:05:36: I love this topic

00:05:37: because Henri Nakarundi had this fascinating point that totally shifts the economic perspective here.

00:05:43: What did he say?

00:05:44: He argues as these AI models become commodities you know digital intelligence just becomes abundant and cheap.

00:05:53: The real trillion dollar value shifting down the stack

00:05:56: Down to the physical hardware

00:05:58: Exactly.

00:05:59: It's moving to everything that intelligence physically depends on, so we're talking energy grids specialized chips and the actual concrete in steel of data centers.

00:06:07: And engineering data completely backs up on Rhys thesis.

00:06:12: Ongtun shared some hard numbers.

00:06:14: what modern infrastructure actually requires.

00:06:17: Honestly this skill is just difficult even comprehend.

00:06:20: Lay it on me How big are you talking?

00:06:23: Modern high-density AI server racks are now pushing power density past one hundred to three hundred kilowatts per single rack.

00:06:31: Okay, let's just pause on that number for the IT professionals listening.

00:06:35: A traditional enterprise server rack runs at what like five to fifteen kilowattes?

00:06:38: Often

00:06:39: less than ten.

00:06:40: Wow so three hundred.

00:06:42: yeah

00:06:42: You are basically looking at condensing the power draw of an entire small neighborhood into the physical footprint a kitchen refrigerator.

00:06:49: That is insane!

00:06:50: The heat coming off that must be unbelievable.

00:06:53: It IS, the physical heat generated by three hundred kilowatts of concentrated compute means traditional air cooling simply hits a physical wall.

00:07:01: you

00:07:01: literally can't blow enough cold air fast enough

00:07:03: No...you cant' blow chilled air through a chassis to stop the silicon from just completely melting down.

00:07:10: TENS points out that cooling is no longer just a supporting facility system.

00:07:14: It's the main event!

00:07:15: Exactly, thermal management is dictating the entire architecture of the data center.

00:07:19: now you have to engineer direct-to-chip liquid cooling or even those full immersion systems right from day one which

00:07:26: means retrofitting old data centers for AI is practically impossible

00:07:30: pretty much.

00:07:31: yeah

00:07:31: You basically have to build them from scratch.

00:07:33: and because these intense constraints on the ground you know, trying to find land.

00:07:39: Finding the water for cooling getting enough power from the grid some engineers are literally looking up.

00:07:45: Oh!

00:07:45: You mean orbital compute?

00:07:46: Yes Yeah

00:07:47: Ken Kwong posted about this.

00:07:49: When people hear Space Data Center I think they immediately picture this giant sci-fi floating Pentagon in space

00:07:56: Right Like The Death Star but For Servers

00:07:58: Exactly.

00:07:59: But Kwong explains that future of Space Compute is actually these distributed starling style satellite racks.

00:08:07: It sounds so exotic, but honestly the mechanics make a lot of sense when you dig into it

00:08:12: Really?

00:08:12: I mean the launch costs is own

00:08:14: Sure!

00:08:14: Launch costs are a factor.

00:08:16: But by placing standardized modular compute racks in low earth orbit You bypass all terrestrial real estate issues and The cooling challenge becomes totally different than vacuum space.

00:08:27: Oh because its freezing

00:08:28: Well, depending on how you manage the solar radiation.

00:08:31: Yeah ambient cooling is a completely different engineering game up there but more importantly it takes edge computing to its absolute extreme limit

00:08:38: just bypassing The entire terrestrial network.

00:08:41: yeah It minimizes latency for global remote operations anywhere On

00:08:46: the planet.

00:08:47: I don't know.

00:08:47: i'm still not entirely convinced.

00:08:48: the launch costs justify quite yet.

00:08:51: But the modular approach kwang is talking about isn't Just For Space that's true.

00:08:55: Sudipta Padacharya and Ibrahim Kisioglu kind of synthesized a similar trend happening right here on the ground.

00:09:02: They're arguing that the era of The massive centralized monolithic hyperscale data center is actually ending,

00:09:10: And it has to end simply due to material and grid limits.

00:09:13: i mean you just cannot find enough concentrated gigawatts Of power On the terrestrial grid To keep building these massive brute force campuses indefinitely.

00:09:22: so what's the alternative?

00:09:23: breaking them apart.

00:09:24: Yeah,

00:09:24: Badacharia and Keseoglu see the physical infrastructure evolving into a highly integrated distributed ecosystem.

00:09:31: you build specialized tiers.

00:09:33: Okay what does that look like?

00:09:34: Well, you'll still have hyperscale AI campuses but they will be dedicated purely to heavy asynchronous model training and there'll be built wherever raw power like a dedicated nuclear plant or geothermal is available.

00:09:48: Okay so way out in the middle of nowhere

00:09:50: right?

00:09:50: But those training hubs will be deeply networked to a massive web of smaller modular edge facilities users, factories and autonomous vehicles to handle all the real-time inference.

00:10:06: Ah!

00:10:06: So spreading out the real time load.

00:10:09: but hold on if you distribute your physical infrastructure To thousands of these extreme edge locations.

00:10:15: And earlier we were talking about handing the operational keys over to Autonomous AI agents.

00:10:19: Yep

00:10:20: I know where you're going with this.

00:10:21: You are blowing your attack surface completely wide open Absolutely

00:10:24: wide open

00:10:25: Which brings us perfectly into our third cluster The terrifying reality of cybersecurity in this hyper-fast era.

00:10:31: And speed really is the fundamental change here, Matthew Rosenquist shared this wild update regarding frontier AI models specifically pointing to a model called Mythos.

00:10:40: Okay what did he find?

00:10:41: He meets that AI is collapsing the vulnerability patch window.

00:10:45: Historically, you know when a zero day was discovered security teams had days maybe even weeks to test and deploy a patch before adversaries could really weaponize it at scale.

00:10:54: Right there's human bottleneck

00:10:56: Exactly.

00:10:57: but now that windows shrinking down literally minutes

00:11:01: Minutes?

00:11:02: Imagine piece of malware breaches your firewall and then dynamically rewrites its own exploit code in the three minutes it takes your security team to go grab a cup of coffee.

00:11:12: It's terrifying because the AI can ingest the vulnerability disclosure, write the exploit code and deploy across network faster than human analysts.

00:11:21: even triage very first alert

00:11:23: The adversary is fully automated.

00:11:24: they don't sleep

00:11:25: Right They process code at machine speed.

00:11:28: But we also have balance this high-tech anxiety with grounded reality check.

00:11:32: Okay please give me some good news.

00:11:34: Well, Jane Franklin pointed out this deeply frustrating truth about enterprise security.

00:11:38: Despite all the industry panic over nation-state threats and AI zero days The vast majority of debilitating breaches still happen for really dumb reasons

00:11:48: Because organizations just ignore the absolute basics.

00:11:51: Yes She says it's usually weak passwords or legacy systems that haven't been patched in a year.

00:11:57: Or you know access rights That hasn't been reviewed since an employee changed departments like three years ago.

00:12:02: So we're essentially so distracted looking up at the sky for these Terminator AI threats that were just forgetting to lock the digital front door.

00:12:12: Exactly, and the A.I doesn't need to invent a brilliant new exploit if it can turn the doorknob on an administrative portal.

00:12:20: you left exposed to public internet one million times per second

00:12:24: which means your defense strategy has to totally shift.

00:12:26: You cant build higher walls but assume they are already breached.

00:12:30: Marcel Velika's insights completely confirm this.

00:12:33: He emphasizes that cyber resilience matters significantly more than pure prevention.

00:12:37: right now, you just simply cannot prevent every single AI speed attack.

00:12:42: So it's about bouncing back.

00:12:43: Yeah,

00:12:43: Valica argues that the organizations that actually survive these incidents are the ones who practice their response routines relentlessly like before the crisis ever hits.

00:12:52: They engineer systems to isolate compromised nodes automatically.

00:12:56: Right!

00:12:56: It is all about minimizing blast radius and recovery time instead of pretending you can block everything.

00:13:02: Now if your listening this a new manager regional IT team You might be thinking You know, I don't have the budget of a hyperscaler to fight these automated threats.

00:13:11: Which is very fair point!

00:13:12: It's true but Mohamed Sayed posted some highly actionable advice for exactly this scenario particularly targeting tech professionals in regions like APAC or Middle East.

00:13:23: I saw that post.

00:13:24: his core argument was so smart.

00:13:26: he says you do not actually need massive Silicon Valley budget to capitalize on security gap?

00:13:31: No no at all because his roadmap leverages the exact same tools.

00:13:35: The adversaries are using.

00:13:37: he lays out this targeted six to eight week sequence, right?

00:13:40: He advises professionals to start by using free cloud tiers and open source AI tools to build their own defensive agents like get hands on with the mechanics of how these models interact with network logs.

00:13:51: but they're real.

00:13:52: kicker in his strategy isn't just learning the tech.

00:13:54: it's mapping those AI capabilities directly to local governance frameworks

00:13:59: And that is the crucial differentiator.

00:14:01: He suggests taking those automated defensive routines and mapping them to frameworks like NIST or ISO, specifically tailoring them to local regulated industries in your specific region.

00:14:12: So if you can automate compliance auditing for a regional bank in the Middle East using free open source AI?

00:14:18: You suddenly become the Regional AI Security go-to person!

00:14:21: Exactly because technology itself is global.

00:14:24: but... Compliance & Liability?

00:14:26: That is always hyper-local.

00:14:28: And mastering that intersection, it's how you make yourself completely invaluable without needing a ten million dollar security budget?

00:14:34: It's brilliant!

00:14:36: So let's shift gears again.

00:14:37: we talked about AI speeding up current threats but lets talk about the technology that threatens to

00:14:42: just

00:14:42: break the locks entirely.

00:14:44: Ah!

00:14:45: Quantum!

00:14:46: Yes our final cluster quantum readiness and digital sovereignty.

00:14:50: For a long time quantum computing felt like just a purely academic exercise.

00:14:54: Yeah, cool physics but not a real network architect problem

00:14:57: exactly.

00:14:58: But Dr.

00:14:59: Aaron C camp highlighted this massive shift from the research lab to hard industrial reality.

00:15:06: the numbers he shared really indicate a tipping point.

00:15:09: The US government and IBM have committed over twelve billion dollars specifically to quantum manufacturing, and fault-tolerant

00:15:16: computing."

00:15:17: And they are targeting at twenty twenty nine timeline

00:15:20: right?

00:15:20: Yeah!

00:15:20: Twenty twenty nine it's right around the corner.

00:15:22: Wait real quick.

00:15:22: for those who might not know what does fault tolerant actually mean in this context?

00:15:26: because quantum states are notoriously fragile.

00:15:29: They're incredibly unstable.

00:15:31: like the slightest temperature change or a little bit of electromagnetic interference causes the quantum bits, quibits to lose their state

00:15:39: which introduces massive calculation errors.

00:15:42: right so fault-tolerant computing means engineering systems that can automatically correct those errors in real time Which allows machine actually run complex algorithms reliably.

00:15:51: That makes sense.

00:15:53: But the most critical detail in Kemp's post isn't even the tech itself.

00:15:57: It is a financial structure of.

00:16:11: When the government takes an equity position, they are signaling that the foundational cryptographic standards—the stuff that protects everything from your banking to military communications —are officially on a ticking clock.

00:16:23: And reality fundamentally changes corporate strategy.

00:16:26: Alison King and Lawrence Belton both posted about this emphasizing that quantum readiness is no longer just problem for the CSO.

00:16:34: No!

00:16:34: It has escalated.

00:16:35: it's now a CFOs' Problem

00:16:36: Because migrating away our current encryption to post-quantum cryptography, requires serious budget allocation.

00:16:44: It is not just a simple software update you run over the weekend.

00:16:47: Belton and King point out that you have to fund deep asset discovery first.

00:16:52: What does Deep Asset Discovery actually entail though?

00:16:54: Are they running a network scan?

00:16:56: No it's far more complex than a scan.

00:16:58: We are talking about hunting down cryptographic libraries that are hard coded into legacy applications from like fifteen years ago.

00:17:06: Oh wow The stuff everyone forgot

00:17:08: Exactly.

00:17:09: You have to find every single instance where vulnerable encryption is used across your entire enterprise architecture, and then you have fund the modernization of those systems without breaking current production.

00:17:22: Good luck with that.

00:17:23: Right If CFO isn't involved in funding that discovery phase right now The organization will be completely flat-footed when the current standards are officially deprecated.

00:17:34: It really is a massive infrastructure overhaul, but it's just disguised as the security update.

00:17:38: Absolutely

00:17:39: And

00:17:39: bringing this conversation back to Europe for a minute The insights curated from leaders at the tech twenty-twenty six conference Were incredibly sharp regarding the geopolitical side of this infrastructure.

00:17:52: Yeah Tim Huttkiss was very blunt on stage.

00:17:54: What did you say?

00:17:55: He stated that Germany needs an AI gigafactory because in his exact words, scale eats sovereignty for breakfast.

00:18:02: Scale eats sovereignty from breakfast?

00:18:03: That is such a great quote.

00:18:05: it perfectly captures the tension and Europe right now he's highlighting this harsh truth.

00:18:10: you simply cannot legislate digital sovereignty if You don't actually own The underlying compute And the physical infrastructure.

00:18:17: Right If all your critical enterprise data and citizen services are just sitting on foreign owned hyperscale servers, your sovereignty is just an illusion on paper.

00:18:27: You are entirely dependent on their terms of service.

00:18:30: there's security protocols and they're national jurisdictions.

00:18:33: to have true sovereignty you need the physical hardware On Your Own soil operating under your own legal frameworks

00:18:39: And Iris Rothbauer captured The solution To this perfectly in her notes from the conference.

00:18:44: She specifically referenced the opening by Lisa Pocosta From Estonia and Carson Wilde Burger from Germany.

00:18:50: Right!

00:18:50: They signed that new innovation partnership Yeah,

00:18:53: and it proves that digital sovereignty isn't just an abstract political ambition anymore.

00:18:58: It's a practical execution challenge.

00:19:00: It requires cross-border trusted infrastructure.

00:19:03: I mean Estonia has been a huge pioneer in digital government services for years.

00:19:07: but they need the industrial scale that Germany can provide

00:19:10: In Wild Burger, Infagos' partnership really shows.

00:19:13: building this sovereign capability requires speed and deep integration.

00:19:17: You have to actually build the data centers, you have to train the sovereign AI models on local languages and local laws.

00:19:24: And you have deploy quantum resistant networks together!

00:19:27: You can't just talk about it in Parliament but go pour concrete and rack servers which really brings all of these clusters we've discussed today full circle doesn't?

00:19:35: It does.

00:19:36: We are entering this era where physical infrastructure is distributed to extreme edge software agents are totally autonomous cyberattacks instantaneous and the underlying cryptography is in a race against quantum decryption.

00:19:52: It's just a massive complex web of dependencies, but I think understanding how the physical power constraints limits the AI models

00:20:09: Absolutely, and if you are a practitioner trying to build resilience systems in this kind of environment You have to assume that the infrastructure.

00:20:16: You deploy today will be operating under completely different stress parameters In just a few years for

00:20:22: sure.

00:20:23: but I do want to leave you with one final thought to mull over Kind of connecting those autonomous agents back to the cyber threats we discussed

00:20:30: okay?

00:20:31: Let's hear it.

00:20:32: If AI is collapsing attack windows from weeks down to literally minutes and quantum computing threatens the very mathematical foundations of our cryptography,

00:20:42: Our only

00:20:42: logical choice is to fully automate our cyber defenses.

00:20:46: Right?

00:20:47: We have to fight machines with machine.

00:20:48: Exactly, but as we hand over the keys to autonomous AI agents To protect our most critical infrastructure letting them make split second decisions to quarantine networks or revoke access How do we ensure?

00:21:00: We don't accidentally automate Our own permanent lockout from these systems when a defensive Machine inevitably misinterprets of signal?

00:21:07: that is a terrifying But very real question.

00:21:10: to end on.

00:21:11: If you enjoyed this episode, new episodes drop every two weeks.

00:21:14: Also check out our other editions on cloud, defense tech digital products and services artificial intelligence sustainability in green ICT Defense Tech and HealthTech.

00:21:24: Thank You so much for joining us On This Deep Dive into the real mechanics driving The Industry.

00:21:29: Keep

00:21:29: Questioning The Consentist Map Your Models To Your Local Frameworks And Don't Forget to Subscribe.