Best of LinkedIn: Cloud Insights CW 05/ 06

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about cloud in calendar weeks five and six.

00:00:08: Frennis enables enterprises with market, technology and competitive intelligence for portfolio and strategy development.

00:00:15: It's good to be back and you know looking at the stack for early February, twenty twenty six.

00:00:21: It feels like we're in a very different landscape than we were.

00:00:24: even six months ago.

00:00:25: It really does.

00:00:26: Our mission today is to unpack the cloud trends from calendar weeks five and six, and I have to say the whole vibe has shifted.

00:00:33: For, I don't know, the last decade cloud was just IT plumbing.

00:00:37: Right, storage buckets, virtual machines.

00:00:39: Exactly, just trying to get a slightly cheaper build in your on-prem data center.

00:00:44: But going through these sources, it feels like cloud has graduated.

00:00:47: It's not just tech anymore, it's It's geopolitics.

00:00:50: That is the perfect way to frame it.

00:00:51: I mean, if you look at the data, we're not talking about features or updates.

00:00:54: We're seeing these three massive, really heavy-hitting themes colliding.

00:00:59: First, you've got the shift from just, you know, simple compliance to hardcore sovereignty.

00:01:06: Then there's the rise of the industrial AI cloud, which is a totally new beast.

00:01:11: And then, well, the brutal reality of cloud economics.

00:01:15: Let's start with that first one, sovereignty.

00:01:17: Because I feel like for the last few years that word just meant, store my data in Frankfurt, please.

00:01:22: Yeah, it was a legal checkbox.

00:01:23: Just to keep the GDPR regulators happy.

00:01:27: But the sources this week suggest it's become something much more... aggressive.

00:01:33: what changed?

00:01:34: it's

00:01:34: moved from a nice to have to a survival strategy.

00:01:37: we're seeing sovereignty go from our regulatory checkbox you know avoiding a fine to a core infrastructure priority.

00:01:43: it's about risk control and frankly national resilience.

00:01:46: I was reading the notes on Bram Verhagen's analysis and he makes this really interesting point.

00:01:50: he says we've been looking at sovereignty all wrong.

00:01:52: we think it's just about location.

00:01:54: right but it's much deeper.

00:01:55: Verhagen's framework is I think essential to understand where this is all going.

00:01:59: Most people stop at data residency.

00:02:00: Where does the file sit?

00:02:02: But he argues, you need three pillars to actually be sovereign.

00:02:05: First is supply chain independence.

00:02:07: Do you actually control the hardware and the code?

00:02:09: Second is legal protection.

00:02:11: Which court has jurisdiction?

00:02:13: Operational

00:02:14: control.

00:02:15: Who is actually pressing the buttons?

00:02:17: Who has root

00:02:18: access?

00:02:18: Okay, that seems to be the one causing all the headaches right now.

00:02:21: Because you can have a server in Berlin, but if the admin is logging in from somewhere else, Does it even matter

00:02:28: precisely and you have risen and made a really sharp point about this.

00:02:31: He argues that this operational control isn't a technical decision anymore.

00:02:35: It's a leadership decision about risk exposure.

00:02:39: So it's not about whether AWS or Azure is technically good enough.

00:02:43: No the question for a CEO now is What happens if there are sanctions?

00:02:48: What happens if there's state pressure from a foreign government?

00:02:51: It's the kill switch question.

00:02:52: It is.

00:02:53: If diplomatic relations break down, does your cloud keep running or are you just digitally stranded?

00:02:59: That context makes the big AWS news this week look even bigger.

00:03:03: Their launch of the AWS European Sovereign Cloud in Brandenburg, Germany.

00:03:08: And looking at the specs shared by Melanie McGrory and Dina Skarda, this doesn't feel like a standard region rollout.

00:03:15: It's not just U.S.

00:03:16: East one, but in Germany.

00:03:18: No, this is a distinct beast.

00:03:20: They're talking about a seven point eight billion investment, physically and logically independent data centers operated entirely by EU residents.

00:03:29: And they're already planning to expand.

00:03:31: It sounds like they're trying to answer that operational control critique

00:03:34: head-on.

00:03:34: They are trying.

00:03:35: And look, for many enterprise customers, this will be enough.

00:03:38: It checks a lot of boxes.

00:03:39: But, and this is a big but, not everyone is buying

00:03:42: it.

00:03:42: I saw that.

00:03:43: Enrico Signoretti had some pretty biting commentary.

00:03:45: He said some CIOs are calling these offerings lipstick on a pig.

00:03:49: It's a brutal visual, isn't it?

00:03:50: It is.

00:03:51: But is it fair?

00:03:52: I mean, seven point eight billion euros is a lot of lipstick.

00:03:55: If they're building separate buildings, hiring separate people, where's the pig?

00:03:58: The pig in his analogy is the legal foundation.

00:04:02: Senority's point, and Martin Hoskin backed this up too, is about the passport of the parent company.

00:04:07: As long as the ultimate owner is a U.S.

00:04:09: corporation, they are subject to U.S.

00:04:11: laws, specifically the Cloud

00:04:13: Act.

00:04:14: Okay, let's pause there, because we hear Cloud Act thrown around so much.

00:04:18: What's the actual fear?

00:04:19: The fear is its extraterritorial reach.

00:04:23: The argument is that the US government can compel a US company like Amazon or Microsoft to hand over data.

00:04:30: Even if the server is in Brandenburg.

00:04:32: Exactly.

00:04:33: Regardless of where the server is or whether the person operating it is a German citizen.

00:04:39: If the parent company wants to do business in the US, they have to comply.

00:04:43: So even if the person pressing the buttons is German, if their paycheck ultimately comes from Seattle.

00:04:48: The risk remains.

00:04:50: That's the skeptical view.

00:04:51: Martin Hoskins said that U.S.

00:04:53: owned providers may never be able to offer full sovereignty in the way a Native European solution could.

00:04:58: It's a spectrum.

00:04:59: AWS is moving along it, but they hit a hard wall called U.S.

00:05:02: law.

00:05:02: So if the hyperscalers are hitting this legal wall, what's the alternative?

00:05:07: We're seeing a different approach from Microsoft and Google, right?

00:05:09: They aren't trying to build it all themselves.

00:05:11: We are.

00:05:12: The partnership model is getting serious traction.

00:05:15: Olivier Breton broke down Microsoft strategy and it's very tiered.

00:05:19: You have the public cloud, a disconnected private cloud, but the most interesting bit is the national partner cloud.

00:05:26: This is where they basically licensed the tech to a local company.

00:05:29: Exactly,

00:05:30: like the blue initiative in France.

00:05:31: You get the Microsoft tech stack, all the tools your people know how to use, but a local European company runs it.

00:05:37: The U.S.

00:05:38: parent technically doesn't have the keys.

00:05:40: It seems like Google's doing the same thing.

00:05:41: I saw Karen Masson posting about S-three-N-S, that's the joint venture with Veils in France.

00:05:46: And that was a huge milestone this week.

00:05:48: Masson noted that S-three-N-S achieved the SecnumCloud three point two qualification.

00:05:53: Okay,

00:05:54: for those of us who don't spend our lives reading French compliance manuals.

00:05:57: How big of a deal is that?

00:05:59: It's the gold standard.

00:06:00: It's arguably the strictest cloud security certification in the world.

00:06:04: So for a hybrid offering US tech but operated by a European defense contractor to hit that level, it validates the whole model.

00:06:12: It says you can bridge that gap.

00:06:13: It's fascinating.

00:06:14: You have this split, the build it here from AWS versus the license it out from Microsoft and Google.

00:06:20: But this isn't just a European anxiety, is it?

00:06:23: Not at all.

00:06:23: Alexander's a hitmare, flagged something happening in Malaysia.

00:06:26: They just dropped RM two billion on sovereign AI infrastructure.

00:06:31: And look at the reasoning he highlighted.

00:06:32: It wasn't just we want better tech.

00:06:35: It was specifically to reduce dependency on US hyperscalers.

00:06:39: That feels like the canary in the coal mine.

00:06:41: It's not about privacy anymore.

00:06:42: It's about not being beholden to one country's tech stack.

00:06:46: It is.

00:06:47: Antonio Irno summed this up perfectly.

00:06:49: He calls it a jurisdictional collision.

00:06:52: On one hand, you have GDPR in Europe treating privacy as a fundamental right.

00:06:57: On the other, the Cloud Act in the US views data access as a sovereign power.

00:07:02: And those two worldviews just they don't mix.

00:07:05: They're fundamentally incompatible.

00:07:07: So you as an organization are stuck in the middle.

00:07:09: You're forced to choose and that friction is reshaping the entire infrastructure map.

00:07:13: Borders on the internet are back.

00:07:15: Which brings us to our second big theme.

00:07:17: Because while the lawyers are fighting over where Dega lives, the engineers are trying to figure out how to use it.

00:07:23: And we saw a massive development in Munich with the industrial AI cloud.

00:07:28: This feels like one of those moments that marks a shift in eras.

00:07:31: I mean, we usually talk about AI in terms of chatbots, right?

00:07:34: ChartGPT writing an email.

00:07:36: But what happened in Munich, this is different.

00:07:39: This is heavy industry.

00:07:40: It was like the Avengers of corporate engineering on one stage.

00:07:43: You had Deutsche Telekom, NVIDIA, SAP, and Siemens.

00:07:46: That lineup tells you everything.

00:07:48: You have connectivity from telecom, hardware from NVIDIA, business process from SAP, and the manufacturing expertise from Siemens.

00:07:55: And they are not playing small.

00:07:56: Yeah.

00:07:57: Ten thousand GPUs dedicated to sovereign industrial use.

00:08:00: One source called it AI meeting the factory floor.

00:08:02: But help me out.

00:08:03: Why does a factory need ten thousand GPUs?

00:08:06: Are the robots mining Bitcoin?

00:08:08: No, think digital twins.

00:08:10: Cedric Nakey from Siemens made a great point about this.

00:08:12: The goal is speed.

00:08:14: If you're designing a wind turbine, you simulate it digitally first.

00:08:17: If you can throw ten thousand GPUs at that simulation, you can run complex models in minutes, not days.

00:08:23: So you iterate faster, bring products to market faster.

00:08:26: Exactly.

00:08:27: But, and this ties right back to our first theme.

00:08:29: You can't do that on a public

00:08:30: cloud.

00:08:31: If you're Siemens or BMW, your engineering data is your crown jewel.

00:08:35: You're not just going to upload blueprints for your next-gen engine to a shared server you don't control.

00:08:41: That's where Dr.

00:08:42: Ferry Abelhassen's comment comes in.

00:08:44: He called this platform a Deutschlandstac.

00:08:46: It's a strong phrase.

00:08:47: It implies total vertical integration.

00:08:50: From the fiber optic cable to the AI algorithm, It's all inside a secure sovereign boundary.

00:08:56: It's a safe house for trade secrets.

00:08:57: And Christine Nakfuss-Nikolik pointed out why this is existentially important for Europe.

00:09:02: What's Europe's raw material?

00:09:03: It's not oil.

00:09:04: It's not rare earth minerals.

00:09:06: Our raw material is intellectual property.

00:09:08: It's engineering.

00:09:09: So if you lose control of that IP because you're running it on insecure infrastructure, the whole economic engine stalls.

00:09:15: Correct.

00:09:16: She argued that Europe needs to be anchor customers for this.

00:09:19: We can't just buy AI from the US.

00:09:21: We have to build the infrastructure that lets our specific economy thrive.

00:09:25: And there was a really interesting note from Marcus W. Hacker about the physical reality of this.

00:09:30: This facility in Munich isn't just a black box sucking up power.

00:09:34: No, and this is becoming a prerequisite for these AI factories.

00:09:38: It's powered by renewable energy, and crucially, it uses waste heat to warm the local district.

00:09:43: You can't just drop a ten thousand GPU cluster into a city without a sustainability plan.

00:09:48: So we have the geopolitical need for sovereignty, we have the industrial ambition of the AI factory, but then we have the bill.

00:09:55: Ah,

00:09:55: yes.

00:09:56: The hangover.

00:09:57: Which brings us to theme three, cloud economics and fine ops.

00:10:01: And honestly, reading these posts, the tone shifts from excitement to, well, pain really fast.

00:10:07: Pain is the right word.

00:10:08: We're seeing this collision of rising costs and bad architectural decisions finally coming home to roost.

00:10:13: Let's start with the elephant in the room that Dwayne Lesnar talked about.

00:10:16: The Broadcom acquisition of VMware.

00:10:18: This has been a seismic event for enterprise IT.

00:10:21: a huge shock.

00:10:23: Lesnar mentioned renewals coming in at five to ten times the previous price.

00:10:27: I mean, you worry about three percent inflation.

00:10:29: How does the CIO even manage a five hundred percent price hike?

00:10:32: You don't.

00:10:32: You panic.

00:10:33: But Lesnar points out the price is just the symptom.

00:10:36: The disease is technical debt.

00:10:38: A lot of this infrastructure is old.

00:10:40: It's out of support.

00:10:42: CIOs are suddenly realizing they're paying a ransom for a tech stack that's actually a liability.

00:10:47: So Broadcom is forcing their hand.

00:10:49: In a way.

00:10:50: Lester suggests this is the moment for a strategic pivot.

00:10:53: He talks about migrating to solutions like Nutanix on AWS.

00:10:56: The idea is to move from CapEx buying hardware that rusts to

00:11:01: OPEX.

00:11:02: Stop hugging your servers.

00:11:03: Stop hugging your servers and start automating compliance.

00:11:05: Use the crisis to force the modernization you should have done five years ago.

00:11:09: But just moving to the cloud isn't a silver bullet for saving money, is it?

00:11:13: In fact, reading Nicholl's fund-readies post, it sounds like it can be a trap.

00:11:18: Oh, Sfondrini highlights a really dangerous dynamic with savings

00:11:21: plans.

00:11:22: These are the deals where you commit to spending, say, a million dollars and you get a discount.

00:11:27: Right.

00:11:28: On paper, it looks like great financial management.

00:11:30: We're going to spend this anyway.

00:11:31: Let's get a discount.

00:11:33: But Sfondrini calls them structural risks, financial lock-in.

00:11:37: Because you're betting on a forecast.

00:11:39: Exactly.

00:11:40: And if that forecast is wrong, you're on the hook for it.

00:11:42: And here's the perverse incentive.

00:11:44: Once you sign that deal, your engineering teams stop trying to optimize.

00:11:49: Wait, why would they stop?

00:11:51: Don't you want to make your code more efficient?

00:11:54: Not if the finance team says... Don't turn off those servers.

00:11:57: We have to burn through this commitment or we lose the discount.

00:12:01: So instead of efficiency, you get waste.

00:12:03: You keep zombie infrastructure running just to hit a spending target.

00:12:07: That is completely backward.

00:12:08: You're optimizing for the contract, not the technology.

00:12:11: And that

00:12:11: leads to bad architecture, which brings us to Mary Simmons and the horseshoe architecture.

00:12:17: I laughed when I read this, but it's actually kind of tragic.

00:12:19: For the listener, what is the horseshoe?

00:12:22: It's a classic anti-pattern.

00:12:24: A company wants to say they're in the cloud, so they move their data there, but, and this is the kicker, their business logic, the heavy processing, is still stuck on an old mainframe on-prem.

00:12:37: So the data travels all the way to the cloud.

00:12:39: Who

00:12:39: realizes it can't do anything and has to U-turn right back to the mainframe to get processed.

00:12:44: And you're paying for data transfer both ways.

00:12:46: You get the latency of the cloud, the cost of the mainframe, and a massive bill for moving data back and forth.

00:12:52: It is the absolute worst of both worlds.

00:12:54: Simmons says this is driving repatriation.

00:12:56: People are pulling workloads back because if this U-turn is just killing them.

00:13:00: It just highlights that you can't lift and shift.

00:13:03: You take a messy process and put it in the cloud, it's just a more expensive, messy process.

00:13:07: You have to re-architect.

00:13:09: And on top of that, you can't even figure out the bill.

00:13:12: Aslan Cater pointed out that teams spend more time normalizing billing data than finding insights.

00:13:18: It's a data wrangling nightmare.

00:13:19: Every cloud provider calls the same thing by a different name.

00:13:23: One calls it an instance, another a droplet, a third a VM.

00:13:26: It's like trying to compare grocery prices, but one store charges by the pound, another by the kilo and a third by the handful.

00:13:33: That's why the industry is rushing towards standards like focus.

00:13:37: It's a fine op standard to normalize this data so you can actually compare apples to apples and make a decision.

00:13:43: So bringing this all together, we have sovereignty, demanding we take control.

00:13:48: We have industrial AI demanding massive infrastructure.

00:13:52: And we have fine ops screaming that we're wasting money on bad architecture.

00:13:56: If you look closely, there's a single thread tying all three of them together.

00:14:00: And

00:14:00: what's that?

00:14:01: Architecture.

00:14:01: Oh, so.

00:14:02: Whether you're designing for legal jurisdiction to satisfy the German government, or designing for high performance GPUs to satisfy Siemens, or designing to avoid that horseshoe effect to satisfy your CFO, the era of lift and shift is dead.

00:14:16: You cannot just move your mess to the cloud.

00:14:18: That reminds me of Fred Earhart's comment.

00:14:20: He said, cloud without enterprise architecture is like a credit card without a limit.

00:14:25: It's the perfect summary.

00:14:26: We are entering the era of design and govern.

00:14:30: You have to design the sovereignty in, you have to design the cost controls in, and you have to design the AI capabilities in from the start.

00:14:38: If you don't, you're just renting liability.

00:14:41: That is a powerful place.

00:14:42: to leave it.

00:14:42: Design and govern or pay the price.

00:14:45: Indeed.

00:14:46: Well,

00:14:46: that wraps up this deep dive into the cloud landscape of early twenty twenty six.

00:14:51: If you enjoyed this episode, new episodes drop every two weeks.

00:14:54: Also, check out our other editions on ICT and tech, digital products and services, artificial intelligence, sustainability and green ICT, defense tech and health.

00:15:02: Thanks for listening.

00:15:03: Make sure to subscribe so you don't miss the next one.

00:15:05: Thanks, everyone.