Best of LinkedIn: Cloud Insights & Sovereignty CW 43/ 44

00:00:00: Provided by Thomas Allgaier and Freiness, based on the most relevant LinkedIn posts about cloud insights in sovereignty in calendar weeks forty-three and forty four.

00:00:08: Freiness has built a sovereign cloud market radar for ICT providers with weekly hot news monthly reports quarterly leadership presentations And AI podcasts For field teams.

00:00:17: you can find more info In The Description.

00:00:20: So glad we got that out of the way because We have A LOT to cover today.

00:00:25: Oh yeah totally.

00:00:27: We are doing a deep dive into the top digital products and services trends that are just buzzing across LinkedIn right now.

00:00:34: And if you're listening, The Core Mission for You Today is really understanding this massive industry shift.

00:00:41: it's basically a shift from wanting capability to do things in cloud To actually demanding absolute control over your infrastructure.

00:00:49: Right because For the longest time we wanted cool features Like give me tools Give me scale.

00:00:57: But now it's like companies are waking up and realizing they don't actually own the house.

00:01:01: They're living in

00:01:02: exactly.

00:01:03: I mean imagine you go out at least this amazing Like two hundred thousand dollar Ferrari,

00:01:09: okay?

00:01:09: I'm imagining it

00:01:11: right.

00:01:12: but then You find out that dealership is legally allowed to just track your GPS And they can you know remotely throttle your engine whenever they feel like a little brutal.

00:01:21: And on top of that, they casually just hold onto a spare master key to your garage.

00:01:25: So yeah you have the technical capability to drive the car but you have absolutely zero actual control over it.

00:01:32: That is exactly what's happening.

00:01:33: and if you're CIO or an infrastructure leader tuning into this You probably nodding along right now because scenario hits way too close.

00:01:40: Yeah for sure.

00:01:42: Which brings us to first major theme we are seeing which is sovereign cloud specifically how people are realizing its no longer about where data center is physically built.

00:01:51: Right, the whole location thing

00:01:53: right

00:01:53: because for years The industry just treated sovereignty as like a geography problem.

00:01:58: Yeah Like oh the servers in Frankfurt were good exactly.

00:02:02: but that physical definition Is it's getting completely dismantled?

00:02:05: Right now?

00:02:06: Trond Ruvan actually posted some really great insights on this recently.

00:02:09: Oh

00:02:09: yeah I saw That one five layers right.

00:02:12: yep Exactly he introduces This Five-Layer Model For European Cloud Sovereignty And He Breaks It Down Into Residency operations, jurisdiction continuity and autonomy.

00:02:23: Right

00:02:24: And when you look at it that way the key takeaway is that residency like just having the server physically sitting inside your country's borders.

00:02:31: That's Just The easy layer

00:02:33: because It's just physical real estate.

00:02:35: exactly location Is Easy Control?

00:02:38: Is the hard part.

00:02:39: the Jurisdiction and Autonomy layers are incredibly difficult Because that's where You have to figure out who actually controls the legal and operational levers.

00:02:47: Right, and Fadi Dasis takes that logic even further in his post.

00:02:51: He basically argues that sovereignty isn't a data center address at all like it has nothing to do with the zip code.

00:02:56: It's entirely about who holds the encryption keys

00:03:00: Which I mean?

00:03:01: That just makes total sense.

00:03:02: Yeah

00:03:02: Okay let's unpack this within analogy.

00:03:05: Is this like renting high security vault but The landlord still legally hold some master key?

00:03:10: i mean how sovereign are you really In that scenario?

00:03:13: You're not At All And that exact tension is just ripping through the European tech sector right now.

00:03:19: Amit Ayal Govran, along with some legal experts at Tagal Florence specifically Cecil Te Arjalu and Gregoire Fushar they pointed out this massive structural crack in the current sovereign cloud offerings

00:03:33: from The Big Hyperscalers you mean?

00:03:34: Right?

00:03:36: take Microsoft for example.

00:03:37: I mean They're expanding their sovereign Cloud footprint Europe like crazy And they give local partners operational oversight.

00:03:45: They give you tamper evident logs that even let you hold the root keys in some cases,

00:03:50: so it looks sovereign on paper right

00:03:52: but as govern points out structurally That really just gives Europe what he calls visitation rights because Microsoft ultimately retains legal custody of the data.

00:04:03: But wait, Microsoft is spending billions pouring concrete building these local data centers in Europe specifically to solve this issue.

00:04:10: Are you telling me that localizing the servers legally means nothing?

00:04:13: Well, legally it gets completely overridden by The U.S.

00:04:16: Cloud Act.

00:04:17: Oh right...the cloud act?

00:04:20: Yeah

00:04:20: That's a crucial context here.

00:04:25: U.S.-based tech providers have to hand over data if requested by a US warrant regardless of where those physical servers are located globally.

00:04:33: Wow So, If Washington comes knocking the legal fight doesn't happen in Paris or Berlin.

00:04:38: it belongs To The Provider Over In Redmond Washington.

00:04:41: so European hosting alone just isn't enough.

00:04:44: not.

00:04:44: if you want true digital sovereignty You need absolute technical legal and organizational control.

00:04:50: That is such a staggering reality check, and you can see organizations reacting to this in real time just by looking at where they're putting their money.

00:04:57: Yeah follow the money right?

00:04:58: Always.

00:04:59: Bart van der Bees highlighted this huge move.

00:05:01: recently The Dutch Ministry of Defense Just signed this specialized agreement with SAP To build a sovereign cloud environment entirely inside the Netherlands.

00:05:09: Oh wow yeah.

00:05:10: And what makes it unique Is that its built on SAP's own Cloud infrastructure Operating completely independent Of any non-European hyperscaler.

00:05:17: Thats'a massive shift And it's not just governments doing this.

00:05:21: On the commercial side, Seb Johnson noted that Starling Bank is actively training a major AI model right now on UK sovereign cloud via SIVO.

00:05:29: Okay so domestic provider

00:05:31: Exactly.

00:05:31: and they're processing over fifteen billion anonymized transactions.

00:05:35: When you are dealing with that kind of sensitive financial data They deliberately chose a domestic cloud to guarantee jurisdictional autonomy.

00:05:41: But here is the catch.

00:05:43: True sovereignty means taking ownership of architecture and keys But ownership of the architecture inevitably means taking ownership of The Bill.

00:05:51: Ah, yes!

00:05:52: The bill always comes to do.

00:05:53: Exactly If you build this highly isolated sovereign fortress how do keep it from just bankrupting your entire IT budget?

00:06:01: Which transitions us perfectly into our second theme... ...the evolution of FinOps and cloud cost management.

00:06:08: Alright because FinOps is having a bit an identity crisis lately A

00:06:11: huge one.

00:06:12: For long time the FinOps team was basically just the IT cleanup crew Like the people who show up on Monday to yell at engineers for leaving expensive test servers running all weekend.

00:06:21: The

00:06:21: cloud police!

00:06:22: Yeah, exactly.

00:06:24: But Diana L brings a really fascinating point in her post.

00:06:28: She says FinOps shouldn't actually own Cloud costs At All?

00:06:32: The business and product team should

00:06:33: Interesting.

00:06:34: So what does FinOps Actually do then?

00:06:37: They

00:06:37: exist To Provide Visibility In Context so they give the engineers the data they need to understand the financial impact of their code decisions.

00:06:44: Okay that makes sense.

00:06:45: In Thomas Valley builds on this too, he noted that a mature FinOps function does way more than just cut costs or report what you already spent.

00:06:54: It actually shapes your technology investment decisions up front

00:06:58: Right.

00:06:58: it's the whole thing to left Exactly.

00:07:00: But here is where gets really interesting and I have push back little.

00:07:03: If FinOps teams are stepping back in providing visibility aren't they building expensive dashboards engineers will completely ignore when busy trying ship of feature?

00:07:14: That is the exact trap.

00:07:16: You nailed it, and Nitin Badaria breaks down mechanically why that happens.

00:07:21: Oh I read his post The Five Steps Thing right?

00:07:24: Yeah!

00:07:24: The five steps to fix a cloud cost problem.

00:07:26: It's detect investigate recommend apply and verify.

00:07:31: Right.

00:07:31: And for decade we've poured venture capital into automating those first three steps.

00:07:36: So We have amazing tools.

00:07:37: they can detect us spike find their root cause even recommended fix But whole thing just stalls completely at action layer.

00:07:44: Right, because actually applying the fix which is step four introduces risk.

00:07:48: Like if an engineer touches a live production environment to resize a server they might bring the whole app down

00:07:54: Exactly.

00:07:56: So what happens?

00:07:57: The tool makes a recommendation gets turned into a Jira ticket and it just rots in the backlog.

00:08:02: The engineers never gonna risk taking down production.

00:08:04: Just save few bucks on a server.

00:08:07: so Step five verifying the savings just never happens

00:08:10: Never.

00:08:11: The only way to actually realize the savings is to automate the action itself.

00:08:15: Take the human hesitation completely out of

00:08:17: it."

00:08:18: And Sean Barker shared a really brilliant example.

00:08:21: how do this in real world, he showed right.

00:08:23: sizing cloud environments into actual workload cycles and automating that resizing inside CICD pipelines cut one his client's annual spend by thirty-one

00:08:34: percent.

00:08:35: It's huge!

00:08:37: Instead of provisioning for absolute peak traffic twenty-four seven, Barker's approach injects resource limits right into the deployment code.

00:08:44: Right because most enterprise apps aren't flat.

00:08:46: they have huge spikes at month end and then just sit idle.

00:08:50: if you automate that resizing in the pipeline You catch the waste before it even reaches production

00:08:55: exactly.

00:08:56: And this optimization is getting so granular now, Arsham Islami actually found a stat that nearly one hundred and eighty US companies are hiring specifically for snowflake cost tuning as dedicated job specialty.

00:09:08: Oh wow!

00:09:09: Just for Snowflake?

00:09:10: Yeah...a dedicated role.

00:09:12: It's wild.

00:09:13: And by the way, just a quick pause.

00:09:15: If you are finding these insights valuable as we map out all of this crazy shifts in the market.

00:09:20: make sure to subscribe so that you don't miss our future deep dives!

00:09:24: We do it every couple weeks.

00:09:25: Definitely hit subscribe.

00:09:27: But back to that snowflake example.

00:09:29: why is cost optimization suddenly becoming hyper-specialized?

00:09:33: It's

00:09:33: because workloads themselves are fundamentally shifting.

00:09:36: We aren't just running predictable web apps anymore.

00:09:39: We are entering the era of AI infrastructure, which is a totally different beast.

00:09:43: Oh man yeah.

00:09:44: let's get into our third theme The AI Infrastructure Reality Check.

00:09:48: So Jason Young Jin Chang talked about this.

00:09:51: He notes there has been massive recalibration over last decade.

00:09:55: we've gone from A cloud-first mentality to an AI first mentality

00:09:59: Which changes all the math

00:10:00: Completely!

00:10:04: And GPUs, they are a scarce resource now.

00:10:06: You have to reserve them way in advance

00:10:08: right?

00:10:09: So Are we basically just bolting the Ferrari engine onto a broken chassis like our?

00:10:14: these companies actually ready for this?

00:10:16: structurally?

00:10:17: Mark Vigoroso argues that there not.

00:10:20: He points out that enterprises are basically repeating The exact same mistakes They made during the first cloud migration There doing a lift and shift but at This time with a gentek AI

00:10:31: Just slapping AI onto old problems.

00:10:33: Right, they're bolting autonomous agents on to broken undocumented legacy processes.

00:10:39: So the true bottleneck isn't the technology its governance.

00:10:42: You can have an AI making hundreds of changes a minute if your change advisory board still meets on Thursdays to manually approve tickets.

00:10:51: Yeah that is a recipe for total chaos.

00:10:53: But you look at this sheer amount of money flowing into AI infrastructure and it feels a bit terrifying.

00:11:00: A lot of people are calling it a bubble.

00:11:01: Yeah, but Ben Edmond pushed back on that.

00:11:04: he compared it to the telecom bubble of two thousand one which was built on you know junk debt and literally zero cash flow.

00:11:10: they were just hoping demand would show up

00:11:13: Right?

00:11:13: The whole build-it-and-they will come thing

00:11:15: Exactly.

00:11:16: But today the hyperscaler AI Capax is being funded by massive operating cashflow from these huge tech monopolies And the demand has strictly supply constrained.

00:11:26: They physically cannot build data centers fast enough.

00:11:29: And to handle that insane demand, the tech managing cloud has to adapt.

00:11:33: Carthike and Jay highlighted a new AWS well-architected agent.

00:11:38: It actually reviews your environments autonomously and prepares fixes for you.

00:11:43: We are moving straight toward Agenetic CloudOps.

00:11:46: But Yulia Petrucin-Toon warned about this, right?

00:11:49: She said giving these agents simple API access isn't enough.

00:11:53: They require deep persistent context of the infrastructure because if they don't understand what downstream dependencies their breaking an autonomous agent is really just a super fast way to accidentally delete your production database

00:12:06: in milliseconds.

00:12:07: Literally, so if AI is demanding entirely new bare metal performance and we need this strict governance companies are being forced to rethink their baseline virtualization strategies.

00:12:18: which brings us perfectly to our final theme the hybrid reality and the massive shake-up happening with VMware AI and sovereignty.

00:12:26: you're actively pushing workloads back to the edge and back on prem

00:12:30: And the stats on this are wild.

00:12:32: Andre's clever shared that nine out of ten VMware customers are currently looking for an exit.

00:12:37: Nine outta ten, and sixty percent are actively looking for multi-hypervisor strategies.

00:12:42: Man is a massive exodus!

00:12:45: It is... Hybrid used to be just a stepping stone like a waiting room before you moved fully into the public cloud but now hybrid is the destination

00:12:55: And Phil DeMarco pointed out that this mass migration isn't just about the sheer price increase after the Broadcom acquisition.

00:13:01: It's about licensing predictability.

00:13:04: What do you mean?

00:13:05: Well, it is shift to per core minimums and forced bundles.

00:13:09: If your are forced pay for a minimum number of cores regardless.

00:13:12: what would actually use completely destroys capacity planning.

00:13:16: So what does this all mean then?

00:13:17: Are companies just fleeing VMware to lock themselves into a different vendor's ecosystem because that sounds like trading one set of handcuffs for another.

00:13:26: You'd think so, but John Hunter pointed out it really depends on the company.

00:13:29: For example, NTT Docomo just signed a ten-year deal with VMware.

00:13:34: Wait...really?!

00:13:35: A ten year deal right now!

00:13:37: Yeah Because when mission critical software is deeply embedded in your operations dependencies shape you decisions.

00:13:43: For them, migrating thousands of workloads is probably way more expensive and risky than just paying the new licensing fees.

00:13:50: That's a great point!

00:13:51: It really comes down to unit economics.

00:13:53: Christoph Popey did this calculation on running five hundred Kubernetes clusters.

00:13:58: He found that true cost differentiator isn't software list price

00:14:02: Right it's billing unit.

00:14:03: Exactly Are you billed per core or VCPU?

00:14:06: Or by head count of?

00:14:07: engineers need manage open source alternative.

00:14:10: That unit dictates your entire strategy.

00:14:13: And Nolan George closes out this whole debate perfectly, he observed that for steady predictable Kubernetes workloads dedicated bare metal servers are just cheaper and better than you.

00:14:22: save

00:14:22: the public cloud for the bursty stuff

00:14:24: exactly.

00:14:25: wow okay for unpredictable demand.

00:14:27: Okay we have covered a lot today

00:14:29: really.

00:14:30: so Just to recap We looked at sovereign cloud and how control matters more than physical location.

00:14:35: we talked about fin ops moving from reporting to actual automated action

00:14:39: Yep, the AI infrastructure reality check

00:14:42: And finally The VMware shakeup proving that hybrid is here to stay.

00:14:46: You know it's a complete rewiring of How we think about IT architecture?

00:14:51: Actually I want leave you with one final thought to mull over.

00:14:53: Oh lay on me.

00:14:55: So as AI agents gain More autonomous Control Over our Cloud Ops Hybrid infrastructure gets wildly complex for humans to manage.

00:15:04: How long will it be until our enterprise AI agents are autonomously negotiating spot pricing and compute resources in real time directly with the cloud providers, AI agents?

00:15:16: Oh man a machine-to-machine financial negotiation.

00:15:19: Exactly we might literally just be building the infrastructure for AI to buy and sell from other AI

00:15:25: because driving itself paying for its own gas.

00:15:27: We just have to make sure we still hold the master keys.

00:15:30: Well, that's all for today!

00:15:32: If you enjoyed this episode new episodes drop every two weeks.

00:15:36: also check out our other editions on digital products and services defense tech AI and agentic systems green ICT in sustainable AI and health tech.