Best of LinkedIn: Cloud Insights & Sovereignty CW 33/ 34

00:00:00: Provided by Thomas Allgeier and Frennus, based on the most relevant LinkedIn posts about cloud insights in Sovereignty in calendar weeks thirty-three and thirty four.

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

00:00:19: you can find more info description.

00:00:22: So imagine waking up tomorrow and finding out that just because of a geopolitical dispute your cloud provider Just deleted your CEO's email.

00:00:30: Oh man Yeah, and they just shut off your entire data center.

00:00:33: And the thing is that isn't some hypothetical scenario.

00:00:35: it It's exactly what is driving this massive eighty billion-year old panic in the cloud market right now.

00:00:42: Right,

00:00:43: which?

00:00:43: why our mission today to extract top digital products and services trends across LinkedIn for tech professionals.

00:00:50: We really need move past you know surface level marketing noise Exactly!

00:00:55: To figure out actually changing architecture and economics of industry.

00:01:00: So we have three core themes curated from the source material for this deep dive.

00:01:04: First, We're looking at the sudden geopolitical urgency of digital sovereignty.

00:01:09: Second...we are going to break down the hard unvarnished truths of SAP and ERP cloud migrations.

00:01:17: And finally were gonna explore how AI is just completely upending.

00:01:20: fine ops

00:01:21: Yeah!

00:01:22: Across all these areas The underlying current Is that infrastructure no longer an IT decision like where you put your data, how you integrate it and how you actually pay for the compute to process.

00:01:33: I mean those are now fundamental business survival strategies.

00:01:36: definitely so.

00:01:37: let's start with that first theme digital sovereignty Because for a long time, you know sovereign cloud just felt like a buzzword.

00:01:44: It was something companies would put on the slide deck to sound secure or maybe it's just checkboxed for compliance audit.

00:01:49: but looking at our sources this has officially moved from a buzz word into full-on boardroom

00:01:54: crisis.".

00:02:04: Yeah.

00:02:05: He pointed to explicit threats from the U S, To actually cut off business with sixty countries and The quote he referenced was essentially that the US has the absolute right to Cut Off all Business And make those Countries no longer Functional.

00:02:19: Wow.

00:02:20: I mean, when you hear a statement like that on the global stage it forces every CIO to look at their vendor contracts in a completely different light.

00:02:27: Oh yeah You realize you cannot run a nation-state or like a hospital network?

00:02:32: Or A major utility on infrastructure That another country can theoretically just turn off with a memo right.

00:02:38: and Wigan shared a very real world example of this mechanism an action When the US sanctioned international criminal court judges their credit cards were canceled within hours.

00:02:48: Within hours?

00:02:49: That's insane!

00:02:50: Yeah, Microsoft deleted the chief prosecutor email Google closed a judges personal account.

00:02:55: even Amazon Alexa stopped responding.

00:02:58: yet these are geopolitical facts and they're directly driving market behavior.

00:03:03: because of this you know immediate switch flipping capability.

00:03:07: European sovereign cloud spending is projected to surge from twenty billion euros.

00:03:12: two one hundred billion euros by twenty thirty-one.

00:03:15: That is a massive jump, but that brings up a critical question though because if the threat is so clear and spending is quintupling are solutions actually protecting anyone?

00:03:26: Right.

00:03:27: Or is sovereignty just becoming a premium label on the same old products like I feel we're seeing the tech equivalent of slapping an organic sticker in a box of junk food.

00:03:36: That's exactly the tension everywhere right now.

00:03:39: Amit Ayal Govran posted about this pointing out sovereign is essentially just

00:03:44: skew Like just a pricing tier.

00:03:45: Exactly, Just A Pricing Tier.

00:03:47: He called out the security industry for spending The whole year just slapping the word on everything like Sovereign's AC For Network Security or sovereign GPU And he specifically mentioned Armors sovereign AI product.

00:03:58: Yeah, I was reading through Govren's breakdown of that and it honestly blew my mind because if an enterprise buys that sovereign AI Product the request actually gets routed to hosted frontier models.

00:04:11: So you're typing your highly sensitive proprietary data into an interface But it still being sent off to a black box AI model owned by a hyperscaler

00:04:19: right running on their hardware.

00:04:20: yeah Which completely defeats?

00:04:22: The purpose yeah govern.

00:04:23: noted That if you run a product like that through the five sovereignty pillars, it scores one out of five.

00:04:31: Wow!

00:04:32: The immediate tell is whether there's a hosted frontier model in the request path?

00:04:37: If there isn't bought sovereignty then just buy governance with sticker

00:04:41: on them.

00:04:42: True sovereignty requires open weights

00:04:44: and for anyone outside the AI trenches Open weights means you actually have full access to The underlying architecture of the AI model.

00:04:52: Yeah, it's download It You put on your own metal in your own basement or data center And you control the data flow entirely?

00:04:58: You aren't just renting access via an API

00:05:01: right.

00:05:01: without that you Just don't have sovereignty.

00:05:04: and it makes me wonder how regulators are attempting To handle this because the technical reality is moving so much faster than the legal frameworks.

00:05:10: yeah Definitely.

00:05:12: And Kaya Hartung posted about Europe's Proposed Cloud and AI Development Act, or CADA.

00:05:18: it sets up four tiers for cloud services okay?

00:05:21: The top level is meant to shut out foreign providers for highly sensitive public services But analysts say American giants can already clear the entry-level floor effortlessly.

00:05:33: right because regulatory checklists usually just focus on encryption standards and localized storage which AWS or Azure can spin up in an afternoon.

00:05:42: Yeah, that doesn't create true legal separation?

00:05:45: No it doesn't.

00:05:45: hard time actually pointed out That past regulatory initiatives like Gaia X were ultimately called a paper monster

00:05:51: of Paper Monster.

00:05:53: Wow

00:05:53: yeah They wrote a mountain of rules and issued certificates.

00:05:56: But you know standard setting doesn't magically build physical servers A badge Or a tier system Doesn't do the hard work Of separating which workloads Can sit under foreign law And Which cannot.

00:06:06: Right, but while the regulations might be messy.

00:06:08: The market moves are very real.

00:06:11: Elk and Earl shared a bit com study showing that ninety seven percent of German companies now explicitly weigh data location when choosing a provider.

00:06:19: Ninety-seven percent?

00:06:21: That's basically everyone.

00:06:22: yeah And one hundred percent of them prefer Germany.

00:06:25: We're seeing major players act on this fear.

00:06:28: the VW group for instance is building a central sovereign infrastructure in Germany via T cloud to keep their most critical data on shore.

00:06:36: But it isn't just about where the data sits, right?

00:06:39: It's about who has the keys to the room.

00:06:41: Exactly.

00:06:41: We saw posts from Jay Chaudhry and Bernd Wagner highlighting a new partnership Where Zscaler is bringing zero trust security To stack its sovereign European cloud.

00:06:52: Yeah And that specific move addresses operational sovereignty.

00:06:56: The service resides in Europe is operated in Europe and specifically targets the growing requirement that no direct access to the environment is available to personnel outside of jurisdiction.

00:07:05: Right, because you can build a sovereign fortress but if a foreign administrator has remote access to maintain servers your Fortress has back door...

00:07:13: Exactly!

00:07:13: ...and perfectly bridges us into this second core theme of this deep dive.

00:07:18: You can build that sovereign fortress But what are actually putting inside it?

00:07:22: Because for most massive enterprises their critical data doesn't live in a neat little cloud bucket.

00:07:28: It lives deep inside their legacy ERP platform, right?

00:07:32: And if you are changing your cloud strategy, You're eventually migrating your ERP which brings us to the harsh reality check of SAP and Sforana public Cloud Migrations.

00:07:44: Yeah

00:07:44: Victoria King posted a fascinating insight about this from the consulting side.

00:07:48: She talked about The Hunt for genuine S-for-Hanna public cloud expertise, everyone claims they have it but very few actually do and she shared the tell that exposes the fake Public Cloud experts.

00:07:59: It's when a consultant with a decade of private cloud experience says public cloud is basically the same thing just easier.

00:08:05: Oh man

00:08:05: if they say its easier.

00:08:07: They haven't done it.

00:08:07: Yeah The disciplines are entirely different.

00:08:10: Private Cloud is all about customizing the software system to fit their business.

00:08:14: public cloud is fit to standard.

00:08:16: Okay, I want to dig into that mechanism because That Is a massive psychological hurdle for a legacy enterprise.

00:08:22: Oh huge under the old private Cloud mentality A company would spend like thirty years customizing SAP To match their specific conmoluted internal workflows.

00:08:33: But with Public Cloud being Fit-to-Standard You basically have to throw away your custom processes and change Your business to match SAP's unchangeable template.

00:08:42: Right, and King noted that consultants from Business One and By Design Backgrounds actually have this muscle.

00:08:48: They know how to guide a client toward the standard process instead of trying to bend their software.

00:08:53: Legacy SAP Consultants just try Breaks it,

00:08:58: but I have to push back on this entirely though.

00:09:00: Okay telling a massive complex enterprise that they have To change their entire internal business process to fit in off the shelf software template Isn't that the exact opposite of what technology is supposed to do for a business?

00:09:12: I mean shouldn't the software serve the company?

00:09:14: It

00:09:15: feels incredibly counterintuitive i know But massive customization Is actually technical debt.

00:09:20: oh i see.

00:09:21: Yeah, when you heavily customize an ERP.

00:09:23: You create this fragile house of cards.

00:09:26: the moment a vendor releases and update... ...you can't install it without breaking all your custom code.

00:09:31: No that makes sense.

00:09:31: So

00:09:32: public cloud forces standard processes so you get continuous innovation in security updates.

00:09:38: But you're right!

00:09:38: It is massive friction point.

00:09:40: Enterprises are struggling SO hard with this shift That SAP just had to change rules

00:09:45: Right And William Sherry posted about this rule change.

00:09:48: SAP just extended their ECC support to twenty thirty three.

00:09:52: For context for those listening, ECC is the old legacy ERP core that everyone's supposed be migrating away from

00:10:00: right which initially sounds like a huge relief.

00:10:02: now Everyone panicking about the deadline.

00:10:04: But Sherry pointed out that this is an illusion, it's not a blanket extension.

00:10:08: It only applies to select large complex customers who are already on a rise journey Meaning they've bought into SAP's commercial package To move ECC to their private cloud edition.

00:10:20: Exactly

00:10:20: So backstop.

00:10:21: for enterprises that're mid-transformation and realizing how slow The process actually is.

00:10:26: What about

00:10:26: everyone else?

00:10:27: Well if you were a Mid Market Shop That twenty thirty three date means absolutely nothing For mid-market companies, twenty twenty seven and twenty thirty are still your real hard deadlines.

00:10:38: where you're legacy systems lose support.

00:10:41: It's a ticking time bomb.

00:10:42: sitting still is just not an option.

00:10:44: Wow

00:10:45: By the way, if you want to make sure you catch our next deep dive into enterprise tech headaches and solutions.

00:10:50: Make Sure To Hit Subscribe!

00:10:52: You definitely wanna follow along because when actually get in the weeds of doing these ERP migrations it is incredibly complex.

00:10:59: Tiago Rebello shared a post about specific project integrating S for Anna to sales cloud V two.

00:11:05: Oh yeah I was looking at his breakdown.

00:11:07: so revealing Because good news that standard SAP packages worked flawlessly for master data and sales processes.

00:11:13: He said configuring them.

00:11:14: It's genuinely simple, but the integration completely broke down around the codelist replication flows.

00:11:25: Well,

00:11:26: these replication flows essentially map the data values between the two systems so they can talk to each other and The mechanism changed on them.

00:11:33: mid project.

00:11:34: Oh mid-project that's brutal.

00:11:36: Yeah,

00:11:37: the older flows were push based basically firing the data over in forgetting about it.

00:11:41: Then the newer ones we're designed to read compare and reconcile.

00:11:45: So pulling the data and syncing it which is

00:11:48: objectively a better, more synchronized architectural design.

00:11:51: It is definitely but SAP marked the old push-based flows as deprecated after the engineering team had already built around them.

00:11:59: Imagine having to rip out your foundational architecture mid sprint.

00:12:03: Oh my god

00:12:03: it costs The Team weeks of delays To reconfigure everything to the new read and reconcile format And his ultimate advice was incredibly pragmatic.

00:12:12: He said just create the code lists manually in the news system.

00:12:16: manually

00:12:16: Yeah, because they're small and rarely change.

00:12:19: So manual configuration is sometimes faster and more reliable than wrestling with standard automation.

00:12:24: That's the exact kind of hard-learned lesson you only get from actually doing the migration.

00:12:29: It really highlights a bigger issue too.

00:12:32: Modernizing these ERPs Integrating them across clouds And building out sovereign architectures Requires astronomical compute power.

00:12:41: Yeah,

00:12:41: especially as you start injecting AI into those systems and how companies are paying for that.

00:12:46: compute leads us to our final theme.

00:12:48: It is fundamentally breaking traditional fine ops.

00:12:51: This is where the economics get really eye-opening.

00:12:54: Dr.. Monica McIterian posted a brilliant breakdown of How AI?

00:12:57: Is acting like a permanent infrastructure lease.

00:13:00: right because think about traditional cloud usage it was highly variable.

00:13:04: yeah You run a pilot program you check the results And if you don't like it you tune it off.

00:13:08: in your compute build goes back to zero.

00:13:10: but

00:13:10: Agentic AI doesn't work that way.

00:13:12: As McIterian pointed out, an AI agent spends tokens on every single step.

00:13:16: just to clarify a token is basically a fraction of the word.

00:13:20: it's The basic unit of compute cost for an LLM and An agent isn't just a chatbot sitting there waiting for a human to type a prompt.

00:13:29: And agentic system runs in a loop.

00:13:31: It breaks down tasks calls tools evaluates its own work and retries if it fails

00:13:37: and this load runs whether a Hume is at the keyboard or not.

00:13:42: Every workflow you automate adds a processing load that never switches off, right?

00:13:46: It creates this permanent constantly growing minimum floor on your cloud bill.

00:13:51: So if traditional Cloud was like a taxi meter You know pay when you're actually moving.

00:13:57: Agentic AI is like signing a fifteen-year lease on building that you can't walk away from, and the landlord just keeps raising the

00:14:03: rent.".

00:14:03: That's

00:14:04: the perfect way to look at it!

00:14:05: The floor of your invoice just keeps rising... Traditional cloud contracts aren't designed for this.

00:14:10: per token pricing commits you nothing ...and reservations maybe give you a discount in one year or three.

00:14:15: But permanent AI workload needs capacity for a decade — no one sells that because hardware changes so fast.

00:14:21: Which is why FinOps has to evolve beyond just looking at the billing dashboard.

00:14:25: Sudeep Khair posted about this, stating that cloud cost is an architecture problem disguised as a finance problem.

00:14:32: Absolutely because if a FinOps team points out a workload is expensive six months after deployment they've already lost the battle.

00:14:40: The biggest costs are decided long before the first bill arrives.

00:14:43: They're decided by how you design the system To scale and move data in

00:14:49: And we saw a highly practical example of solving this at the architectural level from Milan Jovanovich.

00:14:54: He talked about building self-hosted LLM chatbot on GPU infrastructure, and GPUs are the ultimate bottleneck in cost driver right now.

00:15:03: Oh totally!

00:15:04: If you send every single user request straight to the GPU they queue up creating latency... ...and you waste incredibly expensive compute time answering the same question?

00:15:13: Exactly let's walk through his solution because I thought it was brilliant.

00:15:16: Yeah lets hear.

00:15:17: Imagine you have a customer service AI.

00:15:20: If a thousand people asked, what is your return policy?

00:15:24: A traditional setup sends that question to the forty-thousand dollar GPU a thousand times charging.

00:15:29: you compute and token costs every single time

00:15:32: which is just throwing money away right

00:15:34: so Avonimich's architectural solution was to use edge caching via small gateway.

00:15:39: You basically put a bouncer at door.

00:15:40: yes The Edge Gateway is the bouncer.

00:15:43: It intercepts a request before it hits expensive hardware, the gateway recognizes common question grabs cash to answer already generated previously and returns instantly.

00:15:53: So the GPU does zero work?

00:15:55: Zero!

00:15:56: You're solving massive financial problem simply by preventing LLM from generating exact same tokens twice.

00:16:02: It solves cost at root.

00:16:04: that's great.

00:16:04: but aside architecture other major hurdle in fine ops right now visibility.

00:16:10: Nitin Padoria posted about a massive blind spot in object storage.

00:16:14: So object storage is where companies dump all their unstructured data, right?

00:16:18: Like images logs backups and they're stored as blobs or binary large objects.

00:16:23: Yeah

00:16:23: Right.

00:16:24: And padaria noted that phenops teams know exactly what their total object storage costs them every month But they have no idea What is actually inside those hundreds of storage accounts and millions of blobs

00:16:36: and visibility you cannot act on pretty much useless.

00:16:39: Exactly!

00:16:40: A phenops tool might tell you that you spent a hundred thousand dollars on storage, but if it can't tell you which specific folders are holding useless cold data... You can optimize them.

00:16:50: And the technical term for those folds is the prefix level?

00:16:53: Right Because companies lack that prefix-level visibility.

00:16:57: They usually just rely on blanket calendar rules like moving everything older than ninety days to cheaper slower storage tier.

00:17:04: But that's incredibly inefficient

00:17:06: It's just guessing.

00:17:07: Right, Padaria pointed out that the Lucidity platform was just extended to cover object storage.

00:17:13: by actually showing how data is accessed down to that specific prefix level they were able to cut hot-tier waste by up to ninety percent.

00:17:21: Ninety percent?

00:17:22: Wow!

00:17:23: Yeah See when you can see the waste granularly and fix it in the same motion That one story finally gets cheaper

00:17:29: And... When get economics right from start The savings are staggering.

00:17:35: Patrick Newbold shared a success story where CMS became the first federal agency to use The One Gov Cloud Agreement.

00:17:42: Oh, okay!

00:17:43: Yeah they moved a mission-critical contact center and saved forty six percent on the workload.

00:17:47: Forty Six percent on government work load.

00:17:50: They achieved that by being transparent in engaging with general services administration early To create long term predictability.

00:17:58: It proves that when governance, architecture and phenopsa line the cloud can actually deliver on its financial promises even for massive government workloads.

00:18:07: We have covered a vast amount of ground today from international sanctions to the weeds of SAP codelists.

00:18:13: yet

00:18:13: we have if we synthesize all of the sources we examined The clear through-line is that we are witnessing the end of the Cloud as a passive storage layer era?

00:18:23: We've moved into a highly pragmatic complex phase.

00:18:26: infrastructure placement is now dictated by geopolitical borders and sovereignty mandates.

00:18:31: It's dictated the hard, unyielding realities of legacy ERP integration…and it being forcefully shaped.

00:18:39: I

00:18:42: want to leave everyone with a final thought to mull over.

00:18:44: We're seeing these massive shifts collide if AI tokens are becoming a permanent, constantly running utility cost and digital sovereignty mandates dictate exactly where that utility must be generated And who can maintain it?

00:18:57: Are we moving toward a future Where enterprise IT budgets look less like traditional technology spending and much more Like national energy grid contracts?

00:19:06: If you enjoyed this episode, new episodes drop every two weeks.

00:19:09: Also check out our other editions on digital products and services, defense tech AI and agentic systems.

00:19:14: green ICT in sustainable AI and health tech.