Best of LinkedIn: Cloud Insights CW 09/ 10

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

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

00:00:14: Alright welcome to The Deep Dive!

00:00:16: And today we are unpacking top-cloud trends Straight from the front lines of LinkedIn.

00:00:22: Yeah, no fluff here just the most critical insights for you The ICT and tech professionals out there.

00:00:28: we've basically clustered this conversation into four major themes that are Completely shaping the industry right now.

00:00:34: so about multi-cloud strategy in sovereignty AI driven cloud operations phenox and cost governance.

00:00:41: And finally cloud architecture and platform engineering.

00:00:44: okay?

00:00:44: Let's jump right into this first thing because Digital sovereignty has completely morphed.

00:00:49: Oh, absolutely!

00:00:50: It is no longer just this European policy buzzword that gets thrown around in white papers you know?

00:00:55: Yeah it's a massive global operational reality now.

00:00:59: and if you're navigating cloud procurement right now You know there was a really intense debate happening about what sovereignty actually means when the rover meets the road

00:01:08: Exactly.

00:01:10: And to ground this let's look at a really sharp insight from Andrew Byrd.

00:01:15: He points out something fundamental that technical teams often forget, which is that boards of directors don't buy technology.

00:01:23: They by control and they buy risk management.

00:01:26: Which changes the whole conversation

00:01:28: entirely.

00:01:29: when a board looks at cloud strategy today, They aren't looking at compute speeds And they honestly don't care about your multi-cloud container flexibility.

00:01:36: There are looking at jurisdiction exposure

00:01:38: US Cloud Act

00:01:40: us cloud act exactly so.

00:01:41: a sovereign cloud isn't a tech upgrade to them.

00:01:44: It is strictly a control decision.

00:01:46: right

00:01:46: but let's talk About how that plays out in reality because it leads straight To this intense debate we saw around The new AWS European sovereign class.

00:01:54: Oh yeah, the Amazon investment.

00:01:56: You know?

00:01:56: Amazon just spent seven point eight billion euros building a dedicated cloud in Germany.

00:02:02: instead of new GMBH they're hiring EU residents staff.

00:02:06: They are putting up all these walls.

00:02:07: Sounds

00:02:08: great on paper

00:02:09: It does.

00:02:10: But then you read someone like Peon de Bruyne and he argues that this massive multi-billion euro operational theater is legally meaningless.

00:02:19: it sounds harsh when you say Like That but his logic is structurally sound.

00:02:23: Explain

00:02:23: that though.

00:02:24: How is an eight billion euro investment legally meaningless?

00:02:28: Because of corporate structure.

00:02:29: I mean, that German subsidiary is still one hundred percent owned by Amazon Inc.

00:02:34: and for anyone who hasn't read the fine print lately under The U.S.

00:02:37: Cloud Act a US parent company can be compelled to produce data held by its subsidiaries

00:02:43: Regardless of where the servers are

00:02:44: regardless of where that data physically sits.

00:02:46: So if the US government issues a subpoena to Amazon in Seattle It doesn't matter.

00:02:50: If your data is sitting at a bunker and Frankfurt managed by German LLC legally, The parent company has to hand it over.

00:02:58: so Pete young's takeaway Is that?

00:03:00: True bulletproof sovereignty means the vendor never actually possesses the data in the first

00:03:06: place.

00:03:07: Okay, so legal teams you're sweating.

00:03:10: but what about the engineering teams?

00:03:12: because Harry Malone is weighed in on this exact AWS Sovereign Cloud.

00:03:16: Yeah, he had a very different perspective.

00:03:18: He pointed

00:03:19: out that it's only been live for a month and the engineers are already pulling their hair over the friction.

00:03:24: Cross-region debugging is suddenly much slower.

00:03:27: You can't just have US based Tier III support engineer login And quickly troubleshoot an issue in middle of night anymore.

00:03:34: Development velocity actually dropping.

00:03:37: That exactly where disconnect happens between engineering floor & C suite.

00:03:41: How so?

00:03:42: Well,

00:03:42: to an engineer trying to hit a Friday deployment deadline that partition is a nightmare.

00:03:47: But the C-suite dealing with strict national regulations... That friction is a mathematically verifiable feature.

00:03:53: Oh I see!

00:03:54: Think about it.

00:03:55: if a global support engineer cannot cross The Digital Partition To Fix A Bug....A Foreign Government Cannot Cross It To Extract Your Data.

00:04:02: You are actively trading engineering speed for Verifiable Legal Control

00:04:07: Which makes sense on paper.

00:04:08: But then Bram Verhagen jumps into this debate and essentially says, look we're having this ideological conversation.

00:04:14: Fifteen years too late.

00:04:16: he's not wrong

00:04:16: right.

00:04:17: Europe failed to build true massive infrastructure alternatives a decade ago.

00:04:23: so these hyperscaler solutions even with their US legal ties in cloud act exposure are just a pragmatic bridge.

00:04:29: We have to use them because there is no other realistic option at scale.

00:04:33: And Kirsten Wisk actually put that exact theory to the test.

00:04:37: He didn't just theorize, he shared a case study of building Domain.

00:05:01: pricing was randomly higher.

00:05:03: Wow, and the real kicker at the end of day They still had to integrate Google or Apple for social logins And they still needed clawed further AI capabilities.

00:05:12: so you still ended up tied to us anyway.

00:05:14: exactly.

00:05:15: His conclusion is that the sheer convenience still entirely belongs to the US hyperscalars.

00:05:20: Building a truly sovereign stack Is an

00:05:23: active

00:05:23: difficult and expensive architectural choice?

00:05:26: You can't just passively fall into it.

00:05:29: what's fascinating?

00:05:30: This isn't just a European headache anymore.

00:05:32: It's becoming the global standard.

00:05:34: Hitan Mehta highlighted that ASEAN countries are now aggressively tightening their data localization laws.

00:05:40: Right, Southeast Asia is moving fast on this.

00:05:42: Vietnam Indonesia Malaysia they're all raising the floor.

00:05:46: And because of that exact cloud act anxiety, we just talked about national telcos and data center operators in those regions are stepping

00:05:53: up.

00:05:53: They see the market opportunity

00:05:54: huge opportunity.

00:05:56: they don't Just want to provide your fiber cables anymore?

00:05:58: They want to become the custodians Of your sovereign infrastructure.

00:06:02: They Are offering The legal domicile a National Trust That A foreign Vendor Simply Cannot Guarantee.

00:06:09: So whether you're Deploying In Berlin or Bangkok the question of who physically and legally controls your servers is redrawing the infrastructure map entirely.

00:06:17: Yeah, but while we're debating where The Cloud lives how it actually operates as undergoing an arguably bigger revolution which

00:06:24: brings us to our second theme

00:06:26: AI driven cloud operations.

00:06:28: because AI Is no longer just a cute autocomplete feature in your IDE.

00:06:33: It is becoming the core operating model of the infrastructure itself.

00:06:37: Yes, and Fabian Pagel captured this shift Perfectly when he discussed agentic coding.

00:06:42: If you were an engineer listening to this, You know we've been using AI To finish our boilerplate sentences for a year now

00:06:47: sure co-pilot and all that.

00:06:49: but Peggle says We are moving past That.

00:06:52: we're now deploying AI agents that understand entire workflows.

00:06:56: They check your specific architectural context And literally prepare pull requests while the engineering team is still drinking their morning coffee.

00:07:04: It's

00:07:04: a profound shift in daily operations.

00:07:07: He noted that his team is building functional cloud-native prototypes

00:07:12: because that sounds like a recipe for catastrophic production outage.

00:07:20: Well,

00:07:20: that's the evolution of The Engineer's role.

00:07:23: Pagle argues at the bottleneck in development is no longer the physical act of writing code...

00:07:27: Okay!

00:07:28: ...the bottleneck is now our human ability to precisely define the problem.

00:07:32: The developer job is shifting from typing every line to rigorously reviewing and validating the architecture that the agent proposes.

00:07:40: But if we look how massive hyperscalars are institutionalizing this Leon Godwin broke down micro Microsoft's new vision for agent cloud operations, and it is way more structure than just letting a bot loose on your code base.

00:07:52: Much more structured.

00:07:54: Microsoft is embedding six specific AI agents directly into Azure Copilot.

00:08:00: these aren't isolated chatbots.

00:08:01: they're connected system covering the entire life cycle.

00:08:05: you have agents from migration deployment observability optimization resiliency and troubleshooting

00:08:11: And Godwin highlighted the absolute most critical part of that rollout for IT leaders.

00:08:17: These agents strictly honor your existing role-based access controls and policies,

00:08:22: right?

00:08:22: So it's not The Wild West exactly.

00:08:24: if a junior developer doesn't have permission to spin up a massive database cluster their AI agent Doesn't Have Permission either.

00:08:31: It's traceable its auditable and it keeps those strict compliance teams who are just talking about relatively happy

00:08:36: Exactly.

00:08:37: And speaking of compliance and sovereignty, Alman Ibrahimovic shared a massive update that perfectly bridges our first two themes.

00:08:44: Oh the disconnected operations?

00:08:45: Yes Microsoft just made Azure local disconnected operations generally available.

00:08:50: this means you can now run your mission critical infrastructure and even GPT class AI models completely air-gapped an offline.

00:08:57: Ibrahimovich actually called it The Ultimate Cloud Act Killer which is bold claim.

00:09:01: but think about the mechanics.

00:09:03: If your prompts, training data and outputs never touch a public internet cable you physically sever that operational control hook.

00:09:11: that terrifies legal teams.

00:09:13: Because the data literally can't leave the building.

00:09:16: Your intellectual property stays entirely inside of physical buildings but still get the operational power of advanced multi-modal AI.

00:09:24: It is fundamental.

00:09:25: shift in how we deploy intelligence securely.

00:09:28: Sounds like the holy grail for highly regulated industries.

00:09:32: But before we get totally lost in the AI hype cycle, Kishop A offered a really stark reality check on LinkedIn.

00:09:39: He warned that you cannot scale AI operations faster than your cloud latency map allows.

00:09:44: That is such an overlooked constraint?

00:09:46: It really is.

00:09:47: We

00:09:47: have spent the last decade building legacy clouds optimized for efficiency cost consolidation and bash processing.

00:09:54: But AI operations, especially these agentic workflows demand real-time responsiveness.

00:09:59: Right!

00:10:00: If you are relying on an AI agent to troubleshoot a live outage... ...you can't have it hanging because of cross region latency.

00:10:06: Precisely Kashiya points that if your underlying infrastructure is slow the symptoms show up quietly.

00:10:13: at first.

00:10:14: Your automation pauses during peak demand.

00:10:16: Cross Region Workflows start to lag.

00:10:19: Just minor annoyances at first.

00:10:20: Right, but then suddenly your engineering teams start reintroducing manual checkpoints because they don't trust the automated system anymore.

00:10:28: latency stops being just a technical nuisance and becomes a massive leadership issue.

00:10:33: if The System feels slow human trust declines And you're transformative.

00:10:37: AI strategy just becomes an expensive incremental toy

00:10:41: which brings up the elephant in the room cost.

00:10:44: You need sovereign control, you need advanced AI agents and your underlying pipes still need to be lightening fast.

00:10:50: Yeah that's not cheap.

00:10:51: That

00:10:51: sounds incredibly expensive And that pivots us right into theme three Fiannops and cost governance.

00:10:57: Cloud billing is already everyone's favorite headache But with massive AI workloads spinning up globally it is officially the next frontier of cost management.

00:11:05: It is!

00:11:05: And Victor Garcia put The current state affairs very bluntly.

00:11:08: He said cloud cost estimation Is completely broken Completely

00:11:11: broken.

00:11:12: If you've ever managed a cloud budget, You know the cycle.

00:11:15: Teams spend weeks guessing at what infrastructure costs might be.

00:11:18: They deploy their resources and then they sit around in dread waiting for month

00:11:22: to end And then invoice drops.

00:11:24: When that massive invoice finally drops Panic sets in Finance starts demanding answers Trusting the fine ops team just

00:11:32: plummets.

00:11:33: Oh we all knew That.

00:11:34: Friday afternoon Slack message from the CFO asking why the AWS bill spiked to four hundred percent, only to find out a developer left a rogue experiment running.

00:11:45: Exactly Garcia's argument is that we have to shift left using AI engines.

00:11:51: We need to forecast the exact costs of a deployment before a single line of code is committed.

00:11:55: To eliminate that end-of-month panic entirely.

00:11:57: Exactly!

00:11:59: I love the idea of predicting spend instantly at the pull request level, but Diana Remos provided a really necessary reality check here.

00:12:06: She noted simply giving engineering teams shiny new dashboard doesn't magically change their behavior.

00:12:12: Why?

00:12:13: What's the blocker?

00:12:14: Because visibility with that agency is just depressing.

00:12:17: That's great way.

00:12:18: put it Right.

00:12:19: If an engineering team looks at a dashboard and sees their service is burning one hundred fifty thousand euros per month but they don't actually have the ownership, sprint time or architectural space to rewrite it that visibility...is useless.

00:12:34: The cost just becomes this terrible number everyone in department knows about But absolutely nobody feels empowered enough to change.

00:12:42: That aligns perfectly with Jay Shankar out to Purokthu's analysis.

00:12:45: He argues that rising cloud bills aren't actually a phenops tooling problem at all, they're governance gap.

00:12:50: A governance gap?

00:12:52: Meaning what exactly?

00:12:52: Meaning when cloud adoption scales faster than your internal decision rates chaos ensues.

00:12:58: You get duplicate data pipelines, you get multiplying orphaned AI experiments.

00:13:03: It reflects a fundamental lack of discipline in the organization.

00:13:06: who actually owns the architectural standards?

00:13:09: Who is the final approver for data duplication across regions?

00:13:13: if those answers are unclear any cost optimization you do as just temporary relief before the bill inflates again.

00:13:19: so how did we bridge that gap between knowing where wasting money and actually fixing the architecture?

00:13:26: Max Gull shared an amazing leap forward his team is making.

00:13:29: Oh, yeah moving away from the slide decks

00:13:31: exactly they're moving away From Fennop slide decks and manual recommendations And moving directly toward AI driven execution.

00:13:39: That's

00:13:39: a massive operational shift it

00:13:41: really Is.

00:13:42: why using tools that understand?

00:13:44: The deep infrastructure context?

00:13:46: the AI isn't just sending an alert saying hey you have waste over here in this cluster.

00:13:50: It is actually generating a context-aware, ready to merge pull request with the exact code needed to optimize the infrastructure instantly.

00:13:58: Wow!

00:13:58: You review it you merge it.

00:14:00: The waste has gone...it reduces the optimization cycle from weeks of interdepartmental debate down to five minute code review.

00:14:07: That's the promise.

00:14:08: automation realized.

00:14:10: But there's one more angle to frame ops that we absolutely have to discuss, and it comes from Tamor Iswal.

00:14:16: He argues that security professionals the CISOs and the SOC analysts must learn how to read The Cloud Bill.

00:14:22: Wait I need push back on.

00:14:24: asking a highly paid cyber-security analyst sit around auditing finance invoices sounds like massive waste of their time.

00:14:32: How does Teymur justify?

00:14:34: Because sudden unexpected billing spike is often the very first indicator of a compromise.

00:14:40: Oh,

00:14:42: oh I see!

00:14:42: Think about it.

00:14:43: things like unauthorized crypto mining or bad actor provisioning massive GPU clusters might actually show up on the financial bill days before they ever trigger behavioral alert in your security dashboard.

00:14:56: That makes dull sense.

00:14:57: The finance team are off from their first to know you've been hacked.

00:15:00: They just don't realize.

00:15:01: that's fascinating way to look at it.

00:15:03: Furthermore Tamer points out great psychological trick for internal politics.

00:15:07: If you frame a security risk purely as a technical misconfiguration or an open port, You might just get a polite nod from leadership.

00:15:14: Right But if you frame that exact same security risk As a fourteen thousand dollar-a month waste of compute That gets you in immediate meeting with the CFO Because

00:15:22: money talks

00:15:23: Money talks, unused orphaned resources are both a financial drain and an unmonitored attack surface.

00:15:30: Cleaning them up serves the finance team and security at that exact same time.

00:15:34: That is brilliant positioning!

00:15:36: Alright let's bring all of this together with our final theme cloud architecture and platform engineering.

00:15:41: because we've just discussed sovereignty laws AI agents writing code runaway costs security risks.

00:15:51: How are architects actually supposed to design these systems today without losing their minds?

00:15:55: It is incredibly easy to get overwhelmed by the sheer volume of new cloud services launching every week.

00:16:02: But Lefteris Kharjorju offered a very comforting realization for platform engineers, he noted that Cloud architecture isn't about memorizing thousands of individual proprietary services.

00:16:12: Thank

00:16:12: goodness because that's impossible!

00:16:14: Exactly most real-world systems are built using a small number of repeatable patterns.

00:16:19: you have your static web architecture your classic three-tier event driven or microservices.

00:16:25: Once you deeply understand those core patterns, You stop acting like a walking service catalog and start thinking Like A True Architect

00:16:32: And Riaz Syed doubled down on that sentiment.

00:16:35: He pointed out That real architects are not blindly loyal to AWS Azure Or Google Cloud.

00:16:40: Right They shouldn't be.

00:16:42: Their job is Not To Be A Fan Boy.

00:16:43: Their Job Is To Evaluate Options And Defend Trade Offs.

00:16:47: Sometimes that means choosing a massive service like Amazon Transcribe because it integrates easily and other times It means defending the choice to use a smaller, highly specialized tool Like DeepGram Because actually solves your specific latency problem better.

00:17:01: But those architectural trade-offs are increasingly colliding with The messy reality of the physical world.

00:17:06: Both Nicholas Fondrini And Acroft Souk brought up A recent Highly sobering incident on LinkedIn.

00:17:14: Yes The AWS Middle East Central One region experienced a serious disruption after fire which was reportedly caused by external objects, specifically drone attacks.

00:17:25: Which is just stark terrifying reminder that the cloud isn't invisible magic floating in sky.

00:17:31: it's physical servers sitting at buildings requiring power and cooling.

00:17:36: When geopolitical tensions escalate in a region and infrastructure is physically targeted, resilience is no longer just about catching software bugs or load balancing.

00:17:45: You can't patch a drone strike.

00:17:47: Multi-region automated failover suddenly becomes a board level risk

00:17:50: discussion.".

00:17:51: Exactly!

00:17:52: Even the most highly redundant beautifully architected hyperscale environments are ultimately anchored to physical territory.

00:17:59: which brings us to really perfect summary of this entire deep dive from Emile Clopigia.

00:18:03: he noted that cloud computing used to be entirely about centralization.

00:18:07: Right, consolidating data centers.

00:18:08: The whole pitch was putting everything in one massive efficient data center.

00:18:12: but now

00:18:13: it is entirely about distribution.

00:18:15: We are distributing for latency so AI agents can be responsive.

00:18:19: we're distributing for resilience against physical threats and cyber attacks, and we were distributing for sovereignty to keep data legally trapped within national borders everything everywhere all at once.

00:18:30: it is a massive structural shift from where we were even five years ago And as we wrap up this analysis let's connect these dots into a final thought.

00:18:40: We've talked about AI agents automatically writing our infrastructure code.

00:18:44: we've talked About Ai tools generating pull requests to optimize Our cloud bills instantly and we've Talked about the necessity of multi-region architectures Automatically failing over during geopolitical crises or physical attacks.

00:18:56: Yeah,

00:18:57: it begs a fascinating question if ai eventually handles The raw coding the cost governance in the operational troubleshooting Does the cloud architect of the twenty thirties stop being a traditional technologist?

00:19:11: Do they instead become sort of digital diplomat.

00:19:14: That's wild thought.

00:19:15: think about it.

00:19:16: if The National Telcos take over local hosting for sovereignty and AI does all that deploying, does the Cloud actually became heavily regulated public utility

00:19:26: like water company?

00:19:27: are we five years away from cloud compute being managed in regulated exactly like the national Water or Electricity grid?

00:19:34: It's entirely possible that tomorrow's architect spends less time in the command line and more time navigating geopolitical borders, compliance

00:19:58: frameworks.