Best of LinkedIn: Cloud Insights CW 13/ 14

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

00:00:09: 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.

00:00:20: you can find more info.

00:00:23: Welcome to the deep dive everyone.

00:00:25: Yeah, really glad to be here We've got.

00:00:27: we've got a lot too unpacked for you

00:00:29: today sovereign cloud strategies, the operationalization of AI agents and finally FinOps industrialization.

00:00:57: It's a packed agenda but it all connects really beautifully?

00:01:01: It

00:01:01: does!

00:01:01: And I think before we can even talk about the flashy stuff like advanced AI or you know multi-cloud economics... We really have to look at how tech professionals are fundamentally rethinking cloud architecture and security.

00:01:14: Oh,

00:01:15: absolutely!

00:01:15: I mean there is a massive shift happening right now in how we even conceptualize the Cloud.

00:01:20: Right like For years, learning Azure or AWS was basically an exercise in rote memorization.

00:01:27: Yeah just flashcards of proprietary service names?

00:01:29: Exactly.

00:01:30: but we're seeing top architects aggressively pivot away from that like I was looking at an insight for Rajeeb Chanda and he argues that learning shouldn't be about memorizing services at all.

00:01:40: Right it's about the underlying mechanics.

00:01:42: yeah He calls understanding layers responsibility.

00:01:45: you know identity networking compute data Because if you don't grasp the fundamental physics of those layers, memorizing the exact brand name for a specific load balancer is totally useless.

00:01:57: It really is and Lefterskara Giorgio emphasized this same shift.

00:02:02: He had great posts pointing out that engineers struggle when they fixate on a vendor's specific implementation diagram

00:02:08: Instead of actual system behavior

00:02:10: Precisely Like, if you deeply understand a three-tier web architecture where your presentation application and database layers are isolated or how event driven microservices handle queues then the specific cloud services are just interchangeable parts.

00:02:28: Okay let's unpack this bit because it sounds to me like trying to build a house by just memorizing the names of tools at hardware

00:02:36: store.

00:02:37: That is great way.

00:02:37: put it

00:02:38: Right!

00:02:38: You might know what a hammer And maybe you know what it costs, but that doesn't mean you have any clue how the plumbing needs to route through a load bearing wall.

00:02:48: Exactly and honestly if you don't know How The Plumbing Connects To The Framing You Definitely Don't Know How To Secure The House.

00:02:54: Oh Man Yeah.

00:02:55: Which Is Why This Architectural Mindset is Completely Redefining Cloud Security.

00:03:00: Tymer Ejal Made A Brilliant Observation Recently About This.

00:03:03: What Did He Say?

00:03:04: He Pointed Out That Attackers Don'T Break Individual Services.

00:03:07: They Exploit The Pathways Between Them.

00:03:09: So security is really about how services connect and the attack paths those connections create.

00:03:15: Wait, hold on!

00:03:16: If attackers are exploiting pathways rather than individual services doesn't that mean our entire legacy approach to endpoint security kind of misaligned?

00:03:27: It's massively misalign.

00:03:28: I think most modern breaches happen because a series small seemingly okay configuration decisions across different layers

00:03:35: And they just chain together

00:03:37: they connect to create an unintended attack path.

00:03:40: Security isn't just about locking a single door anymore, it's about mapping how trust flows dynamically between you know thousands of microservices

00:03:49: which perfectly brings in this fascinating conceptual model.

00:03:52: I saw from Daniel Spangenberg oh the

00:03:53: time dimensions one yeah

00:03:55: exactly.

00:03:56: he argues that cloud security actually spans three distinct eras.

00:04:01: You have the past, the present and future.

00:04:03: It's such an effective way to diagnose security gaps.

00:04:06: like The Past is governed by CSPM tools Cloud Security Posture Management.

00:04:10: Right because they just scan the infrastructure To tell you what state your already in.

00:04:14: They look backward at misconfigurations that are already happened

00:04:17: Exactly.

00:04:18: And then the Present Is Your Sime Territory.

00:04:20: You know security information and event management.

00:04:23: But I mean, Simon struggles with modern architectures right?

00:04:26: Because in a microservices environment an event stream is basically just fire hose of noise.

00:04:31: Oh it's incredibly difficult!

00:04:33: Simon was trying to catch the thief while they're walking down the hallway which nearly impossible when millions legitimate automated events are happening at that exact same second

00:04:43: Right Which brings us into future.

00:04:45: Yes The Future Is Policy.

00:04:47: As Code You embed the security rules right into the deployment pipeline.

00:04:55: And I think the ultimate challenge for everyone listening right now is unifying those three time dimensions.

00:05:01: Unifying past, present and future.

00:05:03: so your teams aren't constantly switching mental models.

00:05:06: If you want to know just how urgent that unification problem is You'd have to look at market moves.

00:05:11: Oh!

00:05:11: You mean Google & Wiz thing?

00:05:13: Yeah I mean Google's thirty two billion dollar acquisition of wiz Is entirely about solving this architecture problems.

00:05:19: It's a massive number.

00:05:20: And Canna S and Jaco Silva did some great analysis on this.

00:05:24: They noted that Google isn't just buying another security tool, they're buying cut into cloud context

00:05:30: Right!

00:05:30: Their merging whizzes infrastructure mapping with their own AI threat intelligence Exactly...they

00:05:35: are trying to deliver A unified code-to-cloud run time protection layer.

00:05:41: The truly wild part is there aiming across AWS, Azure and GCP simultaneously.

00:05:47: It's a huge play for the Enterprise Foundation.

00:05:50: Yeah But you know IBM is executing a very similar strategy which Michael Morachnick and Catherine Wong pointed out recently.

00:05:57: Oh

00:05:57: with their recent acquisitions?

00:05:58: yeah

00:05:59: over The last eighteen months IBM has heavily integrated Hashi core for infrastructure automation Red Hat for hybrid cloud and Confluent for real-time data streaming.

00:06:08: They're basically assembling an AI-first enterprise stack for highly regulated environments.

00:06:14: But okay, if we are mapping out these attack paths and were securing the flow of data across all this interconnected environment eventually you hit a physical wall?

00:06:22: You do!

00:06:23: Geography

00:06:23: Right...you have to answer where does this data actually live And who legally has authority?

00:06:29: look at

00:06:30: it Which is exactly why Sovereign Cloud has evolved.

00:06:33: I mean its moved from each compliance topic into central strategic pillar for global enterprises.

00:06:40: Dr.

00:06:40: Sun Peterson actually had a great post about this philosophy, he noted that Sovereign Cloud is moving way beyond just checking a data residency box for an auditor.

00:06:49: Oh

00:06:50: especially in Europe.

00:06:51: Yeah!

00:06:51: For Europe it's literally becoming the path to strategic autonomy.

00:06:56: But you know we have to inject a reality check here which Nish Srebhavan provided recently, because organizations are demanding full sovereignty low cost and high speed all at once.

00:07:07: Which is impossible!

00:07:08: Totally impossible.

00:07:09: Sovereign Cloud has fundamentally about making hard choices and accepting trade-offs.

00:07:13: Wait

00:07:13: wait let me push back on that.

00:07:14: Aren't these hyperscalars promising us we can have it all?

00:07:17: Is Soveregn cloud just becoming a marketing buzzword totally ignores the friction of actual deployment.

00:07:22: That's exactly what Ann Leslie highlighted.

00:07:25: Her perspective on this is crucial.

00:07:28: She says Sovereign Cloud isn't a label.

00:07:30: It's an operating reality that only really gets tested when the system is under severe stress.

00:07:35: Under stress, how so?

00:07:37: Like it's not just about where the servers sit... ...it's about who retains decision authority when a geopolitical shift happens or when a massive operational incident occurs.

00:07:47: Oh!

00:07:47: That makes sense and Martin Merce echoed that deeply.

00:07:49: He pointed out that true sovereignty means understanding your dependencies actionable choices rather than just being locked into a proprietary stack.

00:08:00: And we are seeing the hyperscalers and ecosystem actively responding to this demand, like Oracle.

00:08:05: EU Sovereign Cloud hit their three-year milestone.

00:08:08: Yeah, Michele Hustings and Jason Rees shared some wild stats on that.

00:08:12: Their capacity grew by four hundred percent.

00:08:14: They're now supporting NVIDIA GPU shapes for AI directly in those environments

00:08:19: Which is huge for localized inference.

00:08:21: And then you have Microsoft making moves, too.

00:08:23: Douglas Phillips and Lisa Monaco highlighted how Microsoft and Armada are bringing sovereign AI to completely disconnected edge environments

00:08:32: using Azure local right

00:08:33: exactly so You can run a high-end highly regulated mobile settings with zero public internet connectivity.

00:08:40: But we also cannot ignore the European alternatives.

00:08:43: elk Andrew recently celebrated the ten year anniversary of T systems t cloud Public.

00:08:48: That's a true European sovereign alternative.

00:08:51: And Herod Jues mentioned the Neonephas Foundation, which is doing some incredible work building vendor-neutral open source sovereign cloud infrastructure.

00:08:59: Okay so let us take step back and look at what we have built in this conversation.

00:09:03: You've mapped the dependencies, locked down the architecture built a perfectly compliant sovereign fortress.

00:09:09: It's pretty secure right?

00:09:10: and then you deploy A fleet of autonomous AI agents inside it that absolutely break The entire governance model?

00:09:18: It's the ultimate irony isn't it?

00:09:20: we're moving so fast from experimental AI to production grade agentic workflows...the

00:09:25: speed Of deployment is just staggering.

00:09:28: Renato Ramon de Souza posted this incredible example.

00:09:32: He used Claude code to build a fully functional phenops expert agent inside co-pilot studio,

00:09:37: let me guess it didn't take him months?

00:09:39: It took them ten minutes, ten minutes to prove how fast external coding assistants are accelerating deployment.

00:09:45: Wow and you know Riaz Syod added really great point for IT probes here.

00:09:50: You don't need to build neural networks from scratch using complex calculus.

00:09:54: No not at

00:09:55: all.

00:09:55: You just need enough AI knowledge to deploy and secure models.

00:09:59: If you understand how to securely wire up AWS API Gateway, Lambda in bedrock... ...you are incredibly valuable right

00:10:05: now.".

00:10:07: But there's a serious governance crisis brewing here!

00:10:11: Tristan Ingold raised a huge warning flag about operationalizing these agents.

00:10:15: Yeah the IAM problem.

00:10:16: Exactly current identity and access management stacks were fundamentally built for humans in The Loop

00:10:21: Right.

00:10:22: They assume a biological human is typing a password or clicking the token, but AI agents authenticate up to one hundred and forty eight times more frequently than humans do.

00:10:31: One

00:10:31: hundred and fourty-eight times?

00:10:32: And they delegate!

00:10:33: Agent A spins up agent B to do a task.

00:10:36: so Tristan points out this creates a massive confused deputy liability.

00:10:41: Okay here's where it gets really interesting to me because basically what you're saying Is we are giving The corporate credit card To a fleet of hyperactive robot interns.

00:10:50: Yes

00:10:51: But we're still using human ID badges to track them.

00:10:54: How do we keep this from becoming an absolute disaster?

00:10:57: Well,

00:10:57: the ecosystem is starting to scramble for solutions... Per Gustafson shared some critical insights from Google Cloud's Gemini Enterprise events.

00:11:05: What is the strategy they're pushing?

00:11:06: He says, The solution requires an entirely new strategic blueprint.

00:11:10: We need fine ops for agents to manage token consumption.

00:11:14: we open standards prevent shadow agents and crucially agent-to-agent communication protocols.

00:11:20: They aren't just passing around static human centric tokens Exactly!

00:11:24: They need dynamically verify intent in scope

00:11:26: which seamlessly transitions us to our final theme.

00:11:29: Because speaking of phenops for agents and managing all this consumption, operating complex architectures in sovereign clouds with AI agents requires a massive evolution on how we track cloud

00:11:40: spend.".

00:11:41: We really have to industrialize Phenops!

00:11:43: Nicholas Fundrini introduced the brilliant concept that phenops must evolve from cost visibility to consumption intelligence.

00:11:53: I

00:11:53: love that distinction, because cost-visibility just explains the past.

00:11:56: you look at it dashboard and realize You blew your budget three weeks ago

00:12:00: right?

00:12:00: It does nothing to fix the present but consumption intelligence actually governs The present.

00:12:05: it embeds guardrails into the architecture

00:12:08: And Victor Garcia supported this strongly.

00:12:10: he noted that manual tagging is basically a slow death for phenops.

00:12:14: Oh chasing developers four tags via email

00:12:17: the worst.

00:12:17: The absolute worst, the goal has to be making cloud spend a hundred percent searchable and automated by

00:12:23: default.".

00:12:23: And we're seeing technical guardrails being released to solve these exact problems.

00:12:27: like Shotenaka shared that Snowflake just released new cost control parameters for Cortex-Can?

00:12:33: They added twenty four hour rolling credit consumption limits specifically to stop unexpected AI bills from bankrupting.

00:12:41: That's

00:12:41: essentially an emergency break for AI agents.

00:12:44: Exactly, and on the visibility side Pichimani Rajesh Kumar brought up a highly technical but incredibly impactful update from AWS.

00:12:53: Always saw this one.

00:12:54: S-three access logs right?

00:12:55: Yes

00:12:56: AWS S three server access logs now finally include a source region field,

00:13:00: which is huge.

00:13:01: I mean anyone who has dealt with cross-region data transfer costs knows it used to take hours of manual

00:13:07: correlation.

00:13:07: It was the nightmare.

00:13:08: and Now that one field eliminates all of that manual forensics

00:13:11: its pure consumption intelligence Yeah.

00:13:13: And we're seeing this visibility enhancement at the multi cloud level too.

00:13:17: Yuri Picard co announced Fokie S dashboard one point two.

00:13:21: Fokies

00:13:21: is the open billing standard, right?

00:13:24: And this dashboard automates multi cloud cost data consolidation.

00:13:27: so it brings AWS Azure and GCP building into a single pane of glass automatically.

00:13:32: no more manual normalization.

00:13:34: It's So cool how all these pieces fit together.

00:13:36: I mean we've gone from the architecture bedrock to sovereign borders to autonomous agents and finally, the fine ops layer that pays for it all.

00:13:46: It

00:13:46: really is an interconnected puzzle And you know I want to leave everyone listening with a final thought To mull over building on this entire discussion.

00:13:55: Lay it honest.

00:13:56: So we talked about cloud security needing to span three time dimensions past present in future.

00:14:02: We just discussed how AI agents authenticate hundreds of times faster than humans.

00:14:06: Right!

00:14:07: It raises a very important question For The Industry Are we rapidly approaching a threshold where human governance of the cloud is literally impossible without an AI governing it for us?

00:14:17: Oh, wow.

00:14:18: That's slightly terrifying but probably very accurate thought.

00:14:22: to end on if you enjoyed this episode new episodes drop every two weeks.

00:14:26: Also check out our other editions on ICT and tech digital products & services artificial intelligence sustainability in green ict defense Tech and health.

00:14:35: Thanks so much for joining The Conversation today.

00:14:37: Thank You!

00:14:37: And don't forget to subscribe.

00:14:38: Catch ya next time.