Best of LinkedIn: Cloud Insights CW 51 - 02

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

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

00:00:15: Welcome to the deep dive.

00:00:17: Our focus today is cutting straight through that, you know, end of year rush and the early year planning frenzy.

00:00:22: We're looking at the top cloud trends that really dominated conversations on LinkedIn.

00:00:27: Yeah, we have a heavy stack of sources from late twenty twenty five early twenty twenty six and The message, I think, is pretty clear.

00:00:34: Which is?

00:00:35: The whole philosophy of cloud adoption has fundamentally shifted.

00:00:38: It's just not about technology adoption anymore.

00:00:40: It's a strategic portfolio design challenge.

00:00:43: Oh,

00:00:43: portfolio challenge.

00:00:44: Exactly.

00:00:45: If you look across all the posts we analyzed, you see three dominant and often contradictory axes driving every major decision right now.

00:00:52: You've got cost, you have control or data sovereignty, and then the massive architectural rewrite being forced by Applied AI.

00:01:00: That tension right there, control.

00:01:02: cost, complexity.

00:01:04: That feels like the story of this period.

00:01:06: So let's unpack that.

00:01:07: Maybe starting with control.

00:01:08: Let's do it.

00:01:09: We're seeing sovereignty and compliance move from being, I guess, abstract regulatory concerns to really hard, non-negotiable architectural constraints.

00:01:19: Precisely.

00:01:20: And Lars Newman, he really hit the nail on the head here.

00:01:24: He emphasized that cloud is now infrastructure.

00:01:27: It's not just IT anymore.

00:01:28: And if it's infrastructure.

00:01:29: Then the legal frameworks around it, like the US Cloud Act, they become strategic geopolitical risk factors.

00:01:35: This is forcing global enterprises to completely rethink who has ultimate authority over their most sensitive data.

00:01:42: And

00:01:43: nothing demonstrates that shift better than the Airbus case study.

00:01:46: I saw that highlighted by Marine Marcus and Scott Newton.

00:01:48: Yes.

00:01:49: This isn't theoretical.

00:01:49: Airbus is actively looking for a truly sovereign EU cloud to move its most critical systems.

00:01:55: And the sheer scope of what they're moving is the real point.

00:01:58: They're looking to shift ERP systems, manufacturing execution systems, PLM, product

00:02:03: lifecycle management.

00:02:05: Right.

00:02:05: And their core aircraft designs.

00:02:07: I mean, Catherine Justin, their EVP digital was very clear.

00:02:10: They need a sovereign cloud to guarantee that information stays entirely under European control.

00:02:15: Wait, moving ERP and PLM, that is not a casual migration.

00:02:19: Those are foundational systems.

00:02:21: Doesn't that just introduce massive complexity?

00:02:24: I mean, doesn't it undermine the whole cost efficiency goal just for political risk mitigation?

00:02:29: That's the core tension right there.

00:02:31: Suddenly, the cost of risk is higher than the cost of complexity.

00:02:35: But the sources also gave us a fantastic counter narrative.

00:02:39: Benjamin Herman, he effectively corrected the public framing.

00:02:43: He argued Airbus' real genius is focusing on cryptographic control, not geographic theater.

00:02:48: I love that phrase.

00:02:49: What does that mean in practice?

00:02:50: Well, Herman noted that Airbus has been using Google Workspace for years for some sensitive workloads, but, and this is the key, they use external Thiele's key management.

00:02:59: Ah, so they hold the keys themselves.

00:03:01: Exactly.

00:03:02: The architecture ensures that Airbus maintains exclusive ownership of the encryption keys.

00:03:07: The keys never touch the hyperscalers infrastructure.

00:03:10: So in this view, sovereignty is secured through zero trust and key management, not just by drawing lines on a map.

00:03:16: But the physical location debate isn't entirely dead, is it?

00:03:20: The US Cloud Act still looms over everything.

00:03:22: It absolutely does.

00:03:23: I mean, Daniel Popescu and Thomas Worth both confirmed that just hosting data in, say, Germany That doesn't provide reliable legal protection.

00:03:32: US authorities can still compel a US based provider to release data regardless of where the server

00:03:38: is.

00:03:38: And Thomas Worth mentioned something else, something that shifts the risk from legal compliance to hard security exposure.

00:03:45: The BMI legal opinion on executive order.

00:03:48: one, two, three, three, three.

00:03:49: That's

00:03:49: the one.

00:03:50: Yeah.

00:03:50: That allegedly authorizes US intelligence agencies to exploit vulnerabilities in foreign infrastructure without the cloud provider even knowing.

00:03:58: That's a fun.

00:03:59: different risk.

00:03:59: You're not worried about a subpoena.

00:04:01: You're worried about a technical intrusion.

00:04:03: So if we can't rely purely on geography, how do architects even deal with this?

00:04:07: Derek Dobson actually synthesized this into three practical models.

00:04:10: You've got the full national stack, highest sovereignty, but it's the slowest and most expensive.

00:04:16: Then you have a middle layer, foreign derived, but sovereign controlled.

00:04:20: And finally, the foreign hyperscaler with local controls like that cryptographic approach, fastest but lowest inherent sovereignty.

00:04:28: The market isn't choosing one.

00:04:30: It's adopting a mixed model based on the workload.

00:04:33: That makes perfect sense.

00:04:34: So for critical infrastructure, they have to pay that premium for control, which brings us, I think, perfectly to the second axis.

00:04:41: cost.

00:04:42: If sovereignty is driving up complexity, Phenops has to be the answer.

00:04:45: It's

00:04:45: mandatory, especially when you see the waste.

00:04:48: A post from Tier Mahara quantified it.

00:04:51: Eighty-four percent of organizations say optimizing cloud costs is their biggest challenge.

00:04:55: Eighty-four

00:04:55: percent.

00:04:56: And the result is that up to thirty percent of their cloud spend is pure waste.

00:04:59: Just gone.

00:05:00: Without proper governance, that's an astronomical amount of money.

00:05:04: And you can't just fix that with a new tool.

00:05:07: Mohammed C. really underscored that fine-ups has to be a cultural foundation.

00:05:10: He defined it through three disciplines.

00:05:12: Visibility, optimization, and governance.

00:05:15: And you can't optimize what you can't see.

00:05:17: Exactly.

00:05:17: If you don't have visibility and clear accountability, you can't even start.

00:05:22: And so much of that waste comes down to... shockingly basic oversights.

00:05:27: Isaiah Michael highlighted that cost overruns are fundamentally an ownership problem.

00:05:32: Meaning forgotten EC two instances, dev databases left running all night underutilized NAT gateways billing you twenty four seven.

00:05:40: It's the simple stuff.

00:05:41: But focusing on things like NAT gateways, that implies engineers are just using platform defaults, right?

00:05:47: Is that just an ownership problem, or is it also a failure of the platform tooling to enforce smarter defaults?

00:05:53: It's a bit of both, but the ownership mindset is critical.

00:05:55: But we also saw great posts on granular engineering optimization.

00:05:59: Rajesh Khan and Rajendraan showed huge wins just by tuning SQL queries.

00:06:04: Things like filter early, join late, using union all instead of union, and strictly avoiding select and production code.

00:06:11: He showed these simple tweaks can cut runtime by fifty percent.

00:06:15: That's a direct cut to your cloud bill.

00:06:17: Fifty percent.

00:06:18: That's significant.

00:06:18: It's a powerful lesson.

00:06:20: And Thomas Staszleski cited the Netflix example.

00:06:23: They migrated self-managed Postgres to Amazon Aurora.

00:06:26: The result.

00:06:27: Seventy-five percent faster performance and a twenty-eight percent cost reduction.

00:06:31: The operational overhead of managing your own database often kills any perceived savings.

00:06:36: So once you've got the quick wins, the next phase has to be automation.

00:06:40: getting away from manual spreadsheets.

00:06:41: Exactly.

00:06:43: Myth and Pandit stressed that future governance has to be policy as code.

00:06:46: AI-powered guardrails embedded right in the CICD pipeline.

00:06:50: Humans shift from policing resources to managing exceptions.

00:06:54: And

00:06:54: the hyperscalers are making this easier.

00:06:56: Peter Kim highlighted a big AWS feature for executives.

00:06:59: the new AWS cost efficiency metric.

00:07:01: Right, a standardized automated score.

00:07:03: It tracks right sizing, idle cleanup, commitment savings.

00:07:06: It's a real KPI.

00:07:08: It moves all those granular decisions out of the weeds and onto an executive dashboard.

00:07:13: A single trackable number.

00:07:15: That's the value.

00:07:16: And for the teams in the trenches, Loic Fournier detailed the new AWS account tag support.

00:07:22: That's a huge simplification for multi-account visibility.

00:07:26: It tags untaggable costs like credits and refunds and streamlined savings plans attribution.

00:07:31: Simple things like cagging can make or break a Phenops

00:07:33: program.

00:07:34: Without those details, visibility is just perpetually broken.

00:07:37: Okay, so we've talked about Phenops, but the sheer complexity of AI workloads is rewriting the optimization rules completely.

00:07:44: Let's pivot to AI, the third axis.

00:07:46: The

00:07:47: structural force.

00:07:48: Yes, we're seeing flagship programs emerge, like the German industrial AI cloud, a massive effort noted by Irwin Shetty and make bomb guard.

00:07:55: And Dr.

00:07:55: Ferry Abelhausen is pushing this as a secure sovereign ecosystem specifically for industrial AI.

00:08:02: This whole AI shift is transforming the cost discussion into what Nicholas Fundrini called inference economics.

00:08:08: Inference economics, I like that.

00:08:10: The idea is that cost control shifts away from infrastructure capacity, you know, VMs, and moves toward real-time decision costs.

00:08:17: The price of tokens, the volume of prompts, the latency, optimization becomes an architectural problem, not just a server-right-sizing problem.

00:08:24: It's a completely new economic model.

00:08:26: Totally.

00:08:27: And generative AI is even changing how we interact with the spend data.

00:08:31: Dibba Jody Monopatra noted that tools like AWS Q and Azure AI agents are enabling conversational fine ops.

00:08:37: So instead of digging through reports,

00:08:39: you just ask the tool, why did our S three spend spike?

00:08:42: Fifteen percent last Tuesday and get an immediate contextual answer.

00:08:46: And if the AI answers the why, then the fine ops professional can shift from data digging to policy enforcement and strategic review.

00:08:53: The machine reports, the human enforces the rules, which

00:08:57: leads to the strategic view on scaling AI.

00:08:59: Dr.

00:09:00: Ferry Abelhausen stated that agentic AI is the key to scaling AI beyond the ninety-five percent of pilot projects that fail.

00:09:06: Yes.

00:09:07: Agentic AI uses orchestrated agents to find, connect, and securely act on data across different systems.

00:09:13: It's how you move AI from an isolated experiment into productive enterprise IT.

00:09:19: Agents give AI the ability to

00:09:21: act.

00:09:21: And this massive new demand for high-speed AI inference is changing the architectural blueprint.

00:09:27: Obena Icedenso summarized it really neatly.

00:09:29: Sovereign cloud for control.

00:09:31: edge for speed, and central regions for scale.

00:09:34: And AI inference for robotics and things like that, it absolutely demands that low latency performance.

00:09:40: you only get at the edge.

00:09:41: We've covered these three, access control, cost, AI, but they all intersect at the core issue of operational maturity.

00:09:47: Right,

00:09:47: and UiClose's principle is the necessary reminder here.

00:09:51: Clever in the cloud isn't about moving VMs, it's about designing for business outcomes from the start, avoiding that lift and shift

00:09:57: trap.

00:09:57: And the fastest path to that clever outcome is standardized deployment.

00:10:01: Jonathan Vell highlighted using Azure Landing Zones with Bicep.

00:10:05: It acts as an accelerator, giving you a secure, governed foundation using infrastructure as code from day one.

00:10:10: Which leads right to security maturity.

00:10:13: Tamer

00:10:14: Ejlal insisted that visibility is the absolute non-negotiable foundation.

00:10:19: Before you worry about alerts or scores, you have to inventory all your resources, find shadow accounts, and define ownership.

00:10:26: You

00:10:26: cannot secure what you cannot inventory.

00:10:28: That's it.

00:10:29: It's a powerful point.

00:10:30: And that inventory is enabled by the foundational platform tooling.

00:10:34: Rami Alcafaje listed the key AWS tools for this.

00:10:38: Security Hub, which is your central dashboard, GuardDuty for threat detection on API calls, and Config, which records resource settings and checks compliance.

00:10:47: Together they give you that necessary inventory and early detection system.

00:10:51: So bringing this all full circle to the engineering team who has to build all this.

00:10:55: What's the required mindset for twenty-twenty-six?

00:10:58: Lefteris Krijorje had some really excellent advice.

00:11:01: He said aspiring cloud engineers need to stop overthinking and just start building.

00:11:04: Learn by doing.

00:11:05: Yes.

00:11:06: Deploy a static website on S-three, then add CloudFront, then add IAC.

00:11:10: Failure in the build process teaches more than any certification ever will.

00:11:14: And he also laid out five essential traits separating good engineers from the exceptional ones.

00:11:19: Yeah, they were relentless curiosity, always asking why.

00:11:22: systems thinking, understanding how the pieces fit, clarity and communication, which is essential, true ownership, you know, caring about production at three AM.

00:11:31: And the last one?

00:11:32: Polymathie, the ability to bridge across security, networking, and the core business strategy.

00:11:39: That last one is crucial for navigating these three complex axes.

00:11:44: So to synthesize everything, the cloud market right now is defined by this massive tension, the pursuit of geopolitical control through sovereignty.

00:11:53: battling against the insatiable demand for algorithmic efficiency from AI fine ops.

00:11:57: And they constantly pull cost and complexity in opposite directions.

00:12:01: So as you map out your architecture for the next few years, maybe consider the concept of optionality.

00:12:06: Guy Bartram suggested that the best roadmaps are designing for twenty twenty eight.

00:12:09: They're prioritizing optionality through data portability and even quantum readiness.

00:12:13: Building that flexibility in now reduces friction and builds leverage later.

00:12:17: So the question for you to mull over is, how much does optionality factor into your current architecture decisions?

00:12:23: If you enjoyed this episode, new episodes drop every two weeks.

00:12:27: Also check out our other editions on ICT and tech, digital products and services, artificial intelligence, sustainability and green ICT, defense tech

00:12:36: and health.

00:12:36: Thank you for diving deep with us.

00:12:38: Be sure to subscribe so you don't miss our next analysis.