Best of LinkedIn: Cloud Insights CW 49/ 50

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

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

00:00:16: Welcome to the deep dive.

00:00:18: Looking at the sources from the last couple of weeks, it's pretty clear we're at a major inflection point.

00:00:23: We're seeing three huge collisions happening all at once that are, I think, completely rewriting the cloud playbook.

00:00:30: And these aren't small tweaks, are they?

00:00:32: We're talking about sovereignty moving from a talking point to an actual engineering challenge.

00:00:36: Right.

00:00:36: And then there's the AI infrastructure demand, which is just exploding and stressing everything from budgets to power grids.

00:00:42: And finally, that forces a maturity and cost control.

00:00:46: Finops is shifting from a tech checklist to really a company-wide mandate.

00:00:51: Exactly.

00:00:52: So our mission today is to cut through that noise.

00:00:54: We want to give you the strategic shifts you need to understand right now.

00:00:57: This is basically the blueprint for cloud strategy for twenty twenty six and beyond.

00:01:01: OK, let's dive right in then, starting with what feels like the biggest headache for everyone.

00:01:07: Sovereignty.

00:01:08: For years it was just a concept, but now it dictates architecture.

00:01:11: The whole conversation has gotten very technical,

00:01:13: very fast.

00:01:14: That's the key.

00:01:15: All about the engineering now.

00:01:16: Yeah.

00:01:17: We saw... AWS that reinvent, for instance, lay out two totally different models to achieve sovereignty.

00:01:23: This was something people like Jevon George and Renato P were really focused on two

00:01:27: different paths to the same goal, which is what control

00:01:30: control and localization.

00:01:31: Yeah.

00:01:31: The first one is the EU sovereign cloud.

00:01:34: This is like the gold standard isolation model.

00:01:37: Okay, we're building a whole new separate region in Germany.

00:01:40: Full data localization, independent governance.

00:01:43: And this is critical run only by EU based staff.

00:01:47: The gold is maximum separation from US jurisdiction.

00:01:50: But the second model, the one for the UAE, sounds completely different, more like a patch than a rebuild.

00:01:54: Precisely.

00:01:55: The UAE Sovereign Launchpad layers sovereign controls onto existing AWS regions.

00:02:01: It's done through a local partnership with EN Enterprise, so it's a legal and management wrapper over the top of existing infrastructure.

00:02:08: So sovereign T through partnership, not isolation.

00:02:12: It's a much faster way to get there for sure.

00:02:14: But

00:02:14: here's where it gets complicated, and where the online discussion got pretty heated.

00:02:18: The architecture is one thing, but the legal reality is another.

00:02:22: Absolutely.

00:02:23: The law often trumps the physical location.

00:02:26: Marcus Ducey made this point really well.

00:02:28: He basically said, jurisdiction beats location.

00:02:31: Meaning, even if my data is sitting in a server in Frankfurt.

00:02:34: If the provider is US controlled.

00:02:36: U.S.

00:02:36: authorities could still potentially access it.

00:02:39: And

00:02:39: what legal tools are we talking about here, just to be clear for everyone listening?

00:02:42: We're mainly

00:02:42: talking about the Cloud Act and FISA's Seven-O-Two.

00:02:45: These are what give U.S.

00:02:46: authorities that extraterritorial reach.

00:02:49: And that's exactly why someone like Moritz Ott pointed out that AWS's new setup, while a good step, still has to prove its EU fit under really strict EU rules.

00:03:00: Which explains why European companies aren't just waiting around, they're building their own path.

00:03:04: They're building their own path, anchored in open source and collaboration.

00:03:09: You see the strong partnerships like SUSC and Evrock, which Mithias Ulstrom and Dirk Peter Van Luen were talking about.

00:03:16: Right, trying to create an independent stack.

00:03:17: Yeah, a stack built on Linux and Kubernetes so they control their own destiny and can move workloads easily.

00:03:23: And this is already spilling over into the hottest area, which is AI.

00:03:27: We saw SAP launch its own EU AI cloud.

00:03:30: Yes, Martin Mers and Andreas J. Wagner were discussing that.

00:03:34: It's all about providing a trusted, sovereign environment for AI, even bringing in partners like Cohere.

00:03:41: for their models.

00:03:41: And it's not just them.

00:03:43: Dissault systems and Mistral AI are doubling down on running their models on the outscale sovereign cloud.

00:03:49: Tobias Wagner was highlighting that collaboration.

00:03:51: I mean, when you're dealing with sensitive industrial data, you have to.

00:03:55: There's no other way.

00:03:56: The ultimate proof point here has to be the defense sector, right?

00:03:59: Fabian Keenley and Luca Caligari reported NATO is building completely air-gapped, isolated cloud systems with Google.

00:04:06: That says everything.

00:04:07: Zero tolerance for external access.

00:04:10: It's not a surprise when you see that sixty one percent of European CIOs are planning to use more local cloud services precisely because of this geopolitical uncertainty.

00:04:20: It's just smart risk management now.

00:04:22: OK,

00:04:22: so that's where the data lives.

00:04:23: Let's shift to how it's all being powered.

00:04:26: This brings us to the second collision, AI infrastructure.

00:04:30: The demand is just off the charts.

00:04:32: It really is.

00:04:33: Steve D made a great point.

00:04:34: The demand is outstripping capacity.

00:04:36: which increases the risk of major outages, like we've seen recently.

00:04:40: It's

00:04:40: a huge economic problem.

00:04:41: It is.

00:04:42: And Dr.

00:04:42: Monica McTaryon offered this really crucial insight for AI founders.

00:04:46: She said, the core problem isn't just the cost of compute, it's power.

00:04:51: and the long-term lock-in that comes with it.

00:04:53: Hang on, why is power the bigger headache than the immediate cost?

00:04:56: Because running AI at scale, especially for training, doesn't look like the flexible cloud we know.

00:05:02: It looks like old-school high-performance computing, HPC.

00:05:05: So think less paper-drink utility and more like renting a dedicated power plant for three years.

00:05:11: You've got high fixed costs and you need long-term commitments.

00:05:14: That completely goes against the variable on-demand cloud model.

00:05:18: That's a huge distinction.

00:05:20: The whole promise of cloud was elasticity.

00:05:22: But if my GPU usage is constant for months, an on-prem contract suddenly starts to look a lot cheaper.

00:05:29: Exactly.

00:05:29: And that economic reality is what's pushing everyone toward hybrid architectures.

00:05:35: Monica McIterian actually projects, seventy five percent of enterprise AI will be hybrid by twenty twenty eight.

00:05:41: So what does that winning hybrid playbook actually look like in practice?

00:05:46: It's a tiered approach.

00:05:47: You prototype on the public cloud to move fast, but then you move your core steady-state workloads to custom or sovereign infrastructure for cost control.

00:05:55: The public cloud becomes a tactical tool for bursting or spikes.

00:05:58: And the key to making that all work.

00:06:00: Workload portability.

00:06:01: Fred Azar stressed this.

00:06:03: If you can't move your workload, you're trapped.

00:06:04: It's non-negotiable.

00:06:06: which brings us to a really interesting technical point that make Weasmiller flagged.

00:06:10: The hyperscalers are actually making this hybrid world easier to live in.

00:06:13: They have to.

00:06:15: To stay relevant, we're now seeing AWS and Google Cloud connect their platforms with private, dedicated links.

00:06:21: Using things like Direct Connect and CrossCloud Interconnect.

00:06:24: Wait, so

00:06:25: this isn't just running things over the public internet anymore.

00:06:27: This is like a proper engineered.

00:06:31: It's a massive step up.

00:06:33: It lowers the penalty for running distributed systems.

00:06:36: Suddenly running a query in BigQuery against data in S-three isn't a fragile experiment.

00:06:41: It's a reliable low latency pattern.

00:06:44: Multi-cloud is becoming an intentional choice.

00:06:46: Which,

00:06:46: again, reinforces that sovereignty point.

00:06:48: The real power is the ability to switch.

00:06:51: Exactly.

00:06:52: OK.

00:06:52: That leads us perfectly to the third theme.

00:06:55: You have these massive AI bills, complex sovereignty rules, but you still have to master the basics.

00:07:00: Cost control.

00:07:01: Guy Bartram cited a flexor report that cloud waste is still around twenty-seven percent.

00:07:06: Twenty-seven

00:07:06: percent.

00:07:07: That's just unsustainable.

00:07:08: And Richard Vester made the critical point here.

00:07:10: Cloud spend isn't an IT problem.

00:07:12: It's a business problem.

00:07:13: Right.

00:07:14: If you treat it as just a tech issue, costs will always spiral out of control.

00:07:18: And that's the shift.

00:07:20: FineOps isn't a tool.

00:07:21: It's a culture.

00:07:22: It's a cadence that involves finance, engineering, everyone, sharing responsibility.

00:07:27: But the culture needs tools to work.

00:07:30: And Nicholas Fundrini pointed out that Reinvent delivered a lot of new granular cost controls to help with that.

00:07:35: They did.

00:07:36: They're trying to make things more predictable.

00:07:38: We saw new commitment options, like database savings plans.

00:07:41: That's a huge deal because databases are often a source of wild spending swings.

00:07:46: This lets you budget them like you budget compute.

00:07:49: What about storage?

00:07:50: All that cold data just sitting there costing money.

00:07:53: Yep,

00:07:53: they address that too.

00:07:55: S-three has new things like intelligent table tiering and better metadata tools to help you automatically find and move that cold data to cheaper storage.

00:08:04: Stop paying premium prices for data nobody's touched in a year.

00:08:08: And they're even trying to bring that predictability to AI costs.

00:08:11: They are.

00:08:12: Bedrock Reserve capacity is designed to stabilize your unit costs for AI.

00:08:16: When you're dealing with those HPC style fixed costs, having a stable baseline is everything.

00:08:21: But ultimately the best tools in the world don't matter if the discipline isn't there.

00:08:25: That's

00:08:25: the bottom line.

00:08:27: Wu Ming Zhang had a great reflection on this pulling in insights from people like Eric Mark who's focused on just implementing those savings plans.

00:08:34: And William Cooper, who used a great phrase, stop the bleed.

00:08:38: Stop the bleed.

00:08:38: I like that.

00:08:39: It's about hygiene, rigorous forecasting, right sizing, turning off dev environments.

00:08:45: You have to master the basics before you can optimize the complex stuff.

00:08:48: So we've covered three massive shifts, sovereignty, AI infrastructure, and financial discipline.

00:08:55: If you pull it all together, what's the one strategic takeaway for a leader planning for the next few years?

00:09:00: To sum it up.

00:09:01: First, sovereignty is forcing real architectural splits that are legally sound, not just marketing.

00:09:08: Second, AI's power demands are pushing us into a hybrid world for pure economic survival.

00:09:13: And third, FinOps is now the essential business culture.

00:09:16: you need to afford the first two.

00:09:18: It

00:09:18: sounds like a complete break from the lift and shift cloud-first mantra of the last decade.

00:09:23: It is.

00:09:24: I think Guy Bartram's long-term forecast really captures it.

00:09:27: The reality is, Cloud is maturing into a regulated utility and AI is entering its capital intensive industrial phase.

00:09:35: So for the next three to seven years, the winners will be those who prioritize architectural flexibility, economic discipline, and crucially energy aware planning.

00:09:45: It's a decouple your AI and cloud strategies.

00:09:47: They're related, but they're not the same economic problem.

00:09:50: That is the perfect thought to end on.

00:09:53: The focus is shifting from speed to resilience, efficiency and control.

00:09:57: If you enjoyed this episode, new episodes drop every two weeks.

00:10:00: Also check out our other editions on ICT and tech, digital products and services, artificial intelligence, sustainability and green ICT, defense tech and health.

00:10:08: Thank you for diving deep with us.

00:10:10: We'll see you next time.

00:10:11: And don't forget to subscribe.