Best of LinkedIn: Cloud Insights CW 37/ 38

00:00:00: This episode is provided by Thomas Allgaier and Frennis, based on the most relevant LinkedIn posts about cloud in calendar weeks, thirty seven and thirty eight.

00:00:08: For soon, this enables enterprises with market technology and competitive intelligence for portfolio and strategy development.

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

00:00:19: Today, we're really digging into LinkedIn to unpack what's actually happening in the cloud space over the last couple of weeks.

00:00:25: Yeah, the key trends.

00:00:26: Exactly.

00:00:26: Our mission really is to connect the dots between geopolitics, this huge demand for AI, and just the day-to-day operational realities that everyone's facing.

00:00:36: It feels like these forces are really colliding right now.

00:00:38: Oh, absolutely.

00:00:39: It's not just about migrating faster anymore or saving a bit on compute.

00:00:43: We're definitely seeing a major strategic shift.

00:00:46: The whole conversation now seems centered on verifiable control, getting ready for AI, and building infrastructure that's genuinely resilient and, importantly, jurisdictional.

00:00:56: And you can see how public sector needs and these regulated industries are basically forcing providers to rethink their entire ecosystems.

00:01:03: It's fundamental.

00:01:04: OK, so let's kick off with... sovereignty in geopolitics.

00:01:08: That seems to be where a lot of the tension is.

00:01:12: Governments are treating cloud like critical national infrastructure now.

00:01:16: They

00:01:16: are.

00:01:16: But what is sovereignty in this context?

00:01:18: I mean, it's obviously more than just sticking a data center somewhere local, right?

00:01:21: Way more.

00:01:22: And that's the crucial point.

00:01:23: The old definitions just don't cut it.

00:01:26: Andreas Miller pointed this out.

00:01:28: We need a much more sort of granular view.

00:01:31: Think of it like maybe three layers.

00:01:32: You've got data sovereignty, that's the physical location bit, then operational sovereignty, who actually manages the services, who has control.

00:01:40: Right, the who touches it part.

00:01:41: Exactly.

00:01:42: And then the really thorny one, legal sovereignty.

00:01:45: Making sure the whole setup isn't exposed to foreign laws, specifically thinking about the U.S.

00:01:50: Cloud

00:01:51: Act.

00:01:51: Ah, yes.

00:01:52: The Cloud Act.

00:01:53: That tension seemed central.

00:01:54: Andrew Wood and Yuri Fomenoff really highlighted this clash, didn't they?

00:01:59: They did, because the U.S.

00:02:00: law says providers have to hand over data no matter where it sits physically.

00:02:05: Which directly clashes with rules like GDPR in Europe.

00:02:08: Precisely.

00:02:09: It puts global providers in an impossible position.

00:02:12: They either violate US law or they violate European data protection law.

00:02:16: There's often no way to satisfy both.

00:02:18: They're legally cornered.

00:02:19: And this isn't just theoretical legal stuff anymore, is it?

00:02:22: It's actually changing how governments buy cloud services.

00:02:24: It absolutely is.

00:02:26: We saw Joe Knoss detailing this huge shift in Canada.

00:02:29: They've completely rewritten the rules for federal cloud procurement.

00:02:33: Well, to be considered Canadian now, suppliers need fifty one percent Canadian ownership.

00:02:38: Their head office has to be in Canada.

00:02:39: And this is the kicker.

00:02:40: They can't have a foreign parent company that could force data access or expose them to the cloud

00:02:45: act.

00:02:45: Wow.

00:02:46: Okay, that is significant.

00:02:48: That basically draws a line in the sand for the big U.S.

00:02:51: hyperscalers, doesn't it?

00:02:52: It does.

00:02:53: It's procurement policy essentially becoming industrial policy.

00:02:56: Don Spires, Amanda LeClaire, Cole Siorin, they all flagged how serious this push for national digital control is becoming.

00:03:03: Yeah, it's a clear message.

00:03:05: Just being in the country isn't enough anymore.

00:03:07: Control is paramount.

00:03:09: So the big question is how do the global players respond?

00:03:12: They can't just ignore huge markets like Canada or the EU.

00:03:16: Right.

00:03:16: Andrew Woods sort of summarized the pattern we're seeing.

00:03:19: U.S.

00:03:20: vendors, Microsoft, AWS, Oracle, Arcles, they seem to be adapting by... giving up operational control.

00:03:27: Giving it up.

00:03:28: Ow.

00:03:29: They keep the license revenue, the core tech, but they hand over the actual management, the encryption keys, the control plane access to partners based in the EU or, you know, local entities.

00:03:39: Just

00:03:39: to meet the compliance tick box.

00:03:40: Essentially, yes.

00:03:41: It's a way to insulate the operation legally.

00:03:44: But doesn't that create its own risks?

00:03:46: You're basically trusting a third party with the keys, operationally speaking, seems, I don't know, a bit fragile.

00:03:53: It definitely adds complexity.

00:03:54: Yeah.

00:03:54: And that complexity leads to these, well, high control, often high cost solutions for organizations that need absolute assurance.

00:04:01: Like what?

00:04:02: Like SAP's new offering, Michael Herning talked about, sovereign cloud on site.

00:04:06: This takes it to the extreme.

00:04:08: SAP still manages the cloud services, the software layer, but the actual physical infrastructure sits entirely inside the customer's own data center.

00:04:16: So maximum physical and legal control.

00:04:19: Exactly.

00:04:19: You get the SAP innovation, but you don't give up control to anyone externally.

00:04:24: It's the high assurance answer to that cloud act problem, but likely not cheap.

00:04:28: That makes sense.

00:04:29: And it's a good transition because these kinds of specialized controlled environments are probably where a lot of the big AI infrastructure spending is happening now.

00:04:37: AI demand is just exploding.

00:04:39: It really is.

00:04:40: The intelligence phase, as some call it.

00:04:42: Aiman Hussein over at NTT Data noted that combining cloud with what he termed agentic AI is starting to unlock totally new innovation pathways.

00:04:52: Agentic AI, what does that mean exactly?

00:04:54: Is it different from the AI we usually hear about?

00:04:56: It's a step beyond.

00:04:57: maybe, instead of just predicting something or answering a single query.

00:05:00: Agentic AI can kind of... take ownership of a whole multi-step process.

00:05:05: It reasons, it plans, it executes, it course-corrects, often without needing constant human input for every little step.

00:05:12: Let's see, so more autonomous.

00:05:14: Yeah.

00:05:14: He's expecting something like thirty, forty percent productivity boosts from this, especially in complex areas like engineering workflows.

00:05:21: We're seeing this integrated into actual applications.

00:05:23: We are.

00:05:24: Cormac Waters detailed how Oracle Fusion Cloud HCM is embedding AI agents right across the employee life cycle.

00:05:32: So not just suggesting training, but maybe handling the whole internal job matching process or answering complex HR policy questions automatically.

00:05:41: Exactly that kind of thing.

00:05:42: owning the workflow, but you know running these sophisticated AI models takes serious compute power specifically GPUs.

00:05:50: Right, the great GPU scramble.

00:05:51: And critically assured access to those GPUs, often needing to be within specific geographic or sovereign boundaries, which explains these huge investment announcements we've seen often tied directly to government and defense needs.

00:06:04: Like

00:06:04: Oracle in the UK.

00:06:06: That was a big number.

00:06:06: Huge.

00:06:07: Elliott Perkins reported five billion dollars over five years, just focused on expanding AI infrastructure specifically for UK government and defense sectors.

00:06:16: And did NATO also pick Oracle?

00:06:19: They did.

00:06:19: For mission-critical workloads, Jan Rons and Sebastian H. Linnah highlighted that.

00:06:24: It's a massive vote of confidence in OCI security and, well, it's sovereign capabilities.

00:06:29: Then up in the Nordics, there was that Microsoft deal.

00:06:31: Unbelievable scale there.

00:06:32: Sarmar Abilteif and Avan Ericsson were talking about it.

00:06:35: a six point two billion dollar agreement.

00:06:37: Microsoft, Nest Scale, and Acre ASA.

00:06:40: What are they building?

00:06:41: One of the world's largest GPU clusters, apparently, up in Narvik, Norway, and significantly powered entirely by renewable energy.

00:06:50: So hitting the sustainability angle too?

00:06:52: Definitely.

00:06:53: Tying together national sovereignty, strategic AI power, and green energy.

00:06:57: It shows how intertwined these things are becoming.

00:06:59: And T systems is jumping in too.

00:07:00: right in Germany.

00:07:01: Yeah, Dr.

00:07:02: Ferry Abelhessen announced their plan with NVIDIA.

00:07:05: They're calling it an AI gigafactory aiming for the first industrial AI cloud on German soil.

00:07:10: So the message is clear.

00:07:12: Serious AI capability demands nation-level sovereign infrastructure investment, and that costs.

00:07:18: Which brings us right back down to earth, operational reality.

00:07:23: Because while AI demand seems limitless and sovereignty needs are driving up costs, we also saw a couple of major financial shark waves in the post recently.

00:07:33: Okay, let's hear them.

00:07:34: Where's the pain?

00:07:35: Well, first, a potential bombshell for anyone with a multi-cloud strategy involving VMware.

00:07:40: Stephen Kaplan reported on this.

00:07:42: Broadcom's licensing changes for VMware.

00:07:45: They're causing massive cost hikes.

00:07:47: How massive?

00:07:48: Reports are saying up to five hundred percent increases in some cases, specifically for enterprises running Oracle workloads on VMware, particularly within Azure.

00:07:57: Five hundred percent.

00:07:57: Okay, wait, that's that's not an increase.

00:07:59: It's potentially strategy breaking.

00:08:01: Exactly.

00:08:02: Think of all the companies that lifted and shifted Oracle onto VMware in the cloud.

00:08:06: That was their chosen operating model, often for years.

00:08:09: And now the economics are just blown apart.

00:08:11: Completely

00:08:11: destroyed, potentially, by a licensing decision totally outside their control.

00:08:15: It forces a really tough, immediate choice.

00:08:17: You just absorb that massive cost hike, or do you trigger an emergency unplanned migration off VMware?

00:08:23: Right.

00:08:24: People are suddenly scrambling, looking hard at cloud native database platforms like deboss options, just trying to get back to some kind of predictable cost model.

00:08:33: It's a real crisis point for some.

00:08:34: Wow.

00:08:35: Okay, that's one shock.

00:08:36: What was the other?

00:08:37: It relates to cost two, but more about managing it effectively.

00:08:40: Nicholas Fondrini put out a pretty stark warning about fainops.

00:08:45: Oh.

00:08:46: What's the concern?

00:08:47: her view is that fine ops is seriously at risk of making the exact same mistakes that happened with the agile transformation push years ago

00:08:54: meaning Focusing too much on the process not the results.

00:08:58: precisely too many teams are getting caught up in.

00:09:01: You know the vanity metrics like showing off pretty dashboards that claim saving

00:09:05: instead of linking cloud spending to actual business value

00:09:08: exactly like unit economics.

00:09:10: What's the cloud cost per transaction?

00:09:12: or per customer served.

00:09:13: Something that ties directly back to the business outcome.

00:09:16: If you're just optimizing infrastructure costs in isolation, you're kind of missing the point.

00:09:21: So, fine ops needs to be more ingrained culturally, not just a reporting function.

00:09:25: That's the argument.

00:09:27: A true cultural shift, not just ticking boxes on a checklist.

00:09:30: Makes sense.

00:09:31: And alongside cost, there's always resilience.

00:09:33: That's non-negotiable, especially with all this volatility.

00:09:36: Absolutely.

00:09:37: Elliott Layton Woodruff shared some really practical advice.

00:09:39: Learn the hard way after that East US two outage earlier in the year.

00:09:43: Right.

00:09:43: That caused some pain.

00:09:44: What was the takeaway?

00:09:45: Simple, really.

00:09:46: Relying on a single cloud region is just not responsible anymore.

00:09:50: Full stop.

00:09:51: So multi-region is mandatory.

00:09:53: And critically automated multi-region resilience.

00:09:57: His point was you absolutely must use infrastructure as code IAC configured to handle failover automatically.

00:10:04: Manual recovery is too slow, too error prone, especially when compliance is on the

00:10:08: line.

00:10:09: Automation is key.

00:10:10: Okay, and what about security?

00:10:12: Any specific threats highlighted?

00:10:14: Yes, a crucial one around credential theft.

00:10:16: Rami Akifaje raised an alert about the AWS Instance Metadata Service, the IMDS.

00:10:22: Okay, what's the issue there?

00:10:24: Attackers are actively exploiting the older version, IMDSV-One, to basically steal IAM role credentials from EC-II instances.

00:10:32: How do

00:10:32: they do that?

00:10:33: Usually through something called a server-side request forgery, or SSRF attack.

00:10:37: They trick the server into giving up its own internal metadata, which includes temporary security credentials.

00:10:43: And those credentials let them move around inside the environment.

00:10:45: Exactly.

00:10:46: Lateral movement.

00:10:48: The fix, and this is really actionable advice for everyone on AWS, is to migrate from that default IMDS V-one to the newer token-based IMDS V-two.

00:10:58: How does V-two help?

00:10:59: It requires a session token for requests which stops those forgery attempts cold.

00:11:05: It significantly hardens the instance against this type of attack.

00:11:09: Everyone should be looking at this.

00:11:10: Good practical advice.

00:11:12: And this challenge of securing credentials, it only gets harder as things spread out to the edge, right?

00:11:17: Definitely.

00:11:17: Alexander Blanc talked about this, the cloud edge, all those IoT devices, remote systems, low-powered sensors, that's the new really vulnerable attack surface.

00:11:25: Because they're often less managed, less powerful.

00:11:28: And often directly exposed.

00:11:30: Securing that edge, he argued, requires a zero-trust mindset.

00:11:34: network level controls, micro-segmentation, basically assuming every single edge device is untrusted until it proves otherwise through authentication.

00:11:42: The old perimeter security model is dead.

00:11:44: Long gone.

00:11:45: Security has to follow the workload wherever it is.

00:11:48: Okay, so wrapping up the main trends, we also saw continued focus on modernization, especially in core systems like finance.

00:11:54: Yes,

00:11:55: that came through.

00:11:56: A lot.

00:11:56: Kumar and Derek Excel noted the strategic shift away from older finance systems.

00:12:01: They specifically mentioned SAP Business By Design.

00:12:04: users needing to plan, given new customer onboarding is ending, and the move towards SAP Sforhana Public Cloud Finance.

00:12:11: Why the push there?

00:12:12: What does Sforhana Public Cloud offer?

00:12:14: It's really about agility and insight, getting real-time visibility across the business, automating compliance, which is increasingly complex, and embedding AI directly into finance processes.

00:12:24: So moving away from batch processing and historical reporting is something much more dynamic.

00:12:28: Exactly.

00:12:29: Trying to avoid playing, as they put it, yesterday's finance game when the business environment demands real-time decisions.

00:12:36: Okay, that covers a lot of ground.

00:12:38: It really feels like cloud isn't just a tech decision anymore.

00:12:41: Not at all.

00:12:42: Antonio Grasso summed it up nicely, I thought.

00:12:44: Cloud adoption is now fundamentally a strategic business imperative.

00:12:48: It has to be aligned with C-suite goals.

00:12:50: It demands verifiable resilience and continuous upskilling across the organization.

00:12:55: Yeah,

00:12:55: it all ties together.

00:12:56: the geopolitics shapes where you build AI, the vendor policies create sudden cost crises.

00:13:02: It's complex.

00:13:03: It

00:13:03: is indeed.

00:13:04: If you enjoyed this deep dive, new episodes drop every two weeks.

00:13:08: Also check out our other editions on ICT and tech.

00:13:11: digital products and services, artificial intelligence, sustainability and green ICT, defense tech and health.

00:13:17: So maybe a final thought to leave people with.

00:13:19: We're seeing this intense geopolitical pressure demanding real, verifiable operational control.

00:13:25: Phil Dawson mentioned this too.

00:13:27: And at the same time, these unpredictable, massive cost spikes from vendor changes like the Broadcom VMware situation.

00:13:33: Does all this escalate in complexity?

00:13:35: And frankly, the risk involved in managing huge, multi-vendor hybrid cloud estates.

00:13:41: Does it finally push regulated enterprises especially to rethink?

00:13:45: Will they start to narrow their focus?

00:13:47: maybe consolidate onto single purpose built sovereign cloud ecosystems, simply because the multi-cloud dream is becoming too complex and just too costly to govern effectively.

00:13:57: That's a provocative question.

00:13:58: Something to definitely think about.

00:13:59: Indeed.

00:14:00: Thank you for joining us on this deep dive.

00:14:02: Don't forget to subscribe.