Latest / Best of LinkedIn: Digital Transformation & Tech / Best of LinkedIn: Cloud Insights CW 25/ 26

Transcript

Best of LinkedIn: Cloud Insights CW 25/ 26 · Best of LinkedIn: Digital Transformation & Tech

• 0:00
  Speaker 0: This episode is provided by Thomas Allgaier and Frannis based on the most relevant LinkedIn posts about cloud in calendar weeks, twenty five and twenty six.
• 0:08
  Speaker 0: Frannes has built a sovereign Cloud Market radar for ICT providers with weekly hot news monthly reports quarterly leadership presentations an AI podcasts four field teams.
• 0:19
  Speaker 0: you can find more info in the description
• 0:21
  Speaker 1: right.
• 0:21
  Speaker 1: so we are jumping into some incredibly wild shifts today.
• 0:25
  Speaker 0: yeah We really are And You know For this deep dive.
• 0:28
  Speaker 0: Our mission is to unpack the biggest cloud trends that have been blowing up across the professional network over the last two weeks.
• 0:34
  Speaker 1: Exactly, we're cutting through the noise trying give you a real shortcut understanding new reality of cloud transformation phenops and obviously AI.
• 0:45
  Speaker 0: Honestly, if you're in the ICT and tech industry right now You need to stay ahead of this.
• 0:49
  Speaker 0: The old assumption that the cloud is just as you know cheap infinite borderless thing.
• 0:53
  Speaker 0: That's completely falling apart.
• 0:55
  Speaker 1: Oh absolutely falling apart.
• 0:56
  Speaker 1: i mean geo politics And the raw economics of AI are basically forcing the cloud back down to earth.
• 1:01
  Speaker 1: Yeah.
• 1:01
  Speaker 0: So let's start with this whole idea of digital sovereignty because it rapidly moves from like a basic regulatory compliance checklist To Just This massive
• 1:12
  Speaker 1: boardroom priority, right?
• 1:14
  Speaker 1: It's all about control and resilience now.
• 1:16
  Speaker 1: And the network has been intensely focused on this stark difference between data residency and data sovereignty
• 1:23
  Speaker 0: which are definitely not The same thing at
• 1:25
  Speaker 1: all.
• 1:26
  Speaker 1: Tima Halonen and Robin Faber have actually been really vocal about This lately.
• 1:29
  Speaker 1: Faber even uses this term Sovereignty washing
• 1:33
  Speaker 0: Laverty washing.
• 1:34
  Speaker 0: okay I love that term.
• 1:35
  Speaker 0: That's basically when A provider tells you your enterprise data is totally safe just because the physical server is, sitting in a data center and Frankfurt or wherever.
• 1:44
  Speaker 1: Yeah exactly which?
• 1:45
  Speaker 1: legally?
• 1:46
  Speaker 1: well it means very little.
• 1:48
  Speaker 1: The core issue they're pointing to is the US Cloud Act.
• 1:50
  Speaker 1: Right Because that legislation says that data access requests follow the nationality of the corporate operator not the physical geography of the server.
• 1:59
  Speaker 1: So if an American parent company gets a court order They are legally forced to produce that data even if it's technically sitting in some European subsidiary's data center.
• 2:09
  Speaker 0: Okay, so basically like... It is renting a safety deposit box at the local German bank right?
• 2:14
  Speaker 0: Right You feel totally secure walking into your neighborhood branch.
• 2:17
  Speaker 0: Yeah But The Bank Manager Sitting In Washington DC actually holds A master key to Your Box and they don't have To tell you when They open it.
• 2:26
  Speaker 1: That is the exact analogy, and it captures the exact risk profile that's keeping executives awake at night.
• 2:32
  Speaker 1: I can imagine.
• 2:32
  Speaker 1: So Devin Tyseter actually analyzed how the market is responding to this.
• 2:36
  Speaker 1: He noted that genuinely sovereign options do exist now like The AWS European Sovereign Cloud.
• 2:43
  Speaker 0: Right because that uses a totally separate EU legal entity doesn't?
• 2:47
  Speaker 1: Exactly It uses a separate entity And relies exclusively on EU resident staff so that structure physically and legally cuts the ties to US jurisdiction.
• 2:58
  Speaker 0: But wait, Tecedra also pointed out the catch there right?
• 3:00
  Speaker 0: Yeah Because birthing that legal fortress around your data.
• 3:04
  Speaker 0: it's not cheap
• 3:05
  Speaker 1: Oh!
• 3:05
  Speaker 1: Not at all.
• 3:05
  Speaker 1: It comes with a premium of around fifteen to twenty five percent.
• 3:08
  Speaker 1: Twenty-five
• 3:09
  Speaker 0: percent That's huge.
• 3:11
  Speaker 0: And that brings up really good pushback from Tim Reigns.
• 3:14
  Speaker 0: He was looking at threat modeling here and asking well our enterprise is just overreacting.
• 3:20
  Speaker 1: How so
• 3:21
  Speaker 0: Well he points this mismatch If you look actual government data requests.
• 3:26
  Speaker 0: They almost overwhelmingly target consumer accounts, you know personal emails chat histories for criminal investigations things like that.
• 3:34
  Speaker 0: oh right Government subpoenas for enterprise cloud infrastructure data are statistically basically an anomaly.
• 3:40
  Speaker 0: so he's asking is all this just extensive compliance theater?
• 3:43
  Speaker 1: That's a really fair question.
• 3:45
  Speaker 1: I mean why blow up your IT budget by a fifth For a theoretical threat?
• 3:50
  Speaker 1: but Uwe Geyer from INOS offered a really good counter perspective to that.
• 3:53
  Speaker 0: okay what was his take.
• 3:54
  Speaker 1: He argues that true sovereignty isn't just about government surveillance.
• 3:57
  Speaker 1: It's about full-stack operational accountability down to the bare metal.
• 4:01
  Speaker 1: he points to standards like ISO, twenty seven thousand one and BSI C five.
• 4:05
  Speaker 0: okay let's unpack those standards a bit.
• 4:07
  Speaker 0: for anyone outside of the audit world what does BSIC Five actually do?
• 4:12
  Speaker 1: So BSIC five is this really rigorous German Government standard?
• 4:16
  Speaker 1: it goes way beyond generic cybersecurity.
• 4:19
  Speaker 1: it dictates the entire chain of custody.
• 4:21
  Speaker 0: Meaning like who can actually touch the hardware?
• 4:23
  Speaker 1: Exactly, Who Can Physically Touch a Server?
• 4:26
  Speaker 1: How The Hardware Supply Chains Are Audited?
• 4:28
  Speaker 1: And Most Importantly Who Is Legally Accountable When An Outage Happens For Highly Regulated Industries Think Banking Or Healthcare?
• 4:36
  Speaker 1: Failing an Audit On Those Operational Controls Can Literally Cost Them Their License To Do Business
• 4:41
  Speaker 0: Right.
• 4:42
  Speaker 0: So It's An Existential Business Requirement.
• 4:44
  Speaker 0: Its Not Just Geopolitical Paranoia.
• 4:46
  Speaker 0: and This Isn't Just A European Issue.
• 4:47
  Speaker 0: Either No
• 4:48
  Speaker 1: definitely not.
• 4:50
  Speaker 0: Robert Machen and Craig McClellan were highlighting what's happening in Canada up there the Hyperscalers, AWS Azure Google.
• 4:58
  Speaker 0: they control eighty five percent of the market.
• 4:59
  Speaker 0: wow eighty-five
• 5:01
  Speaker 1: percent yeah
• 5:03
  Speaker 0: that translates to roughly sixteen billion dollars flowing out of canada every single year.
• 5:08
  Speaker 0: so digital sovereignty is basically becoming an economic retention strategy for entire countries
• 5:13
  Speaker 1: Which is wild, and we're seeing these massive market shifts actively happening right now because of it.
• 5:18
  Speaker 1: Capgemini just became the first SAP Sovereign Cloud Partner for instance.
• 5:22
  Speaker 1: Oh really?
• 5:23
  Speaker 1: Yeah!
• 5:24
  Speaker 1: And ELCA partnered with Microsoft Switzerland to build out a full lifecycle sovereign cloud solution Plus.
• 5:31
  Speaker 1: look at Volkswagen.
• 5:32
  Speaker 1: They are moving aggressively to adopt T-Cloud so they can centralize their sovereign infrastructure across all different car brands.
• 5:40
  Speaker 0: That makes alot sense thing, if enterprises are suddenly willing to pay these massive you know twenty percent premiums just for sovereignty how on earth they managing the overall explosion in cloud
• 5:52
  Speaker 1: costs?
• 5:53
  Speaker 1: Right especially with AI thrown into a mix
• 5:55
  Speaker 0: now.
• 5:55
  Speaker 0: Exactly!
• 5:56
  Speaker 0: Generative AI is eating everyone's budgets alive.
• 5:58
  Speaker 0: so how does Phenops even work?
• 6:01
  Speaker 1: Well, it's funny you ask because Phenops is going through a total overhaul.
• 6:04
  Speaker 1: Ronak Goyle recently shared some takeaways from the Phinops X- Twenty-twenty six event and people were jokingly calling The whole conference Tokenomicon.
• 6:14
  Speaker 0: That is incredible!
• 6:15
  Speaker 0: It is...because
• 6:16
  Speaker 1: tokens are literally their own financial discipline now.
• 6:20
  Speaker 1: Genii models caused this massive token panic this year.
• 6:23
  Speaker 1: organizations we're hitting there full your cloud forecast by June.
• 6:29
  Speaker 0: So basically, Runaway AI agents are the new Runaways.
• 6:32
  Speaker 0: Lambda functions.
• 6:33
  Speaker 1: Pretty much!
• 6:34
  Speaker 1: Which is why Urielia whose recent breakdown was so brilliant.
• 6:37
  Speaker 1: he mapped traditional phenops directly to this new GNAI world.
• 6:42
  Speaker 1: The matrix has just completely shifted.
• 6:44
  Speaker 0: Okay break that down for me How does an old playbook apply to tokens?
• 6:47
  Speaker 1: Well think about traditional compute like CPU and RAM You used to optimize by rightsizing your virtual machines Right...so
• 6:55
  Speaker 0: you aren't paying for idle capacity.
• 6:56
  Speaker 1: Exactly Well, today Compute is input or prompt tokens so engineers have to right size their payloads.
• 7:02
  Speaker 1: Instead of sending these huge bloated system instructions you have to tightly trim the context window.
• 7:07
  Speaker 1: So that
• 7:09
  Speaker 0: makes total sense.
• 7:09
  Speaker 0: Okay what about storage?
• 7:11
  Speaker 0: Usually you'd move cold data to a cheaper storage tier
• 7:14
  Speaker 1: Right.
• 7:14
  Speaker 1: well in AI Storage translates into I put tokens and model tearing.
• 7:19
  Speaker 1: So you shift simple routine tasks to small language models, SLMs instead of running everything through your expensive premium tier LLM.
• 7:28
  Speaker 0: Oh I see!
• 7:29
  Speaker 0: You tear the intelligence just like you used it to your hard drives?
• 7:31
  Speaker 0: Exactly Okay.
• 7:32
  Speaker 0: so that leaves the network layer...I think he mapped that to prompt caching right.
• 7:36
  Speaker 0: let's definitely explain that because i found it fascinating.
• 7:38
  Speaker 1: Yeah It is a huge deal and bypasses this massive redundant compute tax.
• 7:43
  Speaker 0: Think about it as student taking an open book exam in the old AI model Every single time you ask a question about, say your five hundred page document.
• 7:52
  Speaker 0: The AI reads the entire book from chapter one answers and then closes the book
• 7:57
  Speaker 1: right which takes forever.
• 7:58
  Speaker 1: yeah
• 7:59
  Speaker 0: And when you asked to follow-up of Question.
• 8:01
  Speaker 0: it has to read the whole Five Hundred Page Book all over again.
• 8:04
  Speaker 0: You are paying for that reading Time on every single query
• 8:07
  Speaker 1: Which is just burning cash
• 8:08
  Speaker 0: exactly.
• 8:09
  Speaker 0: but prompt caching means?
• 8:11
  Speaker 0: The AI keeps the book open On the desk.
• 8:13
  Speaker 0: It retains the context so it instantly drops the latency in the cost.
• 8:17
  Speaker 1: It's a great analogy, but here is the catch.
• 8:20
  Speaker 1: Knowing how to optimize tokens was only part of that problem.
• 8:24
  Speaker 1: Victor Garcia had this warning that really stood out – he noted.
• 8:28
  Speaker 1: data platforms are now eating up to fifty percent total cloud bills.
• 8:32
  Speaker 1: Wait!
• 8:32
  Speaker 0: Fifty percent?
• 8:33
  Speaker 0: Just the data platform?
• 8:34
  Speaker 0: Yes.
• 8:35
  Speaker 1: Largely because of undetected super inefficient queries….
• 8:39
  Speaker 1: you know….a bad query scanning a massive database constantly... As Yorda Nova pointed out, just handing a developer spreadsheet of expensive queries doesn't
• 8:49
  Speaker 0: fix anything.
• 8:50
  Speaker 0: Right because data alone does not actually change developer behavior.
• 8:53
  Speaker 1: Exactly!
• 8:54
  Speaker 1: Phenops is eighty percent culture.
• 8:56
  Speaker 0: So how do you fix that culture problem?
• 8:58
  Speaker 0: You can't just yell at engineers to code cheaper.
• 9:00
  Speaker 1: Well, Yordanova suggests shifting to unit economics.
• 9:03
  Speaker 1: so instead of looking at raw spend and look for cost per active user... ...you also gamify the optimization with leaderboards.
• 9:10
  Speaker 1: Make
• 9:10
  Speaker 0: it competitive.
• 9:11
  Speaker 1: But Benjamin Lenaris makes a great point that the ultimate fix is preventative governance.
• 9:16
  Speaker 1: You have to catch the costs before the infrastructure is even provisioned.
• 9:20
  Speaker 0: Catch it in the pipeline!
• 9:22
  Speaker 0: And actually, Loic Fournier shared an update that AWS has finally building tools for this.
• 9:26
  Speaker 0: They rolled out these new AI-powered cost investigations and six new idle resource recommendations.
• 9:33
  Speaker 0: Oh really?
• 9:34
  Speaker 0: Yes specifically for complex stuff like DynamoDB and ElastiCache.
• 9:38
  Speaker 0: so it finds those toxic queries automatically.
• 9:40
  Speaker 1: That is desperately needed, because if you combine all these massive cost pressures with the strict sovereignty requirements we talked about earlier companies are completely rethinking where their workloads actually live.
• 9:53
  Speaker 0: Oh absolutely in fact We're seeing a major return to bare metal hardware.
• 9:58
  Speaker 0: Day-to-day had some fascinating insights on this.
• 10:00
  Speaker 0: looking at the iGaming industry
• 10:02
  Speaker 1: Right that online casinos and gaming platforms Yeah
• 10:05
  Speaker 0: Their core engines run twenty four seven at crazy high utilization.
• 10:09
  Speaker 0: So paying premium public cloud prices by the minute just stops making sense.
• 10:13
  Speaker 0: They're moving those core engines back to bare metal for predictable latency and fixed costs.
• 10:18
  Speaker 1: It's like deciding between owning a car or using a rideshare.
• 10:21
  Speaker 0: Exactly ride shares are great for unpredictable occasional trips.
• 10:25
  Speaker 0: You only pay when you use it, but if you are driving a hundred miles every single day Using a rideshare will bankrupt.
• 10:32
  Speaker 0: you need to own the hardware.
• 10:33
  Speaker 1: That is a perfect way to look at it.
• 10:35
  Speaker 1: And so they keep the elastic workloads in the cloud, but the heavy lifting comes home and Peter Fletcher discussed this recently too regarding that private Cloud Outlook.
• 10:44
  Speaker 1: twenty-twenty six report.
• 10:45
  Speaker 1: The big takeaway Is that AI is pushing enterprises heavily toward private clouds for production workloads
• 10:51
  Speaker 0: because of the cost predictability right?
• 10:53
  Speaker 1: Costs predictability and data sovereignty plus.
• 10:56
  Speaker 1: Juan Carlos Florez noted This massive shift towards running on device small language models SLMs.
• 11:03
  Speaker 0: Right, pushing the compute all the way to edge?
• 11:04
  Speaker 1: Exactly!
• 11:05
  Speaker 1: Running an SLM directly on a user's phone offers zero latency total privacy since data never leaves the device and it drastically lowers central computing costs.
• 11:15
  Speaker 0: It is incredible And vendor ecosystem scrambling to adapt this new hybrid reality.
• 11:21
  Speaker 0: David Tropiano announced that IBM Cloud just launched Red Hat OpenShift Virtualization Service.
• 11:26
  Speaker 1: Okay, what does that do?
• 11:27
  Speaker 1: It's
• 11:28
  Speaker 0: specifically for managing legacy VM workloads via Kubernetes.
• 11:31
  Speaker 0: So it bridges the gap and Herald Juice noted that German companies are basically abandoning the pure public cloud idea entirely.
• 11:39
  Speaker 0: they're heavily favoring complex hybrid and multi-cloud setups now.
• 11:43
  Speaker 1: But wait!
• 11:43
  Speaker 1: That brings up a huge issue if an enterprise is running a sovereign Cloud in Europe.
• 11:48
  Speaker 1: bare metal for core apps public cloud for elastic traffic and local SLMs on edge devices.
• 11:54
  Speaker 1: It's a massive amount of complexity, how on earth does human IT team manage that without just completely burning out?
• 12:01
  Speaker 0: That is the million dollar question!
• 12:02
  Speaker 0: And they don't at least not alone.
• 12:06
  Speaker 0: this what driving us into the agentic era of cloud operations.
• 12:09
  Speaker 1: The agentic Era?
• 12:10
  Speaker 1: tell me more about it.
• 12:11
  Speaker 0: Humsichish D&J Shanth shared some really wild insights regarding Microsoft new Azure SRE Agent co-pilot observability agent, we are shifting from assisted operations to egentic operations.
• 12:25
  Speaker 1: So wait what's the difference between assisted and egentick in this context?
• 12:28
  Speaker 0: So assisted is like having a chatbot A human asks it to pull logs And The Human decides What To Do.
• 12:35
  Speaker 0: Egentik means that AI acts autonomously.
• 12:38
  Speaker 0: It diagnoses the crash Dynamically writes query code to test its own hypothesis And executes workflows all by itself.
• 12:45
  Speaker 0: Wow They're seeing cut root cause analysis down to under three minutes.
• 12:50
  Speaker 0: Under
• 12:50
  Speaker 1: Three Minutes?
• 12:51
  Speaker 1: That's industry changing!
• 12:52
  Speaker 0: It really is, but there's a catch which Jaroslav Panschoha observed at the recent Google Cloud Summit.
• 13:00
  Speaker 0: He pointed out that while tech has essentially solved the build side of AI The enterprises themselves are totally overwhelmed.
• 13:07
  Speaker 0: There's this huge delivery gap.
• 13:08
  Speaker 0: Value is just piling up at the door because companies lack the internal operations and security frameworks to actually run these autonomous agents safely at scale.
• 13:17
  Speaker 1: Oh, I can totally see that Security teams must be terrified of an agent writing in executing its own code in production.
• 13:22
  Speaker 0: exactly It's a operating model problem
• 13:23
  Speaker 1: which really ties back perfectly to a philosophy George K Matthew shared recently.
• 13:28
  Speaker 1: he said And i love this quote technology creates possibilities people create outcomes, and leadership connects the two.
• 13:36
  Speaker 0: Oh that is so stun on!
• 13:38
  Speaker 1: Right... The future of cloud isn't just about infrastructure or picking the right vendor anymore.
• 13:42
  Speaker 1: it's about decision-making It's about setting up that preventative governance.
• 13:46
  Speaker 1: So these autonomous tools can actually be deployed safely
• 13:49
  Speaker 0: because if you don't have the leadership to connect a technology To the outcome You just have a bunch of really expensive complicated Tools.
• 13:56
  Speaker 1: exactly its a leadership challenge above all
• 13:59
  Speaker 0: else.
• 13:59
  Speaker 0: Well, as we wrap up I want to leave you with one final slightly provocative thought to mull over.
• 14:05
  Speaker 0: It builds on an idea David Eves presented where he talked about the cloud eventually becoming a utility kind of like The Power Grid.
• 14:13
  Speaker 1: okay i like Where this is going.
• 14:14
  Speaker 0: so combine that Utility concept With the rise Of the autonomous AI Agents.
• 14:18
  Speaker 0: We just Talked About What if the Future of Cloud Architecture Isn't About Choosing A Provider at all?
• 14:24
  Speaker 1: what do You mean?
• 14:25
  Speaker 0: imagine in near future where your autonomous, agentic AI acts as your company's energy grid broker.
• 14:32
  Speaker 0: It dynamically shifts your workloads minute by minute across sovereign data centers private bare metal and public hyperscalars.
• 14:39
  Speaker 1: And it does all this routing autonomously Completely
• 14:42
  Speaker 0: autonomously based entirely on real-time token pricing legal compliance flags and latency needs.
• 14:49
  Speaker 0: You never pick a vendor again.
• 14:50
  Speaker 0: Your AI just plays the grid for you.
• 14:52
  Speaker 1: That is wow, that's a wild but totally plausible future to think about.
• 14:57
  Speaker 0: Right it changes everything.
• 14:59
  Speaker 0: well if you enjoy this episode new episodes drop every two weeks.
• 15:02
  Speaker 0: also check out our other editions on ICT and tech digital products and services artificial intelligence sustainability in green ict defense tech and health.
• 15:11
  Speaker 1: thank you so much for tuning in.
• 15:12
  Speaker 1: don't forget to subscribe.
• 15:13
  Speaker 1: see y'all
• 15:15
  Speaker 0: next deep dive.