David Sterling: Megan, hey — you see that IBM number this week?
Megan Skiendel: The breach report, yeah — which number stopped you?
David Sterling: Five-point-oh-five million. That's the average cost when a breach spans multiple environments. And 276 days to contain it. I keep reading that back because — well, that's not a security statistic. That's an architecture bill.
Megan Skiendel: It's the cost of indecision, basically.
David Sterling: Right — but here's what actually bothers me. The architecture that's supposed to fix multi-environment exposure — edge computing, keeping data local, off the cloud — that same architecture scatters your data across potentially hundreds of heterogeneous devices you can barely audit consistently. So you solve one problem and you've just... rotated to a different one.
Megan Skiendel: And I don't think that's framed as a security story. Honestly, I think it's an optimization story — two systems built for genuinely different goals that people keep pretending sit on the same dial.
David Sterling: Cloud — AWS first, then Azure and Google Cloud — was built around statistical multiplexing. You pool variable demand from thousands of workloads onto shared infrastructure, spread the fixed capital cost broadly, that's the whole model. Edge inverts it. You're not amortizing anything. You're buying dedicated responsiveness in one place.
Megan Skiendel: So when an enterprise says 'we're doing hybrid' — are they making an architectural decision, or are they just... deferring one?
David Sterling: Deferring one — that's the honest read. But before we get to that, I want to back up one step, because I think the way most people picture these two things is wrong from the start.
David Sterling: Cloud is a shared power grid. Everyone draws from it, the fixed cost gets divided among all users, and pay-as-you-go pricing is just what that pooling looks like on an invoice. Edge is a generator in your own building. You give up the efficiency of the shared grid, but you get power even when the lines go down — and you get it instantly. That's the whole thing.
Megan Skiendel: And the lines going down isn't metaphor — that's the 50 to 200 milliseconds the cloud round-trip actually costs you, versus sub-10 milliseconds at the edge. That gap isn't a tuning parameter. You can't engineer your way out of physics.
David Sterling: Which — and this is the part that stops me — none of this is new. CDNs were doing exactly this in the early 1990s and through the 2000s. Placing servers near users to deliver content faster. That's edge logic. Karim Arabi stood at the IEEE DAC keynote in 2014 and gave it a formal definition — computation outside the cloud at the network's edge — but the infrastructure instinct is thirty years old.
Megan Skiendel: Wait — so the 'new paradigm' framing is actually just... rediscovery?
David Sterling: That's what bothers me. If edge is rediscovered CDN logic — and I think it basically is — then the 'hybrid convergence is sophisticated architecture' story starts to look less like strategy and more like cost discipline catching up to a pattern that was already solved. Cloud got expensive in specific pockets, and organizations noticed.
Megan Skiendel: Honestly — that reframe changes who's driving the decision. It's not a CTO seeing the future. It's a CFO seeing a line item and asking why it's still climbing. And nobody in that room is calling it a thirty-year-old idea.
David Sterling: But the CFO framing only holds until you hit a workload where cost doesn't even enter the calculation. Take a CNC machine on a factory floor — 1,000 sensor readings per second. The machine needs a response in under 10 milliseconds or the line stops. Cloud round-trip is 50 to 200 milliseconds. That's not a cost problem. That's physics disqualifying the option before anyone opens a spreadsheet.
Megan Skiendel: The decision was already made. Before the meeting.
David Sterling: Exactly — and autonomous vehicles are the same structure. Sub-10ms isn't a preference. It's the architectural minimum. Cloud is disqualifying before any cost analysis begins.
Megan Skiendel: And regulation does the same thing in a completely different direction. The hospital CIO I'm thinking of — they didn't choose edge because the latency math compelled it. HIPAA compliance officers told them patient data cannot leave the building. Full stop. Data sovereignty isn't a design preference, it's a structural constraint. Edge is the only compliant architecture.
David Sterling: Which — wait, that actually changes the shape of the decision tree. For those workloads, the 'hybrid versus cloud-first' debate is, I mean — it's irrelevant. The workload already told you the answer.
Megan Skiendel: And then there's a third forcing function people don't name cleanly — connectivity. Remote industrial sites, agricultural sensors, infrastructure in places where cloud connectivity is intermittent. Edge devices keep making local decisions when the network's gone. Accenture has mapped this in factory and agricultural deployments — it's not optional redundancy, it's the baseline requirement.
David Sterling: Right — and the bandwidth pressure compounds it. Edge nodes pre-process, filter locally, transmit only the relevant result. At IoT scale, honestly, you can't push raw sensor volume to cloud — the transfer costs alone close the argument. That's actually the thread back to 1999, when the IoT concept entered supply chain management. The data-volume problem was visible that early.
Megan Skiendel: ETSI and 3GPP saw the same pressure coming and defined MEC — Multi-access Edge Computing — to embed compute right at the cellular network edge for 5G applications. That's not a vendor pitch, that's a standards body formalizing the inevitability.
David Sterling: So strip it back: physics-constrained workloads, sovereignty-constrained workloads, connectivity-constrained workloads — those aren't choosing edge. They're assigned to it. What's left in the middle is the only space where 'hybrid strategy' is actually a decision. And that middle — we'll get to why calling it 'hybrid' doesn't actually resolve the security problem it was supposed to fix.
Megan Skiendel: But that middle space is uncomfortable. Because the IBM number, the $5.05M, 276 days — everyone reads that and thinks 'keep data local, edge solves this.' Except distributed edge means now you've got fifty, maybe a hundred nodes, heterogeneous hardware, variable physical security, inconsistent patch cycles. You haven't reduced your attack surface. You've just... multiplied it into places you can't uniformly audit.
David Sterling: Hold on — that IBM figure applies to poorly architected multi-cloud sprawl. A disciplined centralized deployment doesn't look like that.
Megan Skiendel: Sprawl IS the default outcome. That's not the worst case — that's the realistic baseline. I've watched it happen. You start with three edge nodes for compliance, ops adds two more, a vendor installs one you didn't fully sign off on. The CISO can't see inside them consistently. That's not architecture failure, that's just how organizations actually run.
David Sterling: And there's no TCO framework that captures that. Edge management complexity — the dispersed hardware, the per-device operating costs — there are no standardized benchmarks. Nobody's measured the crossover point where managing that fleet exceeds what amortized cloud infrastructure would have cost.
Megan Skiendel: That's — yeah. That's why hybrid won.
David Sterling: Well, frankly — Accenture maps edge deployment in hospitals, factories, retail as a business transformation driver. Springer surveys call hybrid the target state. DigitalOcean frames it as the practitioner consensus. But none of them quantify what hybrid actually costs to operate. So the 'inevitable convergence' narrative is doing a lot of work without a number underneath it.
Megan Skiendel: And vendors selling both cloud and edge infrastructure — I mean, AWS, Azure, Google Cloud all have edge offerings now — they have a direct economic interest in 'hybrid is inevitable' being the conclusion. That's not cynicism, that's just incentives.
David Sterling: The load-bearing assumption nobody names: that hybrid resolves the tension between utilization-optimization and responsiveness-optimization. It doesn't. It inherits both cost structures simultaneously. That's not a third path — it might be a third operating model with the costs of both and the optimum of neither.
Megan Skiendel: And nobody's measured it. That's the whole problem — no comparable TCO figures exist. So 'hybrid' stays the answer because it's the one nobody can disprove yet.
David Sterling: Cloud was built to optimize utilization across diverse workloads, edge was built to optimize responsiveness in one place. Hybrid doesn't resolve that tension. It just... inherits both, simultaneously, without the TCO math that would tell you whether the tradeoff is worth it. That number doesn't exist. Not at scale.
Megan Skiendel: And I think that's exactly why hybrid won. Not because anyone ran the comparison and it came out ahead. Because nobody could disprove it. You can't lose an argument you haven't had the data to have.
David Sterling: Frankly — that's an uncomfortable place to stop. The decision is being made at scale, across hospitals and factories and retail floors, and the number that would make it falsifiable just... isn't there yet.
Megan Skiendel: Good conversation. Genuinely unsettled by it, which means we did something right.
