451 Curiosities

In recent years, AI-powered automation has become deeply embedded in IT operations. Orchestrators like n8n make it possible to integrate data from multiple sources, trigger AI-driven decisions, and communicate seamlessly with tools such as Microsoft Teams or ServiceNow.

This opens enormous opportunities for efficiency. But it also raises a fundamental question: what happens if automation fails?

---

The automation paradox

The more efficient a system becomes, the more dependent we grow on it.
What used to be a minor routing error can now have a multiplied impact when it occurs within an AI-driven workflow—potentially affecting dozens of downstream processes.

Automation is no longer a “nice to have.” It is now a critical component of IT infrastructure. And that’s why it needs a solid Business Recovery plan.

---

The specific risks of AI in automation

Integrating AI into workflows introduces new classes of risk:

• Misclassification errors: a model that labels a log or a ticket incorrectly may route it to the wrong team.
• Model unavailability: Ollama services offline, cloud APIs unreachable, or cost and quota limits exceeded.
• External dependencies: if integrations with Teams or a critical database fail, the entire flow may stop.
• Audit and compliance: AI decisions must be explainable; otherwise, they may not withstand regulatory or governance reviews.

---

Key elements of Business Recovery for AI automation

A well-designed continuity plan doesn’t eliminate risk, but ensures that the organization is never paralyzed. Core elements include:

• Regular backups and fast restores: both for workflows and persistent volumes (e.g. n8n_data in Docker).
• Manual failover procedures: documented steps for handling critical processes without AI.
• Error handling in workflows: catch nodes, alerts, and centralized logging.
• High availability: replicated n8n services in Kubernetes or Docker Swarm, clustered databases for persistence.
• Periodic testing: simulated failures (LLM unreachable, APIs down, corrupted databases) to validate the plan.

---

Beyond technology: culture and governance

Business Recovery is not just about technology. It is equally about organizational culture and governance.

• Teams must know how to intervene manually.
• Procedures must be documented and easily accessible.
• AI should be positioned as a supporting tool, not the single point of decision, especially in high-impact areas.

This approach makes automation not only more resilient but also more ethical and transparent.

---

Conclusion

AI automation is not black magic. It’s an integral part of modern IT.
But without a Business Recovery plan, it risks becoming a single point of vulnerability.

Preparing backups, manual fallbacks, failover procedures, and a culture of resilience is what transforms an experimental POC into a production-ready, reliable solution.

And it’s in this maturity step that the real value of the Ethical Web emerges: innovation, yes—but always with responsibility, resilience, and continuity at its core.

AI infrastructure energy is becoming the hidden backbone of the new computational economy — where cooling, power, and grids define scalability.
Artificial Intelligence is not just algorithms, datasets, and GPUs. Instead, it is a thermodynamic phenomenon, a new heavy industry built on silicon, copper, and megawatts.
Every large language model, every generative AI experiment, represents digital entropy that must be contained — and containing entropy requires cooling systems, massive energy generation, and robust grids.

The mainstream narrative often focuses obsessively on NVIDIA’s quarterly results or the latest AI startup valued in billions. However, the true bottleneck is not the availability of GPUs. Rather, it is the capacity of our energy and infrastructure systems to absorb and sustain the demand of AI at scale.

---

From chips to physics: the hidden constraint

An H100 cluster can have a thermal footprint comparable to a steel plant. When you multiply that across hyperscaler campuses in Virginia, Dublin, Singapore, or Frankfurt, the scale becomes clear: AI is no longer an incremental demand on energy, it is a systemic load.

The holy grail of data center efficiency — the Power Usage Effectiveness (PUE) ratio — has improved. Nevertheless, it can never reach the mythical 1.0. Cooling is never free; on the contrary, it always adds overhead, a structural energy debt that grows with every additional watt consumed by GPUs.

This is why focusing exclusively on chips misses the point. Ultimately, the real battlefield of AI is thermodynamics.

---

AI Infrastructure Energy: Cooling & Efficiency as a Silent Multiplier

Heat dissipation is the dark matter of AI infrastructure: invisible in public discourse, yet dominant in physics and cost.

Without liquid cooling, immersion cooling, or direct-to-chip solutions, the density of next-generation AI workloads would collapse under its own heat.

• Vertiv (VRT) is consolidating global leadership with strategic acquisitions (Waylay, Great Lakes Racks) and is even experimenting with partnerships in modular nuclear power.
• nVent Electric (NVT) is less flashy but financially healthier, straddling both cooling and grid infrastructure.
• Asetek (ASTK), a fragile microcap, remains a pioneer of liquid cooling and could be a technological outlier.
• Powell Industries (POWL) stands on the defensive side: debt-free, conservative, but essential in distribution.

Cooling is not an ancillary cost. In fact, it is the physical multiplier that defines whether AI can scale at all.

---

AI Infrastructure Energy: the geopolitical constraint

AI exposes with brutal clarity what energy geopolitics already knew: generation capacity is not a competitive advantage — it is a condition of existence.

• Renewables: Brookfield Renewable (BEP) and NextEra Energy (NEE) guarantee long-term pipelines, though intermittency remains the Achilles’ heel.
• Nuclear: Cameco (CCJ) and emerging Small Modular Reactors (SMRs like Oklo and NuScale) re-emerge as base-load anchors for AI-powered economies.
• Natural Gas: EQT and Petrobras represent the bridge fuel, filling gaps when renewables falter.

AI data centers are not marginal additions to demand. These facilities have become new industrial poles, requiring dedicated, continuous energy capacity. Moreover, utilities serving them do not simply sell kilowatt-hours. Instead, they sell computational uptime — the most valuable commodity of the 21st century.

---

Grid & Transmission: the invisible bottleneck

Energy generated is not automatically energy delivered. Therefore, the weakest link in the chain is the grid, designed for distributed consumption, not for 100+ MW concentrated loads in a single hyperscale campus.

• ABB and Siemens dominate High Voltage Direct Current (HVDC) and smart grid solutions.
• Eaton (ETN) focuses on power management and microgrids, essential for resilience.
• nVent (NVT) bridges cooling with electrical infrastructure.

Without upgraded transmission networks, we face the paradox of stranded capacity: abundant generation that cannot reach the very AI factories that need it.

---

Investing in the chain, not in the mirage

Financial narratives are hypnotized by compute: NVIDIA, SMCI, CRDO. Yet valuations here are stretched, fragile, and exposed to any slowdown in hyperscaler CAPEX.

By contrast, Cooling, Energy, and Grid are less glamorous but more antifragile:

• They generate recurring, infrastructure-like revenues.
• They benefit from secular investment cycles, independent of short-term hype.
• They remain indispensable regardless of which LLM wins tomorrow.

Betting only on GPUs is playing on the surface. By comparison, betting on the AI Infrastructure & Energy chain means investing in the substrate that no algorithm can circumvent.

---

Conclusion: AI as heavy industry

AI is not just math or software. It is industrial thermodynamics shaping the future of global energy and infrastructure.

The fate of generative AI will not be decided solely in Silicon Valley’s labs. Instead, it will also be determined in power plants, cooling systems, and transmission lines.
The real investment thesis is not to chase hype stocks, but to recognize that AI is the new heavy industry of the 21st century.

A recent review by researchers, including Peter A. Noble and Dr. Alex Pozhitkov, explores the concept of a “third state” that exists between life and death, challenging traditional definitions. This state is characterized by the continued functioning of cells from deceased organisms, which can develop into new multicellular forms under specific conditions, such as nutrient availability and biochemical cues. For instance, studies have shown that skin cells from dead frog embryos can reorganize into functioning xenobots, capable of movement and even self-replication. Similar phenomena have been observed in human lung cells forming anthrobots that can repair themselves and nearby neurons. This suggests that cells may not die immediately upon the organism’s death, opening up new avenues for medical applications, such as using these cells for drug delivery systems. The researchers propose that specialized channels in cell membranes may function as electrical circuits, facilitating communication and growth among cells even after death. Their work raises profound questions about the nature of life and death and could lead to significant advancements in regenerative medicine and our understanding of biological processes. The review highlights ongoing investigations into the transcriptional activity of cells post-mortem, revealing that certain mRNA transcripts can increase after death, indicating potential cellular activity. This research aims to uncover deeper insights into the physiological limits of life and redefine concepts surrounding legal death