DevOps (2009) — Coexistence: "Collaboration between Dev and Ops." Both professions coexist. The infrastructure engineer is present, visible, named. Their role is recognised. The DevOps movement breaks delivery silos — not competence silos.

Serverless (2014) — Material spoliation: "No more servers to manage." Servers exist — but you no longer touch them. The infrastructure engineer becomes progressively superfluous in the narrative. Their scope shrinks. They remain in the org chart but lose decision-making ground.

NoOps (2016-2018) — Nominal annihilation: "Build, we handle the rest." The profession itself is eliminated from vocabulary. "Linux Administrator" and "Infrastructure Engineer" job postings disappear from recruitment platforms. Replaced by "DevOps Engineer", "Cloud Engineer", "Cloud Architect" — with fundamentally different required skills: AWS certified, Kubernetes, CI/CD. Not Linux. Not networking. Not storage.

Job titles change → required certifications change → universities adapt programmes to job postings → training firms adapt curricula to certifications → in five years, the "neutral infrastructure" skill pool collapses (−34% Linux positions documented in The Newspeak). The labour market was reconfigured by job titles — which were reconfigured by certifications — which were reconfigured by hyperscalers. The complete causal loop that the Algorithmic HR Sabotage (Anatomy of Digital Perdition) documents as an effect. The semantic erasure of job titles is the cause.

"NoOps" provided managerial vocabulary to dismiss people. "Not agile." "Resistant to change." "Legacy profile." These formulas — never written in dismissal letters, always present in boardrooms — transformed rare competence into professional handicap. When a senior system administrator with 20 years of network experience is pushed out in 2019, they are not replaced. The technical memory leaves with them. The organisation orchestrated the programmed loss of its own technical memory. That is the deep explanation for why companies currently cannot reverse course: the people capable of configuring a network switch, reading a pcap, opening a patch panel no longer exist in the org chart.

Operations do not disappear with NoOps — they migrate. AWS, Google and Microsoft employ tens of thousands of infrastructure engineers. Their datacentres have maintenance teams, network engineers, storage specialists, physical security experts. "NoOps" for the client means "MoreOps" for the cloud vendor. This is a competence capitulation: the company transfers operational expertise for its material survival to a third party it cannot audit, cannot control, and cannot read the runbooks of. It outsourced what it no longer understands.

NoOps sells: "We manage operations on your behalf, you are at peace." The Terms say otherwise. In the event of a major hyperscaler outage — an entire AWS region down for 12 hours, a GCP incident impacting thousands of clients simultaneously — the hyperscaler contract provides:

What the hyperscaler reimburses: cloud consumption credits equivalent to a derisory percentage of the monthly invoice (typically 10-30% for a major interruption). Sometimes a few tens of euros on an invoice of several thousands.

What the client bears 100%: the actual operating loss · contractual penalties toward their own clients · data breach and its GDPR consequences · customer trust loss · commercial failure in extreme cases.

This is the legal tour de force of NoOps: externalise the operational act while retaining the entirety of the vital risk. The company surrendered control — but kept the responsibility for economic death.

NoOps promised: less operational complexity. Deploy your code — we handle the rest. The cloud era reality (2018-2026) produces the opposite: an explosion of abstract complexity that neither developers nor retrained former infrastructure engineers fully master.

Service meshes (Istio, Linkerd) for inter-service communication. Hyper-fragmented IAM policies with hundreds of roles, permissions and conditions. Distributed observability (traces, metrics, logs in three different tools). Real-time cost management with alerts nobody monitors. Kubernetes clusters with hundreds of deployments whose origin nobody knows. NoOps did not eliminate the need for Ops — it made Devs responsible for an invisible infrastructure they do not master, without giving them the skills to do so.

Application code is designed to be modified, updated, sometimes discarded. Physical infrastructure follows a lifecycle of maintenance, depreciation and mechanical resilience. NoOps, by making infrastructure seem like an ephemeral script applied with a click, destroyed the notion of Operational Condition Maintenance (MCO).

The result: thousands of cloud resources instantiated by obsolete IaC scripts, abandoned in production because no developer dares delete them for fear of bringing down the application. The Terraform resource instantiating a Load Balancer unused for 18 months? It runs. It bills. It waits. Nobody touches it. The developer who created it has left. The script has no owner. NoOps transformed infrastructure into a graveyard of unaudited scripts — invisible technical debt that nobody accounts for because it looks like code, not infrastructure.

NoOps aligned perfectly with the short-term accounting vision of finance teams. Internal infrastructure teams were fixed cost centres — salaries, employer contributions, training, hardware renewal: visible CAPEX on the balance sheet. Managed cloud services were variable consumption costs — OPEX on the income statement, more flexible, more "agile" in financial vocabulary.

The CFO who validated NoOps to convert CAPEX to OPEX thought they were making a rational decision. The trade-off was accountably defensible short-term. Medium-term, it is disastrous: the "savings" on infra team salaries were far exceeded by uncontrolled cloud invoices, unanticipated exit costs, and the absence of Capacity Planning that would have allowed forecasting them.

The decision to eliminate infra teams was not only organisational. It was financially motivated: a senior system administrator costs €60,000 to €90,000 per year. Eliminating three senior positions = saving €200,000 in payroll. HR has the vocabulary: "legacy profile", "not agile", "resistant to change." Finance has the numbers. The board approves.

What nobody accounted for: the value of the human capital destroyed. The memory of architectures. The knowledge of system dependencies. The diagnostic capability for outages outside the managed perimeter. These assets do not appear on the balance sheet — their destruction neither. Until the outage happens.

Between 2024 and 2026, NoOps found its second wind with AI code copilots. "AI writes the YAML and configures Serverless — you no longer need an infra architect." Generative AI is not intrinsically linked to hyperscalers. But Amazon Q generates AWS code by default. GitHub Copilot (Microsoft) generates Azure code in Azure context. Not AI — algorithmic product placement. AI completes the dissolution of the last traces of hardware consciousness in the application engineer: they no longer understand what they deploy, and now they no longer even write it.

If the loss of the name on job postings preceded the loss of expertise in competence audits — then recovering the name is the first act of resistance. Before the rack. Before the migration. Before the tender.

In org charts: "Infrastructure Engineer", "Systems Engineer", "Systems Administrator", "Network Engineer", "Network Administrator" must reappear as distinct titles — not subsets of a generic DevOps role. An infrastructure engineer is not a developer who operates. A developer is not an infrastructure engineer who codes. Two professions with two temporalities and two incompatible risk types.

In budgets: reintroduce a distinct cost line for infrastructure — not buried in a global DevOps budget. What has no budget line has no existence in the boardroom.

In tenders: require neutral and verifiable skills — Linux Foundation Certified, Red Hat, CompTIA Network+ — not only hyperscaler certifications. The AWS Solutions Architect certification does not replace mastery of physical networking.

Every infrastructure component must be able to answer "yes" to the question: "Can it run outside AWS without a rewrite?" If no — deliberate decision, documented, assumed by both parties. Not a default choice because "it was easy to deploy." This rule is the meeting point of the infrastructure engineer and the developer — it does not say "no" to cloud, it says "conscious" of cloud.