Compensation management research identifies precise thresholds. Less than 10% gap between a new hire's salary and that of a 3-year employee in the same role: danger signal. Less than 5% gap: active retention risk. Negative gap — the new hire earns more than the senior: immediate flight of top performers. The cost of a key employee's departure represents 50 to 150% of their annual salary in replacement costs — recruitment, onboarding, productivity loss, tacit knowledge loss. Compression is never a long-term economy.

In France in 2026, a confirmed SRE earns on average €67,500 according to PayScale — versus €43,500 for a DevOps engineer (same source). The €24,000 gap reflects bare-metal competence scarcity. But this gap does not benefit the senior in post — it benefits the new recruit on the external market. The SRE engineer with five years of tenure in a company receiving 3% annual increases rapidly finds themselves below market rate. Their employer pays market rate to recruit their potential successor — without ever having paid this rate to retain them.

Intense competition for rare tech profiles has pushed entry salaries up 8 to 12% per year in the infrastructure sector, while existing employees received 2 to 4% increases. Over five years of accumulation, the gap can exceed 30 to 40%. The engineer recruited in 2021 at €55,000 with 4% annual increases is at €65,000 in 2026. Their potential market successor is offered €75,000 to €80,000. The compression is 15 to 20% — sufficient to trigger flight.

Before the review, there is the form. The self-assessment of 15 to 30 questions on competences, objectives, company values, strategic alignment. The infrastructure engineer who has just managed 12 production incidents, migrated two critical servers, and documented the architecture nobody else knows — this engineer spends an hour filling in a form designed to evaluate commercial profiles. The implicit signal is clear: your work is not measurable in our system. Fill in the form anyway.

The review itself lasts 45 minutes. The manager is well-meaning. The feedback is positive. "You do exceptional work, you are indispensable to the team." Then comes the pay rise: €50 gross per month. €600 per year. A 1.1% increase for an engineer whose external market has grown by 10%. The dissonance between the valorising speech and the derisory pay rise is one of the most documented forms of demotivation — it says two contradictory things simultaneously: "you are valuable" and "your value does not merit real investment".

The organisation does not do the inverse calculation. Replacement cost for this engineer: 50 to 150% of annual salary, i.e. €31,000 to €93,000. Cost of recruiting their successor at market rate: €72,000 to €75,000 salary. Tacit knowledge loss: unmeasured but documented in "The Departure of the Last One Who Knows" as irreplaceable in the short term. Investment in the pay rise that would have avoided all of this: €10,000 per year. Choosing not to revalue is a decision that costs €100,000 to save €10,000.

The engineer resigns. They take 2 to 4 months of gap — either to decompress, to do short freelance or interim missions, or simply to reposition on the market. They then apply — to other companies or sometimes to the same employer — with an "updated" CV and immediate availability. They are recruited at market rate, i.e. €10,000 to €15,000 more than their previous salary. The gap lasted 3 months. They recovered in one negotiation what 10 years of annual increases would not have given them.

The system rewards departure, not loyalty. It rewards breaking a contract, not continuity. It rewards the absence of company knowledge — the new recruit does not know the systems, past incidents, undocumented configurations — more than the presence of that knowledge. And it punishes loyalty — the engineer who stays sees their market value increase while their salary stagnates. This is not an anomaly: it is the signal the system systematically sends to its best people.

Each voluntary gap temporarily removes a rare profile from the market. If 20% of bare-metal SREs take a 3-month gap per year — a plausible figure under severe compression — 5% of the body's total capacity is unavailable at any given moment. In a body already in structural shortage, this chronic unavailability worsens the shortage, drives up entry salaries, and deepens compression for those who stayed. The loop tightens.

A market in shortage of bare-metal competences is a market where organisations cannot build their own sovereign infrastructure — because they lack the engineers to operate it. Dependence on hyperscaler cloud is directly proportional to the shortage of local competences. Hyperscalers did not create the infrational loop — but they benefit from it structurally. Every SRE who takes a gap or recycles themselves as a "cloud architect" is one fewer engineer capable of building outside the cloud.

High turnover in infrastructure profiles is a windfall for specialist tech recruitment agencies. An agency placing a senior SRE receives 15 to 25% of the annual salary — i.e. €10,000 to €18,000 per placement. A market under chronic compression produces chronic turnover that produces chronic placements. This is not a conspiracy — it is an economic interest aligned with the status quo.

The infrastructure engineer who stays in their company is the first victim of the loop. Their salary stagnates while their market progresses. Their tacit knowledge deepens while their nominal value regresses. Their engagement deteriorates — ADP documents that salary compression produces "low motivation, reduced productivity and difficulty retaining staff". They end up leaving or resigning themselves — both are a loss.

OVHcloud, Ecritel, Scaleway, Clever Cloud — the European sovereign actors who depend precisely on bare-metal profiles to operate their infrastructures — are victims of the loop as much as their clients. They compete for a reduced pool at inflated salaries, lose profiles to financially stronger hyperscalers, and cannot absorb the potential demand if a digital sovereignty policy were implemented tomorrow.

Replace the internally budget-based pay review with an annual market benchmark-based review. If the market has grown 8%, the internal increase must at minimum follow — independently of the standard 3% budget. The cost of this policy is a fraction of the turnover cost it prevents. An engineer correctly valued has no reason to take the gap.

For profiles under extreme tension — bare-metal SRE, senior systems administrator, infrastructure architect — introduce an annual or biannual retention bonus, conditional on remaining with the company. This bonus is not a salary — it is an explicit recognition of competence scarcity and its replacement cost. It must be calculated on the differential between the internal salary and the external market.

Self-assessment on generic criteria is unsuited to technical profiles whose value is operational not commercial. Replace it with a review of incidents managed, systems maintained, knowledge transmitted — metrics that correspond to the reality of the infrastructure engineer's work. The generic form is a signal of incomprehension. The operational review is a signal of respect.

Calculate annually the ratio between in-post employees' salaries and the hiring salary for the same role. Publish this ratio internally. Set a minimum gap threshold — for example, a senior's salary must be at minimum 15% above the entry-level hiring salary for the same role. If this threshold is breached, it automatically triggers a correction budget. What is not measured is not corrected.

Lever 5 — Reduce the shortage at source: the levers documented in "The State and the Invisible Body" — Foundation of Iron in BTS programmes, France Travail convention, RNCP protected title — reduce the structural shortage fuelling compression. The more qualified bare-metal engineers there are, the less entry salaries explode, the less violent the compression.

Lever 6 — Updated digital sector collective agreements: digital sector collective agreements have not been revised to account for extreme tension on infrastructure profiles. Collectively negotiated pay scales at sector level, integrating scarcity coefficients for bare-metal competences, would mechanically reduce compression by setting more realistic floors.