Gas pipelines on the floor of the South Caspian have to cope with a set of loads that seldom occur together anywhere else: an aggressive, CO₂- and H₂S-bearing produced fluid, multiphase flow, a steep temperature drop between the wellhead and the seabed, and a tectonically restless basin dotted with mud volcanoes. This paper asks a focused question — for a representative rather than field-measured 70 km line, how do two standard CO₂-corrosion correlations behave, how sensitive is the answer to the main operating variables, and how does the resulting wall loss propagate into mechanical and operational risk? The de Waard–Milliams (1991) and NORSOK M-506 (2017) models are applied along the route using the correct total-pressure form of the fugacity coefficient and NORSOK's tabulated temperature interpolation, and H₂S is handled where it actually matters — as a sour-service and cracking constraint screened against ISO 15156 / NACE MR0175 rather than as a general-corrosion term. Because the inputs are assumptions, the numbers are illustrative of the framework, not predictions for a specific asset. With that caveat, the two models give inlet corrosion rates of roughly 11–16 mm/yr under the hottest, highest-CO₂ conditions, falling below 1 mm/yr at the cold outlet. The more interesting result is that the models do not simply differ by a fixed margin: NORSOK M-506 is far more aggressive than de Waard–Milliams through the 20–60 °C band — by up to three- to fourfold — but the two cross over near the seabed temperature, where NORSOK's temperature constant collapses and de Waard–Milliams becomes the more conservative of the two. A one-at-a-time sensitivity study ranks CO₂ partial pressure, temperature and total pressure as the dominant levers (factors of ×1.5 to ×2.8 over realistic ranges), and a 10,000-sample Monte Carlo propagates the input ranges into full predictive distributions. Wall thinning is then carried into a Timoshenko–Gere collapse check and a DNV-RP-F101 burst assessment, and the operational hazards are consolidated into a 5×5 risk matrix with explicit likelihood and consequence definitions. Internal CO₂ corrosion emerges as the controlling degradation mechanism, and — given the crossover — the safe design choice is the envelope of the two models rather than either one alone.