Drilling rigs and frac spreads (services)

The TAM is built from a few live variables, through two expressions of the same physical base (well capex and frac stages). Each variable carries its freshness seal — what changes often and what barely moves. The verification by routes outside that base is in the cross-check, at the foot.

Robustness check by three routes. The two methods this page publishes both point at the all-in figure; what neither touches is the final step —the 45-60% discount that separates the pure service from the inputs—, and that step is, as the model itself states, our own assumption with no direct source. It is the link that decides the headline figure, so all three routes aim there. (a) Subtracting the complements, and this is the route that bites. The model states that sand, water, chemicals and tubulars are counted as niches of their own and are not added here. Those four have a TAM calculated in this work through physical routes that use neither the cost of the well nor the 45-60%: sand delivered to the well comes from 5-7M t × USD 165/t (825-1,155M), OCTG from ~475 wells × 600 t × USD 1,800/t (450-600M, excluding line pipe, which is surface pipeline and not well capex), and fracturing and drilling chemicals from the stages and the wells (250-430M). Subtracted from the all-in figure, what remains for service plus water is USD 2,815-4,975M/year, midpoint 3,895. Water cannot be isolated —its niche publishes 350-800M/year with hazardous waste inside and with no Argentine tariff per m³—, so it is treated as an unknown and its entire range is swept: even assigning the whole water niche to the cost of the well, which overstates it, the pure service gives 3,095; with water at its midpoint, 3,370; with water at its floor, 3,545. For any plausible value of water, then, the subtraction leaves the pure service between USD 3,095 and 3,545M/year. The published band is 2,500-3,500: the subtraction brackets its upper third and excludes its entire lower half, and the published midpoint —3,000— falls below the whole range. Read against the assumption it set out to attack, the pure service is 54-62% of the all-in figure versus the 45-60% assumed: the independent route validates the upper half of the assumption and leaves its 45% floor unsupported. If this number is off, it is off on the low side. The caveat that makes it honest: the subtraction shares the all-in figure with the published method, so it validates the discount, not the base. (b) The volume counted by company, which is a genuinely foreign route. The stage count by provider for January-April 2026 —Halliburton 4,378, SLB 2,428, Tenaris 1,307, Calfrac 1,149, SPI 452— adds up to 9,714, and it is built by adding companies, not wells. Annualized it gives 29,142 stages against the ~28,000 the model projects: they converge within 4% along two routes that share not a single input. It is the cleanest check this page has, and it holds for volume, not for price. Its limit is the seal: the split by provider is a trade-press count and its seal is probable — what it adds is a foreign route, not a primary source. What that same count leaves strained is the fleet: 29,142 stages spread across the ~13 frac spreads pointed at the basin give 6.1 stages per spread per calendar day, a record pace sustained all year. The reasonable reading is not that the volume is inflated —the company-by-company sum sustains it— but that the count of 13 spreads has gone stale, and the niche itself gives the reason: there were 8 three years ago and Halliburton is adding four Zeus units. (c) The perimeter check, which here changes what the page can promise. The model's two bottom-up methods are not two independent routes, and the arithmetic shows it: 470-500 wells at 50 stages per well are 23,500-25,000 stages, which is the same base as the 23,896 of method B; and the cost per well divided by those 50 stages, taken at 85%, gives USD 238k per stage, inside the 210-270k band that method B uses as its price. Method B is A divided by fifty and multiplied back — an identity, not a verification: they converge because they cannot fail to. That does not touch the number, which stands on (a) and on (b), but it does order what the page can claim. What stands: the volume converges with an independent count within 4%; the pure-service discount —the link with no source— is bracketed by subtracting the four inputs, in the direction that the published midpoint is conservative; and the base of the all-in figure rests on a single method, because the two that are published are one. The gap has the shape of two concrete documents: an Argentine tariff per m³ of frac water —what today prevents isolating the one remaining input, and what keeps the subtraction a band rather than a point— and a day rate or a per-stage tariff from a local contract, which would be the first observed price of the service in the whole chain.

The number rests on a few variables. The formula shows how it moves when each one changes; and each variable carries its freshness seal — how often it is worth revisiting. estim