Latinomics
Private memorandum · Nº 15

The bottleneck was never labor.

$110.7B of investment was announced for Mexico in 2023; the new investment that arrived was $4.8B. The binding constraint is management density per line, not labor supply — and an AI-native manufacturing ERP removes it.

$110.7B
Announced FDI · MX 2023 · SE MX
13%
New share of realized · SE MX
13.5%
MFG % of GDP, GT · World Bank
Thesis

The argument

Nearshoring promised. It underdelivered.01 · THE PROMISE

In 2023, Mexico logged 378 investment announcements worth $110.7B. New investment that year was $4.8B — 13% of realized FDI. In 2024 the new share fell to 8.6%. The demand is real. The conversion is not.

USD BN · SE MX 2023
Announced '23New, realized '23
The constraint is the middle layer.02 · THE BOTTLENECK

Inputs are not the problem. Manufacturing already accounts for 15.9% of employment in Guatemala (INE), and industrial power in Guatemala City is the cheapest among Central American capitals at 15.0¢/kWh (CEPAL). The gap is the middle layer: managers are 0.7% of manufacturing employment in Guatemala against 3.5% in Mexico — a fifth of the density in the market that captured nearshoring. Supervision, not labor, is what a new plant cannot hire in weeks.

0.7%
Managers, MFG · Guatemala
3.5%
Managers, MFG · Mexico
ILOSTAT · Managers, share of MFG employment · 2024
13.5%
MFG % of GDP · Guatemala
20.0%
MFG % of GDP · Mexico
24.5%
MFG % of GDP · Vietnam
World Bank WDI
Inference substitutes for scarce management.03 · THE THESIS

An AI-native manufacturing ERP runs scheduling, quality, inventory and exceptions as inference. The scarce asset — an experienced plant manager — supervises many lines instead of one. The constraint clears.

Floor
Orders
Sensors
Operators
Inference layer
Scheduling
Quality
Inventory
Exceptions
Management
One manager, many lines
Capital

Allocate

Institutional readers receive the memoranda series and the desk's data annexes. Circulation is limited; access is reviewed, not sold.

Talent

Build

Operators and engineers work with the desk on the inference layer itself — plant-side, in Guatemala. State your case in one line.

Research

Memoranda

Nº 15 · New

The middle layer

Why nearshoring stalled: $110.7B announced, $4.8B of new investment realized. The management-density evidence, plant by plant — and what an inference layer changes about the arithmetic.

ThesisAUG 2026 · SE MX · INE
Data

Evidence, in numbers

Data as of 02 AUG 2026
Manufacturing value added, share of GDP% OF GDP · WORLD BANK
GT 13.5%MX 20.0%VN 24.5%
Remittances set the floorUSD BN · BANGUAT
'19'20'21'22'23'24
Key indicatorsANNUAL · SOURCES CITED
Indicator20232024ΔSource
Real GDP growth3.5%3.7%+0.2 ppBANGUAT
Manufacturing, % of GDP13.5%WORLD BANK
Manufacturing employment15.9%INE
Industrial electricity, ¢/kWh15.0CEPAL
FDI inflows$1.61B$1.69B+5.1%BANGUAT
Remittances$19.8B$21.5B+8.6%BANGUAT
Inflation, eop4.2%1.7%−2.5 ppINE
GTQ/USD, avg7.837.76−0.9%BANGUAT
Fiscal balance−1.3%−1.0%+0.3 ppMINFIN
The desk

About

Latinomics is an independent research desk in Guatemala City. We work from primary sources — BANGUAT, SECMCA, INE — and from the plant floor: the desk visits the factories, ports, and corridors it writes about.

The research funds and informs one operating position: that an AI-native manufacturing ERP clears the region's management-capacity constraint. We publish what we can defend with a number, and mark what we cannot.

Method
Primary sources: BANGUAT · SECMCA · INE
Fieldwork: Guatemala City · Puerto Quetzal · Northern corridor
Restricted circulation