Colombia's Amazon fiber tender died at the underwriting desk, not the survey

About 11.4 million pesos per household connected, and no more than 62.5 households per route-kilometre. That is the density a bidder was being asked to finance and then keep alive for ten and a half years ^[1][2][3].

The build cost is the smaller half of that arithmetic. The source is explicit that low population density and the absence of physical infrastructure raised both construction cost and the risk attached to operating and maintaining the network ^[10]. On sub-river crossings in the Putumayo and the Amazon ^[5], the recurring expense is the money and elapsed time involved in getting a crew to a fault, not the glass in the ground, and no survey returns that figure. A tender either names the party who absorbs it or it does not close.

The sequencing is the part worth keeping. CONPES 4195 guaranteed resources for implementation, operation and maintenance through 2035, and the Ministry of ICT presented the scheme as one of its principal digital infrastructure initiatives ^[8][9]. A 10.5-year financing term running from that May approval expires at roughly the same moment the funded horizon does ^[5], so there was no operating period on the far side of the debt in which anyone could earn. The fiscal appraisal cleared the money before procurement settled who carried the exposure, and procurement never produced a contract ^[1].

What the record does not say is whether bids arrived. DCD's account attributes the identified risks to the government rather than to responses from the market ^[1], and the difference sets the redesign problem: a ministry that priced its own scheme and flinched is in a different position from one that was quoted a number it could not accept.

One structural point for the redo. The tender carried two products at once: last-mile service in Amazonas and Putumayo, and an international corridor tying the Colombian backbone to Brazil's InfoVias network and to the South American Crossing cable landing at Buenaventura, with alternate Atlantic-to-Pacific routing as the payoff ^[2][6][7]. Those have different customers and different revenue. The corridor's economics do not depend on 62 households per kilometre, and the households do not need a Brazilian handoff. Bundling them produced a single procurement large enough to be called strategic and, on this evidence, too correlated to underwrite.

What the government has to choose now is a deployment model, not a route ^[11]. Everything already drawn stays drawn.