Creuto is now an OpenAI Select Partner Read More

AI & Machine Learning

AI route optimization: the per-shipment bill

AI route optimization, priced per shipment: 100k fleet-routed shipments cost $2,970 on Google's global list, $223 on India's. OR-Tools, and when to build.

AI route optimization: the per-shipment bill

AI route optimization is mostly a vehicle-routing solver with prediction and re-planning wrapped around it, and the solver is now priced per shipment. Google's Route Optimization API bills $30 per 1,000 shipments for fleet routing on its global list and $2.40 on its India list. That turns build-versus-buy into arithmetic.

This post is for the head of operations or CTO at a 3PL, distributor or last-mile delivery company running a few dozen to a few hundred vehicles. By the end you will know what the solver layer costs at your shipment volume, why re-planning can multiply that bill, when the free OR-Tools solver is enough, and which layers around the solver are worth building yourself. Prices are as of November 2026.

How AI route optimization works: one solver, three layers around it

The core of any route optimization product is the vehicle routing problem (VRP). Google's OR-Tools documentation defines it as finding "optimal routes for multiple vehicles visiting a set of locations", and its variants add the constraints real fleets have: vehicles with a maximum carrying capacity, and locations that must be visited within a time window. That part is mathematics, and it is now a commodity you can rent or download.

What vendors call "AI" usually sits in three layers around the solver:

  • Prediction. Estimating travel times, service times and demand, so the solver works on realistic inputs rather than straight-line distances.
  • Re-planning triggers. Deciding when the plan is wrong enough to solve again: a new order, a late vehicle, a sensor reading.
  • Dispatch and telemetry. Getting the plan to drivers, tracking them by GPS, and feeding what actually happened back into the first two layers.

The solver is the part you should almost never write yourself. The three layers are where your operation differs from everyone else's, and where custom work pays.

Google route optimization API pricing: billed per shipment, by vehicle count

The Google Route Optimization API bills per shipment. According to Google's usage and billing page, the per-request charge is the number of shipments in the request multiplied by the SKU's unit price, and the SKU is set "only" by the number of vehicles in that request. A request with one vehicle bills to Single Vehicle Routing, a Pro SKU. A request with two or more vehicles bills to Fleet Routing, an Enterprise SKU.

Here are both SKUs on both price lists, per 1,000 billable events a month:

SKU and listFree each monthUp to 100,000100,001-500,000500,001-1M1M-5MAbove 5M
Fleet Routing, global1,000$30.00$14.00$6.00$2.40$2.10
Fleet Routing, India7,000$2.40$2.40$2.40$2.40$2.10
Single Vehicle Routing, global5,000$10.00$4.00$2.00$0.80$0.70
Single Vehicle Routing, India35,000$0.80$0.80$0.80$0.80$0.70

Sources: Google Maps Platform global pricing list and India pricing list, read 20 November 2026. The India list has only two paid tiers: everything above the free cap up to 5 million, and above 5 million.

Cost per shipment: what 100,000 fleet-routed shipments a month costs

A route optimization API costs about 3 cents per fleet-routed shipment at the global list's first tier, and about a quarter of a cent on the India list. At 100,000 fleet-routed shipments a month, after the free caps are deducted, the solver bill is $2,970 on the global list (99,000 × $30 / 1,000) and $223 on the India list (93,000 × $2.40 / 1,000). That is our arithmetic from Google's published tiers, before taxes and any negotiated discount.

Monthly volumeGlobal listIndia list
100,000 shipments, Fleet Routing$2,970$223
100,000 shipments, Single Vehicle Routing$950$52
500,000 shipments, Fleet Routing$8,570$1,183

The 500,000 row on the global list is 99,000 shipments at $30 plus 400,000 at $14. On the India list it is 493,000 at $2.40. Across these rows the global list costs roughly 7 to 18 times the India list, depending on SKU and volume.

Which list your billing account is charged on is a question for Google, not for this post. Google's India page describes itself only as containing "pricing tables for Google Maps Platform APIs for India", and states no eligibility rule on the page we read. Confirm with Google or your Maps reseller before a business case assumes the India column.

Rate limits that shape the integration

The usage page sets 60 queries per minute for both OptimizeTours and BatchOptimizeTours, with up to 100 requests per batch and 100MB per batch. When a project hits the quota, Google says the service stops responding. For a fleet re-planning every few minutes, that is a design constraint on how you group vehicles into requests, not a footnote.

Dynamic route optimization can multiply the bill

Dynamic route optimization means solving again when the world changes, and on a per-shipment price list every solve is billed. Because the charge is shipments-in-the-request times unit price, our reading of Google's billing page is that re-optimizing a 100,000-shipment month four times bills 400,000 events. On the global Fleet Routing list that is $7,170 instead of $2,970. On the India list it is $943 instead of $223. Test this in a sandbox project before you rely on it; the billing page does not discuss re-optimization explicitly.

The same rule gives you a lever. Because the SKU depends only on the vehicle count in each request, a disruption that affects one vehicle can be re-planned as a single-vehicle request, at $10 per 1,000 shipments instead of $30 on the global list. Solve the fleet assignment once in the morning, then re-sequence individual vehicles as events arrive. Deciding which events justify a fleet-wide solve and which justify a one-vehicle fix is exactly the re-planning layer, and it is where an AI model earns its keep.

OR-Tools vs Google route optimization API

OR-Tools is Google's free solver, described in its repository as "Google's software suite for combinatorial optimization" and licensed under Apache 2.0. It solves the same class of problem with capacities and time windows, and it costs nothing per shipment.

The catch is what you supply. The OR-Tools VRP guide states that "all you need to solve the problem is the distance matrix", which in the example is pre-computed, and it suggests the Google Distance Matrix API as one way to get it. So with OR-Tools you own the travel-time matrix, the servers, the solve-time limits and the tuning. With the Route Optimization API, Google owns those, and you pay per shipment.

Is OR-Tools enough for vehicle routing in production?

OR-Tools is enough for production when your volume makes per-shipment fees material and you have an engineer who can own the solver configuration. At 100,000 fleet shipments a month on the India list, $223 a month is less than the engineering time to run OR-Tools well, so the API wins on cost. On the global list at 500,000 shipments, $8,570 a month, or about $103,000 a year, starts to pay for that engineer. Your own travel-time matrix has a cost too, so price it before you switch.

Where the AI actually pays: what we built for Binimise and FlashNow

The solver was never the hard part in our own routing work. For Binimise, a smart city waste-management platform, ultrasonic fill-level sensors report bin capacity to the cloud backend. When a bin reaches 80%, the system flags it for collection and adjusts the day's route plan for the nearest available vehicle. Vehicles report GPS coordinates every 30 seconds, and geofences mark authorized zones, depots and routes.

That dynamic route optimization reduced unnecessary vehicle trips by 25-30% in pilot deployments. Before the platform, collection routes were static and vehicles made trips to bins that were empty. The improvement came from knowing when to re-plan. The 80% threshold, the sensor pipeline over MQTT and AWS IoT Core, and the rule that re-plans around the nearest available vehicle are the parts no routing API sells you.

For FlashNow, a quick commerce platform, the problem was dispatch rather than collection. Structured dispatch assigns delivery partners by proximity to the vendor and by availability, route optimization calculates the path from vendor to delivery address, and the estimated delivery time updates from actual route progress. Here, the routing is one leg per order, and the value is in assignment and ETA.

Both cases depend on location data being right. A solver given a wrong pin produces a confidently wrong route, which is why your address field matters more than your algorithm in delivery work.

Build vs buy route optimization software

The strongest case for staying on a route-planning SaaS is that it ships the whole stack: planner, driver app, proof of delivery and support, already working. Building means owning dispatch software, and a team that has never run one underestimates the edge cases. If your routes are stable and your planner's constraints fit, keep paying for it.

OptionSolver costRight whenWrong choice when
Route-planning SaaSBundled in the subscriptionRoutes are stable and standard constraints fitYour re-planning triggers come from your own sensors or systems
Own dispatch, Google Route Optimization API$223 to $2,970 per 100,000 fleet shipmentsYou need custom dispatch and do not want to run a solverYou re-plan the whole fleet many times a day on the global list
Own dispatch, OR-ToolsFree, plus your matrix and serversHigh volume on the global list, with an engineer to own itVolume is low enough that API fees are below an engineer's time

Building is the wrong choice when the only thing you want to change is the solver. That is the cheapest layer to rent. Building is worth it when your advantage is a trigger the SaaS cannot see, like Binimise's fill-level sensors, or a dispatch rule specific to your network.

Start AI route optimization with the trigger, not the solver

Before you compare quotes, write down three numbers: shipments a month, vehicles per planning request, and how often a day the plan has to change. The first two give you the solver bill on either price list from the tables above. The third tells you whether re-planning will multiply it, and whether the event that triggers it is something only your systems can see.

If the trigger is yours, that is the part to build. Our AI engineering team can scope the re-planning and dispatch layers against your shipment volume, and model the solver bill on the Route Optimization API and on OR-Tools side by side, before you commit to either. A focused MVP typically takes 6-10 weeks.

Frequently asked questions

Google's Route Optimization API costs $30 per 1,000 fleet-routed shipments on its global list up to 100,000 a month, about 3 cents each, and $2.40 per 1,000 on its India list. Single Vehicle Routing is $10 and $0.80 per 1,000. Both lists deduct a free monthly allowance first, as of November 2026.

The Google Route Optimization API bills each request at the number of shipments in it multiplied by the SKU's unit price. The SKU depends only on the vehicle count: one vehicle bills to Single Vehicle Routing, two or more to Fleet Routing. So each re-optimization request is billed on the shipments it contains.

OR-Tools, Google's free Apache 2.0 optimization suite, solves vehicle routing with capacities and time windows, but you supply the distance or travel-time matrix and run the solver yourself. It suits high volumes where per-shipment API fees exceed an engineer's time, and is usually overkill at low India-list volumes.

A logistics company reduces vehicle trips by re-planning routes when real conditions change rather than running fixed routes. In Creuto's Binimise platform, bins flag themselves at 80% fill and the day's plan adjusts for the nearest vehicle, which cut unnecessary vehicle trips by 25-30% in pilot deployments.

Route optimization software is worth building when your advantage is a re-planning trigger or dispatch rule a SaaS planner cannot see, such as sensor data. Keep a SaaS planner when routes are stable and its constraints fit. In either case, rent or download the solver rather than writing one.

Written by

Akash Mohapatra

Akash Mohapatra

Co Founder & Director

19 Nov 2026

·

9 min read

Share

LET'S CONNECT

Connect with Creuto!

Ready to take the first step towards unlocking opportunities, realizing goals, and embracing innovation? We're here and eager to connect.