Ninety-nine per cent across 85,000 orders leaves roughly 850 deliveries that landed outside their window [5][1]. That is the figure a grocery buyer actually prices, because a delivery window is a promise to whoever is staffing the dock, and it is a metric autonomy companies rarely put in a release. Gatik put it in one.
Set the money against the work. Capital raised since the company left stealth in 2019 is about $500m [2]. Contracted revenue is above $600m [4], about $100m more than everything ever raised [6]. In January the chief executive, Gautam Narang, described that as $600m of deals over five years, an average of roughly $120m a year [6][3]. Contracted is not collected. He would not say how much the backlog has grown since, or whether the company is profitable [6]. This round alone is about 40 per cent of all the capital Gatik has taken in [2], which is a lot of outside money for what Narang called "less about building a business" and "more about scaling the business that we have today" [18].
The constraint doing the commercial work here is not distance. Gatik began on fixed routes under ten miles and now runs dynamic routes with dozens of stops covering up to 400 miles, about forty times the original reach [8][7]. What has not changed is the pair of ends: a distribution centre and a store, on a schedule that repeats [7]. The company does not do long-haul, robotaxis or sidewalk delivery, and says so [7]. It sells a lane, not a trip.
The asymmetry inside the fleet is the tell. In Ontario, running for Loblaw, there is still a person behind the wheel [12]. Same third-generation Isuzu box trucks, same Gatik Driver software, same near-continuous operation on surface streets and motorways in light rain and snow [9], different jurisdiction, and the driver stays. Driverless is not a product state the company reached once; it is a permission granted route by route, and pulling the drivers off US commercial routes last year is what turned three metros into capacity [16].
ARK's Tasha Keeney said autonomous trucking should lower the cost of transporting goods [19]. Nothing disclosed this week tests that: no valuation, no per-mile number [3]. What has been demonstrated is narrower and more useful than a forecast. A fleet counted in dozens, hauling ambient, refrigerated and frozen freight for Fortune 50 retailers [13], has run a fixed problem to a schedule 85,000 times [5]. The $600m is still a contract.
