Starship Technologies is a robotics company with a program of developing robots that will deliver takeaway meals etc. using small robots such is this:

small white robots, not even knee high

One particular press release on their website makes this claim:

Starship’s robots are powered by zero carbon electricity, with an average delivery for a Starship robot consuming as little energy as boiling a kettle to make just one cup of tea. Orders are made through the Starship food delivery app, which is available for download on iOS and Android, with groceries picked fresh in local Co-op stores and delivered quickly and conveniently in as little as one hour or less.

I can't find any justification for this claim, which is being repeated around the media.

Is the claim about energy usage true? One assumes that the 'average' delivery isn't a test run of a few metres.

  • Please use the answer box for answers and the comment section to improve the questions.
    – Oddthinking
    Commented Sep 30, 2022 at 21:38

3 Answers 3


tl;dr: The claim is nearly true. Based on publicly available data, the robot uses the energy required to heat water for one cup of tea every 27 minutes. This is probably a bit less than the time needed for an "average delivery", but it's feasible.

Starship robot metrics

Back in 2018, Starship partnered with Swiss Post to conduct pilot testing of the Starship robot in Dübendorf. During this test, Swiss Post published a fact sheet about the robot including the following metrics:

  • Vehicle weight: 23 kg
  • Max payload: 10 kg
  • Average service speed: 3 km/h
  • Battery: 8,000 mAh, 18.5 V
  • Range: 6 km / 2 hours
  • Charging time: 45 min (0.75 hours)
  • Charge power consumption: up to 250 W

Hourly energy consumption

Based on the battery capacity and voltage we can calculate that the energy consumption is 148 Wh over it's entire 2 hour range:

8 Ah * 18.5 V = 148 Wh

But this ignores charging efficiency. Per the latest US and EU charger efficiency requirements, an 18.5 V charger up to 250 W must have an efficiency of at least 88%, giving a total consumption of about 168 Wh:

148 Wh / 0.88 = 168.18 Wh

How many cups of tea is that?

Thanks to Schwern's answer, we know that it takes 37.5 Wh to boil water for one cup of tea.

If that's the unit of measure for an "average delivery", that means the robot should be able to complete about 4.5 deliveries on a single battery charge:

168.18 Wh / 37.5 Wh = 4.48 deliveries

Since a charge lasts for two hours, that gives an approximate time of 27 minutes per delivery, covering 1.3 km:

120 minutes / 4.48 deliveries = 26.76 min/delivery
  6 km      / 4.48 deliveries =  1.34  km/delivery

Is 27 minutes reasonable for an "average delivery" by a Starship robot?

The press release doesn't define "average delivery" -- it just says that packages are delivered "in minutes". But there are some clues elsewhere:

So, let's say the range is 20 to 45 minutes. But that includes the transaction and packing time, during which the robot isn't using any energy -- so the actual driving time is probably closer to 10 to 35 minutes.

And that doesn't include the return trip -- so in order to complete a delivery, the robot needs 20 to 70 minutes.

Conclusion: The claim is almost true

For the claim to be true, the robot would have to complete an "average delivery" in 27 minutes. Based on published wait times for delivery, this is close to the low end of travel time. Unless most trips are very short (bringing the average down), it's likely that each trip is using a bit more than the energy to heat water for one cup of tea.

At the high end, the robot uses enough to make almost three cups of tea:

(70 min / 120 min) * 168.18 Wh = 98.11 Wh
98.11 Wh / 37.5 Wh = 2.62 cups of tea

For anyone curious, I found my way to this answer by way of a research paper on efficiency of delivery robots: Vepsäläinen, Jari. 2022. Energy Demand Analysis and Powertrain Design of a High-Speed Delivery Robot Using Synthetic Driving Cycles. Energies 15, no. 6: 2198. Their calculation uses a lower value for power consumption (63.3 Wh/h) because they ignore anything that isn't directly related to locomotion: lights, sounds, charging efficiency, electronics, etc.

  • 2
    The quoted press release says " boiling a kettle to make just one cup" so it could mean wastefully boiling multiple cups of water in the kettle to produce just one cup to be consumed--you could get 2-6x the energy budget.
    – Dave X
    Commented Sep 30, 2022 at 14:06
  • 3
    @Dave I generally agree with you, but tend to be conservative in how I determine whether a claim is true or not. Since I can't say "this claim is undoubtedly true", it doesn't get the "true" label from me.
    – LShaver
    Commented Sep 30, 2022 at 17:25
  • 13
    Rather than convincing me that the robots are efficient, this has me thinking boiling water eats up more energy than most people think. Commented Sep 30, 2022 at 20:02
  • 5
    A joule, at 1Ws, is a perfectly fine unit of energy. 37.5 Wh *3600s/h = 135kWs = 135kJ. Burning coal produces 30,000kJ/kg, or 30kJ/h, so the energy to boil a cup of water is about 136kJ/(30kJ/g) = 4.5 grams of coal. Burning the equivalent energy of 4.5 grams of coal while delivering your groceries isn't as sexy a marketing hook.
    – Dave X
    Commented Sep 30, 2022 at 20:11
  • 5
    You're calculating for the wrong size of cup. You're using an English "cuppa", but drink sizes are way bigger here in Murica and this is a US company. Almost all the insulated cups sold here are in the 16-24oz range (don't believe? Check out "hot cup" in the Starbucks online store) and I always make my tea like that here. So if you update the calculation to be boiling(0.5-0.75l) of water (+ a bit more since people don't fill up their kettles exactly) their claim works out to be completely true
    – Eugene
    Commented Oct 1, 2022 at 17:04

tl;dr Plausible.

Let's compare their robot with a roughly equivalent electric vehicle, a hoverboard.

First, let's turn the marketing claims into numbers.

How much energy to boil a cup of tea?

It takes a lot more energy than you think. Water has one of the highest specific heat capacities, meaning it takes a lot of energy to raise its temperature. For every degree you heat water you could heat the same weight in iron by 10 degrees.

I put a 250 ml of cold tap water into my 1500 W electric kettle and it took 90 seconds to boil. That's 37.5 Wh. An MSN article agrees, they use 3000W for 45 seconds which is also 37.5 Wh.

To put this into perspective, this is the same amount of energy as being hit by a small car at 55 km/h. The claim is very clever marketing; the layperson thinks it's a minuscule amount of energy and makes it seem much more efficient than it actually is. Not to say it isn't efficient, it is much more efficient than driving that same car to get your stuff.

How big and how fast is the robot?

Starship Technologies states that...

Starship’s robots move at pedestrian speed and weigh no more than 100 pounds.

Typical pedestrian speed is 4km/h. 100 pounds is 45 kg. To include the weight of the order, let's round that up to an even 50 kg (you're going to super-size your order, admit it).

Comparison with Hoverboards

Hoverboard battery packs are typically made up of 18650 cells. Each cell carries about 9.4 Wh. So boiling a cup of tea uses the power of roughly 4 18650 cells.

A hoverboard battery typically carries 100-150Wh, 3 to 4 times more than our robot. However, they carry twice the weight (a 90kg adult + the 10 kg hoverboard) at 2 to 4 times the speed (10-20 kph).

Starship Hoverboard
Mass 50 kg 100 kg
Velocity 4 kph 10-20 kph
Energy 37.5 Wh 100-150 Wh
Range ? 8-20 km

This is not a simple linear relationship, for example kinetic energy and drag are the square of velocity; I'm not going to try to derive an exact range.

However, at half the weight and less than half the speed of a hoverboard, it seems plausible that this robot can have a range measured in kilometers on its smaller battery capacity. In my city of Portland, Oregon an 4 km range would cover a round trip in most of the downtown area, sufficient to deliver groceries and takeout.

  • 4
    @WeatherVane Assuming it is linear (it isn't, kinetic energy and drag are squared with velocity), something is wrong. Take the energy required for a hoverboard to go 8km. With 1/2 the mass you can go 2 times further: 8km * 2 = 16 km. With 1/4 the speed you can go 4 times as far: 16km * 4 = 64 km. With 1/3 the energy you can go 1/3 as far: 64km / 3 = 23.3 km. I'm not going to hazard a real guess except to say it's plausible the range is in kilometers not meters.
    – Schwern
    Commented Sep 29, 2022 at 21:03
  • 5
    @WeatherVane really, transporting 50 kgs is twice as difficult as transporting 100? That’s not how physics works…
    – Dan Romik
    Commented Sep 29, 2022 at 21:40
  • 1
    @DanRomik I know it isn't linear too, but the answer did not say so (now edited). There was just a table. Aren't "back of the envelope" answers invalid? I like the answer but am hoping for some justification from the manufacturer or an independant study as to how this "one cup of tea" is calculated. All we have so far is "this seems about right." But we don't even know what "the average journey" is. Commented Sep 29, 2022 at 22:28
  • 15
    @WeatherVane I’m saying you got the sign of the effect wrong, never mind its magnitude. Pushing a smaller mass surely requires less work than pushing a larger mass, but your 1/24 figure implies that you think it’s harder to push a smaller mass than a larger one. Whether the relationship is linear or not is beside the point, your mistake is much more basic than that.
    – Dan Romik
    Commented Sep 29, 2022 at 23:08
  • 5
    Nitpick: the fact that "ice cubes work much better than whiskey stones" is more due to the fact that ice cubes undergo a phase transition (melting) and absorb a huge amount of heat in the process. The same amount of heat that it takes to melt 1 gram of ice into water could also raise the temperature of 1 g of liquid water by almost 80°C. Commented Sep 30, 2022 at 14:17

@LShaver’s excellent answer crunches the numbers and concludes the claim is almost correct. I want to point out several additional factors that actually tilt the numbers further in favor of the company’s claim being true:

  1. Most electric kettles have a minimal water level required for safe operation. An example:

    For example, an electric kettle with a 1L capacity like Amazon Basic’s Portable Electric Hot Water Kettle requires a 0.5L minimum water level.

    0.5L means each time you boil the kettle to make yourself a cup of tea, you’re actually boiling at least two cup-fuls of water. That’s twice as much energy as the cited value of 37.5Wh, with half of the energy being wasted, unless you only drink tea in the company of other people, or like to drink half-liter mugs of tea.

  2. Regardless of the minimal required level of water in a kettle, I’m willing to bet 99.9% of people overfill their kettles slightly (or a lot) when they’re boiling water to make tea. So again, there’s probably a wastage factor of around 25%-100% that needs to be factored in to the calculation.

  3. Finally, in the US at least an embarrassingly large number of people still use traditional (non-electric) kettles that are heated on a gas stove. That is considerably less efficient than using an electric kettle (according to this source, a gas stove only transfers around 35%-40% of the burning energy into useful heating of pot contents). So again, this makes the company’s comparison more favorable to the delivery robot.

In summary, with these factors taken into account, I think it’s reasonable to label the claim as correct rather than almost correct.

  • That would mean the amount of energy to make multiple cups of tea not just one. Information would be needed on how many people are using a kettle like that and overfilling it versus something designed to make a single cup of tea.
    – Joe W
    Commented Sep 30, 2022 at 18:33
  • as little energy as boiling a[n entire] kettle to make just one cup of tea, +1, doesn't sound so good anymore. Gas, +2 : It doesn't take a lot more energy than I think. IDK what the BTU on one of my stove burners is, but it's a lot. And if it takes 2~3m to boil a pot of water, then it's consuming an actually rather insane amount of energy. A spaceship company that makes cars, now it's the thing from Fight of the Navigator? That little thing, that goes 2mph? sure. Or you could have one cup of coffee. That should put your drinking problem into perspective, not vice versa.
    – Mazura
    Commented Oct 1, 2022 at 1:49
  • @JoeW: I can't think of any inexpensive water-boiling appliance in the US which would could safely boil 150ml of water without having to boil much more. Further, a lot of Americans would routinely use a gas stove to boil water--something which would be energy efficient in wintertime, but very inefficient in summertime.
    – supercat
    Commented Oct 2, 2022 at 15:40
  • @supercat depends on what you consider inexpensive. Small micwowave ovens can be got for under $40.
    – barbecue
    Commented Oct 2, 2022 at 15:56
  • 1
    @AI0867 that’s a good point, thanks for keeping me honest… I agree that my point number 3 was the weakest one of my answer.
    – Dan Romik
    Commented Oct 3, 2022 at 14:40

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .