Update 17 Dec 2025: Times have changed, and I have realised that this article needs a short framing intro. Climate science is in trouble. Funding is being cut. Information is disappearing. To draw a little attention to this and other issues, I returned to a format I first introduced in 2003: a darkly humorous “Letter to Santa Claus”. Humorous, but very serious. In this social media-dominated era, the algorithms weren’t spreading my notices about this Letter. Maybe they didn’t understand what I was trying to do. Hence this pedantic intro.

Now, here’s the letter, complete with links to sources and a detailed calculation on the cost, in melted-ice terms, of the accompanying, AI-generated illustrations.

Dear Santa Claus,

It’s been nine years since we last checked in with each other. Twenty-three years since I wrote my first, semi-viral letter to you about global warming. And though you never answer my letters, I still like to think of this communication as an exchange. I write to you when I feel desperate for help with this huge problem we now call “climate change”. I give you my big wish list — things like a new energy system, better electric cars, rapid CO2 emissions reduction — and eventually, you give me some of the things on my list. Partly. Sort of.

I don’t mean to sound ungrateful. I know I’m asking for a lot. Anything you can do to help is greatly appreciated.

But lately, Santa, it seems like you have been taking back some of your earlier gifts of climate progress. It’s like you’re flying around with a big vacuum cleaner, sucking solar panels back up the chimney. And leaving little bags of coal.

Okay, I know you’re not doing that — though I certainly wouldn’t blame you if you got really mad at us. The sea ice around your North Pole home is not getting any thicker. Worse, some people just don’t seem to want to know what’s happening up there where you live, and don’t really want anyone else to know, either.

The National Snow and Ice Data Center (NSIDC) of the United States, for example, used to maintain something called the “Near-Real-Time Sea Ice Concentration Climate Data Record.” It’s also known as “Sea Ice Today”. It’s just what it sounds like, a website where you could easily see what’s happening right around your polar house, right now. It’s also called a “data product”, with animated maps and everything — but the NSIDC has been forced to “retire” that “data product” this year, for “budget” reasons (though fortunately, the data itself is still being maintained). Ho, ho, ho.

To be honest, it’s getting harder to know what we’re actually doing to you, the elves, and the reindeer. Even the fabled CO2 monitoring station on the top of Hawaii’s Mauna Loa volcano, source of the famous “Keeling Curve” of atmospheric CO2 measurements dating back to the 1950s, could get axed because of US budget cuts. If the US stopped all of its CO2 measurement — which is suddenly, weirdly possible — that action would remove 40% of the world’s monitoring capability. The world would still know what we’re doing wrong, of course. Just not so exactly. Maybe Rudolf could help us fill the gap, by blinking that shiny red nose of his, really fast.

And Santa, I hesitate to tell you about the results of the most recent round of UN climate negotiations in Belém, Brazil. Because if I do, and you compare what we should be doing with what we’re actually doing, I’m afraid you’ll declare all of us humans to be naughty. Maybe you’ll cancel the whole sleigh ride.

Please don’t do that, Santa. Really, most of us don’t deserve to share the blame for the melting of your beloved Arctic ice. (I probably do. I flew from Sweden to France twice this year.) There’s even research showing how a pretty small number of companies are responsible for at least half of the increasing intensity of heat waves that we saw between 2000 and 2023.

Santa, a side note: some people would like us to stop spreading knowledge about stuff like that. That’s what the budget cuts are all about. Got any climate research funding, and maybe support for investigative climate journalism, in that big gift bag of yours?

Also, given that a bunch of rich countries have severely cut their financial support to the poorest countries over this past year, I can think of many places where your generosity is needed now more than ever. So hey, if you want to skip over Sweden in 2025, that’s alright with me. I can give you a list of countries that I think you should give extra attention to, instead. Plus a suggested gift list. You know, schools, clinics, clean water, HIV-medication, that kind of thing.

Otherwise my wish list hasn’t really changed in 23 years. It’s all about getting us off fossil fuels and onto clean renewables. And we need that for every energy-dependent, absolutely-essential industrial system that we have. Food. Transport. Electricity. Heat. And home appliances, of course. Especially dishwashers.

I might like to add a couple of things this year, such as extra protection for the Amazon. And maybe an additional sense of urgency, given the rapid growth of energy-hungry AI data centers. By the way, did you like the illustration I made … er, prompted for you? The same AI that created the picture also did a massive amount of research to calculate the cost of generating that lovely image of you. The resulting CO2 emissions will melt only “a few hundredths of a square centimeter” of your precious sea ice. (See the “PS” for the actual calculations.)

Worth it, right? I want to stress that replacing your fingers with miniature reindeer horns was not my idea. Those AIs can certainly be creative.

While we’re at it, how about leaving a little more financing to support climate adaptation under the tree this Yuletide? I’ve accepted that you are a little slow to deliver on mitigation actions, Santa, despite my increasingly desperate pleas. So when you do get around to granting my wishes, make sure to throw in a little extra drought and flood protection, okay? Our insurance companies don’t like to cover that kind of stuff anymore.

So, I’m very sorry about what’s happening to the North Pole, Santa. But frankly, that’s the least of our problems. We have a whole bunch of climate-related “tipping points” to worry about, scattered elsewhere on the planet. So it is well within your self-interest to make a grand gesture this year. Call attention to your plight. Maybe leave little cards under every tree, showing a photo of a cute elf, stranded on the melting sea ice.

(See my AI-generated example. Again, I want to stress that the polar bear swimming threateningly towards the elf was the AI’s idea, not mine.)

Finally, Santa, remember when I wrote you in 2016, hinting that I would appreciate an upgrade to my electric car, and subtly suggesting that you give me a Tesla?

Well, you can forget about that. Strike it from my list. Long story.

Looking forward to your climate-positive response,

Sincerely, nervously yours,

Alan

PS: Estimate regarding cost in sea ice terms of generating an AI image

Here are the calculations I promised, generated by ChatGPT 5.2. The AI thinking process took 1.32 minutes, which is pretty long in AI terms. (It would have taken me something 1.32 months.) The AI seemed to use reputable sources, which are included. But I have no idea if this estimate is correct.

My prompt:

How much arctic ice is melted by one AI-generated image? Can you find sources that would permit an estimate?

ChatGPT’s response:

You can back-of-the-envelope this, but the uncertainty is dominated by “how much electricity did your image actually take?” and “what grid (or carbon accounting method) powered it?”

Sources that let you estimate it

1) Convert CO₂ → Arctic sea-ice area loss

There’s a well-known near-linear relationship between cumulative CO₂ emissions and Arctic sea-ice area. Two commonly cited formulations:

• ~3 m² of September Arctic sea ice per metric ton of CO₂ (often quoted from Notz & Stroeve’s work; e.g., UCL’s writeup). University College London
• A more detailed seasonal framing: ice loss per ton CO₂ varies across the year, roughly ~1 m² in winter to >3 m² in summer. repository.library.noaa.gov

So you can use:$$\text{Sea-ice loss (m²)} \approx \text{CO₂ (tonnes)} \times (1\text{ to }3)\,\text{m²/tonne}$$Sea-ice loss (m²)≈CO₂ (tonnes)×(1 to 3)m²/tonne

[I am not sure why ChatGPT repeated this. /AA]

2) Convert electricity used per image → CO₂

You need two inputs:

(a) kWh per image. This is highly model/hardware/resolution dependent. Examples of sources that provide (or compile) values:

• A 2024 conference paper discussing inference energy uses 0.003 kWh per image as an example value (and explicitly notes large uncertainty and lack of standard methods). clagtee.fi.mdp.edu.ar
• A 2025 paper empirically measures diffusion-model image generation energy and finds large variation (up to ~46× across models), and resolution effects of ~1.3× to ~4.7× when increasing size (useful for bounds even when you don’t know the exact model). arxiv.org+1

(b) Grid carbon intensity (gCO₂/kWh). For this you can use:

• Country/year grid-intensity datasets (e.g., Ember/OWID). Our World in Data

Then:$$\text{CO₂ (g)} = \text{kWh per image}\times \text{(gCO₂/kWh)}$$CO₂ (g)=kWh per image×(gCO₂/kWh) $$\text{CO₂ (tonnes)} = \text{CO₂ (g)}/10^6$$CO₂ (tonnes)=CO₂ (g)/10⁶

[More ChatGPT repetition, probably to make sure its formatting was clear enough. /AA]

A worked example (order-of-magnitude)

Let’s pick:

• Electricity per image: 0.003 kWh (a commonly used estimate in the literature and commentary) clagtee.fi.mdp.edu.ar
• Grid intensity: 455 gCO₂/kWh (placeholder; you’d ideally plug in the actual grid where inference ran—see OWID/Ember for country values). Our World in Data
• Ice-loss factor: 3 m²/ton CO₂ (September framing) University College London

Calculation:

• CO₂ per image = 0.003 kWh × 455 g/kWh = 1.365 g CO₂
• In tonnes: 1.365 / 1,000,000 = 1.365×10⁻⁶ t CO₂
• Sea-ice loss: 3 m²/t × 1.365×10⁻⁶ t = 4.10×10⁻⁶ m²

That’s 0.0409 cm² of September Arctic sea ice per image (since 1 m² = 10,000 cm²) — roughly “a few hundredths of a square centimeter” per image, under those assumptions. University College London+2Our World in Data+2

What makes the estimate swing the most

• Model + settings: Different diffusion models and pipelines can differ by tens of times in measured energy use; higher resolution can increase energy materially for some models. arxiv.org+1
• Where it ran: The carbon intensity of electricity varies massively by country/region and time; using a location-based grid factor vs market-based accounting can change the implied emissions. (This disagreement shows up a lot in recent reporting about AI footprint transparency.) The Verge+1
• Which ice-loss coefficient you choose: seasonal vs September-only can change the factor by ~3×. repository.library.noaa.gov+1

A practical “plug-in” estimator you can reuse

If you tell me (1) your best guess of kWh per image (or model + resolution + steps), and (2) the country/grid you want to assume (e.g., Sweden vs global average vs US), I can compute a tight range. The structure is:$$\boxed{\text{Ice loss (cm²)} \approx \text{kWh/img}\times \text{gCO₂/kWh}\times \frac{(1\text{–}3)\times 10^4}{10^6}}$$Ice loss (cm²)≈kWh/img×gCO₂/kWh×((1–3)×10^4​​/10⁶)

or numerically:$$\text{Ice loss (cm²)} \approx \text{kWh/img}\times \text{gCO₂/kWh}\times (0.01\text{–}0.03)$$Ice loss (cm²)≈kWh/img×gCO₂/kWh×(0.01–0.03)

(That last factor comes from unit conversions and the 1–3 m²/ton relationship.) repository.library.noaa.gov+1