Microsoft spends billions on AI, converts just 3.3% of Copilot Chat users

Only 3.3 percent of Microsoft 365 and Office 365 users who touch Copilot Chat actually pay for it, an awkward figure that landed alongside Microsoft's $37.5 billion quarterly AI splurge and its insistence that the payoff is coming. That single percentage stat undermines the company's carefully polished Copilot success story. …

[Groo The Wanderer - A Canuck]

*snicker* PAY for something I removed your entire operating system to escape from? Now that's good for a laugh!

[Yet Another Anonymous coward]

Isn't that Oracle's proprietary business plan ?

[Rich 2]

Most people don’t want to pay for a word processor or spreadsheet or other notionally useful application. So asking the same people to spend money on something that is considered by most people to be a bit of fun (at best) or the work of the devil is going to be a tough sell

[Kurgan]

Everyone is paying for Copilot

Even if you don't pay for copilot, if you pay for MS products, you are paying for copilot.

And even if you don't use AI, if you buy hardware, you are paying for AI.

And even if you don't buy hardware but just grocery, you are paying for AI. Because computers, software, and power are used everywhere on every supply chain, so in the end AI is making everything more expensive.

[Anonymous Coward]

Re: Everyone is paying for Copilot

Yes and we're all going to be paying for it for years when it pops.

[IGnatius T Foobar !]

the only use of copilot...

Honestly, the only reason I use copilot at all is because of its generous free tier. I'd never pay for it, and I'm on Linux so I don't have to sit through all those forced integrations.

They're not gonna make money on this.

[steviesteveo]

Re: the only use of copilot...

It's first one free rules but that relies on the product being addictive

[TVU]

"Microsoft spends billions on AI, converts just 3.3% of Copilot Chat users"

I think it is safe to say that recruitment initiative does not represent good value for money. Indeed, that money would have been much better spent on Win 11 update quality control.

[Doctor Syntax]

Improving W11 quality doesn't enable them to promise wonders for investors so no chance of that happening.

[TVU]

^ Now that is a very valid and relevant point!

[FirstTangoInParis]

> Improving W11 quality doesn't enable them to promise wonders for investors

How about they spend money on fixing ALL their products? Perhaps the shareholders would like less negative press? But then I guess Zero Days don’t get reported in the broadsheets. MS CEO and first line have no shame.

[Alumoi]

What quality control?

[DJV]

What quality control?

Precisely!

[Dan 55]

I don't know who would pay for this crap

As time is always scarce, I told it to write a quick script for me. It generated the script but there were a number of problems with it. I told it what it what the problems were so it generated a new script and claimed the problems were allegedly solved however new problems had appeared so I couldn't test it. After five tries like this I gave up.

And I couldn't even fix the script myself as it contains an awk call in the middle which is completely incomprehensible.

I'm supposed to live in fear of this waste of electricity, water, bricks, and mortar taking my job away. I've never slept easier in my life.

[Like a badger]

Where there's a will there's a way

Worth bearing in mind that for home users of O365, the "free trial" runs until they renew their subscription. Then the renewal page only shows renewal options with Copilot, and for my family 365 sub that comes with a 33% (+£25) price increase on last year. Many people will just suck that up without thinking, especially because the cheaper-without-Copilot option is not shown unless the user selects "cancel my subscription". Having done so, my sub will lose Copilot at the renewal data because to me it's not even worth a few pence per month.

I suspect a good chunk of those 3.3% who are paying didn't realise they could avoid paying for Copilot, and I further guess that the number paying for Copilot will rise over the year for the same reason. Although on the subject of things I suspect and guess, I'll posit that 70p per user per month* is never going to pay off Microsoft/ChatGPT's ludicrous levels of LLM investment. Business users are already paying much more per user, but in both cases paying nominal amounts for Copilot is merely the thin end of the wedge, and in future years MS will need the price to go up dramatically.

* My household calcs: £25 annual increase for Copilot/(three users x 12 months)

[Dan 55]

Re: Where there's a will there's a way

I suspect a good chunk of those 3.3% who are paying didn't realise they could avoid paying for Copilot

And 96.7% home users don't realise they could avoid paying for Office as well?

[Like a badger]

Re: Where there's a will there's a way

Probably true for most of them.

In my case it's centred on the fact that the youngster is doing medicine, that's hard enough and I don't want to cause any problems with coursework that's all been created by MS applications and which she's been educated with. Once they've finished, then I'm a free agent, and Open Office goes onto the family fleet, and hopefully her indoors won't notice.

[PCScreenOnly]

Re: Where there's a will there's a way

But the 1TB of onedrive is damn cheap when done this way

[Charlie Clark]

Re: Where there's a will there's a way

Looking at the 6TB drive in the corner, I beg to differ… And performance on OneDrive is awful in comparison to Dropbox or Google Drive because everything has to go through the clusterfuck that is SharePoint – something that was happily chewing data and cycles while Sam Altman was still wetting his bed…

[blu3b3rry]

Re: Where there's a will there's a way

Agreed about crap performance on OneDrive. It's not unusual for it to develop random sync errors for seemingly no good reason, either....

I haven't ever managed to workout exactly what the OneDrive client is doing to hog over 80% of CPU resources for anywhere between 5 and 15 minutes at boot while it syncs with the company SharePoint sites. I used to use the GDrive client routinely at home on a machine with about 1/3 of the processing power and it never chugged anywhere near that much.

[sal II]

Re: Where there's a will there's a way

You can't compare on-site storage with offsite storage. It's apples to oranges. For the average user the easiest offsite backup is cloud and out of these M365 Family is by far the cheapest $/TB option by a mile even if you exclude the Office licenses.

[sarusa]

Re: Where there's a will there's a way

Yeah, MS has been doing everything it can to cook the books on this, like counting Office sales as 'AI Revenue'. And even with that the best they can come up with is 3.3%? I suspect it what's you said.

[Anonymous Coward]

Re: Where there's a will there's a way

Remember, AI is in Windows, so every Windows activation counts as using AI.

[Charlie Clark]

Re: Where there's a will there's a way

I think it's fair to say that subcription prices already contain something to help cover the cost of the free tier and that pricing will continue to be adjusted until enough people buy it officially. Whatever happens, subscription prices for most online services are likely to keep going up because AI is increaing costs both indirectly and indirectly.

[Richard 12]

Re: Where there's a will there's a way

Not really. Even the most expensive Copilot subscription doesn't cover the cost of the servers.

The run rate is genuinely incredible.

[Tron]

Imagine that.

It's like Wednesday following Tuesday.

Lay your bets, readers. Will the bubble pop or will it deflate?

[Wellyboot]

Re: Imagine that.

Quietly ignored as soon as the next money making bandwagon appears.

Has AI found an actually helpful place for blockchain yet?

[AMBxx]

Re: Imagine that.

Only in the Metaverse.

[Groo The Wanderer - A Canuck]

Re: Imagine that.

*Badum-cha!*

[Doctor Syntax]

Re: Imagine that.

All the quantum folk must be upset that AI's keeping them away from their turn for a bubble.

[Anonymous Coward]

Re: Imagine that.

Oh I've already seen a fair bit of huffing and puffing to get the QuantumAI bubble going.

[FrogsAndChips]

"plans to streamline or even remove some AI features where they don't make sense"

Notepad would like a word.

[VoiceOfTruth]

This is amusing

Years ago I worked at a company where certain customers were effectively loss-makers - it cost more to have them as customers than they paid. The fact that they were loss-makers was understood, but we tolerated them because they were about 3 or 4% overall. It would also look bad if we stopped their service. In time some churned away. The 95% of customers who paid were effectively subsidising the loss-makers.

According to Google, 5 to 6%% of customers pay for ChatGPT. MS is even less. Perplexity is about 1%.

There is no business case where 3 or 4% pay to cover the costs of the rest. But that is what the AI 'market' seems to be. There are no 'investments' in AI. They are throwing money into the bin.

[xcdb]

Re: This is amusing

Well, "freemium" business models typically convert at around 2%, so in that sense ChatGPT is doing really well. Those companies don't typically start $50bn in the red though :D

[Filippo]

Re: This is amusing

You make a good point, but "freemium" models work where the marginal per-user cost of the service is extremely low - I think it's mostly for software that's costly to develop, but really cheap to run. If each user costs you a penny per month in hosting, it's fine if almost all of them don't pay.

LLM users very much do not cost pennies per month. It's more like dollars, or even tens of dollars. Those inference runs are really expensive. There is no way that model works for LLMs.

[Groo The Wanderer - A Canuck]

Re: This is amusing

There is also a dramatic difference in the capital spend required to run those LLMs compared to traditional services that could get away with a freemium model. Capital investment requires loans, which cost interest from Day Zero, so the bill is growing as astronomically as the capital expenditures themselves are. How anyone can see this whole infrastructure of insane spending on low-return "product" as a valid business model is beyond me.

And don't get me started on my whole "Artificial Ignorance" diatribe. Garbage in, garbage out, and it is the distillation of the internet's collection of the "knowledge" of the masses.

As George Carlin was to have said "Just think how dumb the average person is, and remember, half the people are dumber than that." (paraphrase)

[VoiceOfTruth]

Re: This is amusing

I think that's the killer here. I can make an app or whatever. Even if it takes me a year, the costs are not astronomical. So I can get away with 2% fremium upgrades. Or if like Adobe there is the free version online, that is paid for by having huge numbers of paying customers. But this AI stuff is just crazy sums of 'investment'. Who knows, perhaps lots of people will pay for it.

[Anonymous Coward]

Re: This is amusing

Come on old chap. You really really don't understand, do you? Let me spell it out:

1) Get AI

2) Profit

See? It's easy, isn't it?

/Sarc

[Charlie Clark]

Re: This is amusing

Er, the underpant gnomes have a three point plan:

• Get AI
• …
• Make Profit

[David 132]

Re: This is amusing

More AI-triggered job-losses and economic destruction, I see.

In this case, the Underpants Gnomes, who would ordinarily be handling the intermediate ("???") step.

Edit: Hah, I left the page open for 40 mins whilst dealing with another issue; this'll teach me to refresh the page before commenting. I see that in the interim, Charlie Clark beat me to a very similar point!

Ah well, I'll leave this up anyway as a testament to my poor commenting-discpline.

[Long John Silver]

If you are a widow or an orphan,

steer well clear of investing in AI.

[Anonymous Coward]

Useful A.I.

The only useful feature of A.I. at the moment that has serious uptake seems to be the generation of pictures and short videos. But this has little or no impact on the bottom line of these companies investing tens if not hundreds of billions in new data centers. It's a huge loss leader and it's only a matter of time before Wall Street reigns them in.

[VoiceOfTruth]

Re: Useful A.I.

I made a post recently about the Google AI watermelon elephant advert. I have limited use or need for watermelons as elephants. It's fun for about 2 seconds, but I have already moved on.

[blu3b3rry]

Re: Useful A.I.

Well, quite. It's one thing to get people to fuck about with the image generation and the like - but pay for it? Hardly anyone is likely to do so for what is effectively seen as a toy.

[Sorry that handle is already taken.]

Re: Useful A.I.

And it's only really "useful" in the sense that it's revolutionised the generation of spam and scams

[Anonymous Coward]

Just take it round the back of the barn and shoot it.

[Inventor of the Marmite Laser]

Double the fun

An AI enabled 3D TV with a built in videodisc player.

[Brl4n]

bankrupt them all

The only positive I see re: AI is for it to bankrupt all of big tech. But our corrupted officials will most likely remain on trend and socialize the loses/bankruptcies as recent history has shown for the sake of the insane argument of economic and national security. This is a bizarre campaign of forcing terrible products on to the majority of paying customers that don't want it...repeatedly, year after year. Customers have to start voting with their pocket books because they don't seem to be listening.

But I guess I'm the crazy one...

[David Hicklin]

Re: bankrupt them all

> But our corrupted officials will most likely remain on trend and socialize the loses/bankruptcies as recent history has shown for the sake of the insane argument of economic and national security

I'm not so sure of this mainly because the countries of those officials are neck-deep in debt already and getting deeper - most can't afford to bail them out and unless it is going to cause a banking crisis (which the stress tests are supposed to eliminate) they will have no choice but to let them fail this time.

[nobody who matters]

"CEO talks momentum ...."

CEO talks out of his arse more like.

[Scotthva5]

That "momentum" almost certainly includes punters that used it one time and said "thank you but no". To a CEO desperate to to justify the massive capex that counts as momentum.

[FirstTangoInParis]

On TV this morning some c suite nob from Starbucks was talking about adding AI to their coffee. It was a proper buzzword bingo interview where he riffed on leveraging headwinds to up the customer experience on tariff-free green coffee or some such bollards. I’d just like some coffee, thanks. I don’t want to pay for their massive splurge on AI that nobody wants and will have no measurable positive impact on their profits.

[ComicalEngineer]

Would you like AI with that?

"The company is trialling an AI-powered chatbot, which can help match drinks with customer moods, and is introducing the ability to schedule orders in a bid to reduce customer waits."

I wonder if an AI serving bot would be able to guess that I am in the mood to destroy it with a fire axe?

My choice has been to bypass any service station on the motorway that is Starbucks unless absolutely desperate for coffee, as I find them miserable and uncomfortable compared to the Costa competition (that's just my preference). Now it seems that they want ot enshittify the coffeee experience even more by having an AI bot serving? Will it talk to me and ask if how I am and then tell me to have a good day when it serves my drink?

Got to be one of the decisions that will kill the business.

[Chet Mannly]

Starbucks should try adding actual coffee to the brown water they serve before messing with AI...

[Jimjam3]

Should read;

“CEO takes a motion”

Then preferably flushes it afterwards.

[Joe Gurman]

Nice to see that the market....

....was not amused by the mismatch of all that AI spending and such minimal adoption (shares off 10% since the day after the earnings call).

[ecofeco]

Manufacturing consent

Isn't cheap!

[Daemonik]

If it worked, they might see more

It's a terrible advert for a product when the free trial offers nothing but frustration. We tried it, it was awful, we stopped. No plans to use it, or anything like it, anytime soon. Hallucinations (or possibly just very poor sourcing, we couldn't be bothered to check) aplenty and the general output was of a quality that could best be described as "garbage".

A lot of the code output read like the op in a stack overflow titled "just started coding, why doesn't this work?"

[Blue Screen of Bleurgh]

Win or fail, CEO Satya Nadella will still pick up a nice fat bonus.

That's how it seems to work in business: big corporations make huge losses, and yet the bosses still get their cut for being utterly incompetent and out of touch

[Excused Boots]

Well there’s always hope.

See Sirius Cybernetics Corporation.

[Pen-y-gors]

Another planet...

"plans to streamline or even remove some AI features where they don't make sense."

Start with M$ Swiftkey Keyboard. Which now lets you use AI to make pictures. It's an f***ing KEYBOARD! I could see a use-case for some very focussed LLM work that improves auto-corrupt, based on previous typing, but drawing pictures?

Don't forget you can 'downgrade' your Office 365 sub to the non-Copilot 'Classic' - for now at least. A lot cheaper - i.e. the old price.

[DS999]

The interesting numbers will come over the next year

How many of the people currently paying for Copilot will continue doing so? I expect a lot of businesses are signing up because of all the AI hype, but unless they see some tangible benefits the beancounters are going to say "do we really need to keep paying for this?" Or maybe limit it to a small selection of employees who are actually being helped by it, rather than signing up everyone who has a desk job which is I'm betting what a lot of CEOs/CIOs wanting to be seen as "early adopters" have been doing.

I don't expect Microsoft to make those numbers public, but it might be possible to tease partial information of this out if they say something like "10 million new Copilot signups so far this year" in Q3 but then at the end of the year they announce the total and have less than 10 million more than the 15.5 they claim now.

If they stop telling you anything about Copilot signups that'll be a good clue that the numbers are going south fast.

[Mostly Irrelevant]

I'm guessing that almost all uses of Co-pilot Pro have it bundled with Office 365. It's bundled with every version. I have it at work and I never use it.