A week ago, analyst TD Cowen revealed that Microsoft had canceled leases "totalling a couple hundred MWs," with "at least two private data center operators across multiple US markets." The report also details how Microsoft "pulled back on converting negotiated and signed Statement[s] of Qualifications (SQQs)," which it added
I’m gonna disagree - it’s not like DeepSeek uncovered some upper limit to how much compute you can throw at the problem. More efficient hardware use should be amazing for AI since it allows you to scale even further.
This means that MS isn’t expecting these data centers to generate enough revenue to be profitable, and they’re not willing to bet on further advancements that might make them profitable. In other words, MS doesn’t have a positive outlook for AI.
Exactly. If AI were to scale like the people at OpenAI hoped, they would be celebrating like crazy because their scaling goal was literally infinity. Like seriously the plan that openai had a year ago was to scale their AI compute to be the biggest energy consumer in the world with many dedicated nuclear power plants just for their data centers. That means if they dont grab onto any and every opportunity for more energy, they have lost faith in their original plan.
More efficient hardware use should be amazing for AI since it allows you to scale even further.
If you can achieve scaling with software, you can delay current plans for expensive hardware. If a new driver came out that gave Nvidia 5090 performance to games with gtx1080 equivalent hardware would you still buy a new video card this year?
When all the Telcos scaled back on building fiber in 2000, that was because they didn’t have a positive outlook for the Internet?
Or when video game companies went bankrupt in the 1980’s, it was because video games were over as entertainment?
There’s a huge leap between not spending billions on new data centers ( which are used for more than just AI), and claiming that’s the reason AI is over.
If a new driver came out that gave Nvidia 5090 performance to games with gtx1080 equivalent hardware would you still buy a new video card this year?
It doesn’t make any sense to compare games and AI. Games have a well-defined upper bound for performance. Even Crysis has “maximum settings” that you can’t go above. Supposedly, this doesn’t hold true for AI, scaling it should continually improve it.
So: yes, in your analogy, MS would still buy a new video card this year if they believed in the progress being possible and reasonably likely.
Like games have diminished returns on better graphics (it’s already photo realistic few pay $2k on a GPU for more hairs?), AI has a plateau where it gives good enough answers that people will pay for the service.
If people are paying you money and the next level of performance is not appreciated by the general consumer, why spend billions that will take longer to recoup?
It’s still not a valid comparison. We’re not talking about diminished returns, we’re talking about an actual ceiling. There are only so many options implemented in games - once they’re maxed out, you can’t go higher.
That’s not the situation we have with AI, it’s supposed to scale indefinitely.
Current games have a limit. Current models have a limit. New games could scale until people don’t see a quality improvement. New models can scale until people don’t see a quality improvement.
I’m supposed to be able to take a model architecture from today, scale it up 100x and get an improvement. I can’t make the settings in Crysis 100x higher than they can go.
Games always have a limit, AI is supposed to get better with scale. Which part do you not understand?
But really the “game” is the model. Throwing more hardware at the same model is like throwing more hardware at the same game.
Which part of diminished returns not offering as much profit did you not understand?
Current models give MS an extra 30% revenue. If they spend billions on a new model will customer pay even more? How much would you pay more for a marginally better AI?
This doesn’t really work, because the goal when you buy a video card isn’t to have the most possible processing power ever and playing video games doesn’t scale linearly so having an additional card doesn’t add anything.
If I was mining crypto, or selling GPU compute (which is basically what ai companies are doing) and the existing card got an update that made it perform on par with new cards, I would buy out the existing cards and when there are no more, I would buy up the newer cards, they are both generating revenue still.
But this is the supposition that not buying a video card makes you the same money. You’re forecasting free performance upgrades so there’s no need to spend money now when you can wait and upgrade the hardware once software improvements stop.
And that’s assuming it has anything to do with AI but the long term macroeconomics of Trump destroying the economy so MS is putting off spending when businesses will be slowing down because of the tariff war.
I’m gonna disagree - it’s not like DeepSeek uncovered some upper limit to how much compute you can throw at the problem. More efficient hardware use should be amazing for AI since it allows you to scale even further.
This means that MS isn’t expecting these data centers to generate enough revenue to be profitable, and they’re not willing to bet on further advancements that might make them profitable. In other words, MS doesn’t have a positive outlook for AI.
Exactly. If AI were to scale like the people at OpenAI hoped, they would be celebrating like crazy because their scaling goal was literally infinity. Like seriously the plan that openai had a year ago was to scale their AI compute to be the biggest energy consumer in the world with many dedicated nuclear power plants just for their data centers. That means if they dont grab onto any and every opportunity for more energy, they have lost faith in their original plan.
If you can achieve scaling with software, you can delay current plans for expensive hardware. If a new driver came out that gave Nvidia 5090 performance to games with gtx1080 equivalent hardware would you still buy a new video card this year?
When all the Telcos scaled back on building fiber in 2000, that was because they didn’t have a positive outlook for the Internet?
Or when video game companies went bankrupt in the 1980’s, it was because video games were over as entertainment?
There’s a huge leap between not spending billions on new data centers ( which are used for more than just AI), and claiming that’s the reason AI is over.
It doesn’t make any sense to compare games and AI. Games have a well-defined upper bound for performance. Even Crysis has “maximum settings” that you can’t go above. Supposedly, this doesn’t hold true for AI, scaling it should continually improve it.
So: yes, in your analogy, MS would still buy a new video card this year if they believed in the progress being possible and reasonably likely.
Like games have diminished returns on better graphics (it’s already photo realistic few pay $2k on a GPU for more hairs?), AI has a plateau where it gives good enough answers that people will pay for the service.
If people are paying you money and the next level of performance is not appreciated by the general consumer, why spend billions that will take longer to recoup?
And again data centers aren’t just used for AI.
It’s still not a valid comparison. We’re not talking about diminished returns, we’re talking about an actual ceiling. There are only so many options implemented in games - once they’re maxed out, you can’t go higher.
That’s not the situation we have with AI, it’s supposed to scale indefinitely.
Current games have a limit. Current models have a limit. New games could scale until people don’t see a quality improvement. New models can scale until people don’t see a quality improvement.
I’m supposed to be able to take a model architecture from today, scale it up 100x and get an improvement. I can’t make the settings in Crysis 100x higher than they can go.
Games always have a limit, AI is supposed to get better with scale. Which part do you not understand?
You can make Crysis run at higher fps. You can add polygons. (remember ati clown feet?) You can add details to textures. https://research.nvidia.com/publication/2016-06_infinite-resolution-textures
But really the “game” is the model. Throwing more hardware at the same model is like throwing more hardware at the same game.
Which part of diminished returns not offering as much profit did you not understand?
Current models give MS an extra 30% revenue. If they spend billions on a new model will customer pay even more? How much would you pay more for a marginally better AI?
If buying a new video card made me money, yes.
This doesn’t really work, because the goal when you buy a video card isn’t to have the most possible processing power ever and playing video games doesn’t scale linearly so having an additional card doesn’t add anything.
If I was mining crypto, or selling GPU compute (which is basically what ai companies are doing) and the existing card got an update that made it perform on par with new cards, I would buy out the existing cards and when there are no more, I would buy up the newer cards, they are both generating revenue still.
But this is the supposition that not buying a video card makes you the same money. You’re forecasting free performance upgrades so there’s no need to spend money now when you can wait and upgrade the hardware once software improvements stop.
And that’s assuming it has anything to do with AI but the long term macroeconomics of Trump destroying the economy so MS is putting off spending when businesses will be slowing down because of the tariff war.