Nvidia's subscription software empire is taking shape In the wake of the AI boom, Nvidia has seen its revenues skyrocket to the point at which it briefly became the most valuable corporation in the world. That growth was overwhelmingly driven by demand for its datacenter GPUs to train and run the ever-growing catalog of better, smarter, and bigger AI models. But as much as …
What worries me is the more hardware they rent out, the less will be available for sale.
Demand will drive up prices on a dwindling supply. In and of itself, that's manageable. However, another e-Coin price spike will deplete that inventory even more. Great news for nVidia, sucks if you are a consumer.
Well I am talking here about personal pensions in the UK, which come with varying degrees of risk.
If you are saying that even the "low-risk" products are _forced_ to buy this, it makes the higher risk ones look extra scary!
Or are you saying we should support the hype to protect pensions? ;)
Personal pensions come in various types, most SIPPs will expect you to select your own funds and shares, you can plump for safe 1% bonds which is where most company schemes are put as the risk is almost non-existent and growth is minimal, which has led to so many people getting into their 60s and realising they're screwed.
If you entertain any hope of retiring at a sensible age then you have to play in the risky big boys playground Im afraid, take the risk with equities if you want to realise 7%+ minium returns year on year 'cos you got to beat away the 3% inflaiton chewing at your heels every single day.
This is isn't just pensions this goes for your ISAs too, you can take the crappy 4% your bank offers or, as the man on the Saturday night gameshow used to say, "Try for the big money prize.".
That's because they are greedy idiots.
nVidia is fabless. So nVidia 'hitting the wall' of growth is inevitable as TSMC has finite production ability - ability that it shares with customers beyond nVidia, at that. There simply isn't enough capacity at TSMC to allow nVidia to 'double growth' for every quarter from here, it's impossible.
Add in the fact that nVidia is indeed "fabless", which means that they are at the mercy and abilities of the aforementioned TSMC. Can we say "earthquake", for one, never mind the political tensions in the region. To give a fabless company, read a company that doesn't make its own products the world's greatest market cap is, simply, stupid.
What you are seeing with nVidia is a combination of "market exuberance" - FOMO - and plain ignorance in understanding nVidia's place. Do they sell a lot? Oh, heck yes. But that can't last forever and, as essentially designers and not manufacturers, they are confined by other business's functional abilities.
Where have you been ?
Havent you noticed a trend in todays society where bullshit trumps reason every single time.
The most valuable companies on earth TODAY all sell bullshit as their primary business practice. By bullshit i mean advertisements.
The leaders of today dont actually have any knowledge or practical experience. By leaders i mean corporate leadership, basically none of them could actually get a job as a regular employee at their own companies. THey couldnt even get a job as a cleaner, and yet the media falls over making heroes out of people whose only ability is to lie and talk bullshit.
Being "fabless" is perfectly normal. Most high end CPU and GPU companies are fabless, with the notable exception of Intel.
Manufacturing chips, and designing high end chips, are very different skills. It's perfectly reasonable to specialise.
The same way that no electronics company makes it's own PCBs, that gets outsourced to a company that specializes in that process. (The electronics company likely populates the components onto the PCB, but doesn't make the actual PCB).
As far as capacity limits go, of course every manufacturer has capacity limits. You can plan and invest to avoid those issues. This is the same for in house and outsourced manufacturing.
And similarly, the natural disaster risk applies both to in house and outsourced manufacturing.
Anyone who'd experienced what they did with GPU virtualisation won't have been shocked at the AI subscriptions.
It used to be enough to just charge eyewatering money for the 'special' version of what was basically the same commodity GPU, but then they worked out they could charge the inflated price *and* slap a hefty license subscription on top too.
Left a bad taste then, still hasn't improved. Shame there isn't enough competition to keep a lid on the wallet plundering.
CUDA, that four letter word that means I have to buy an NVidia graphics card for my development work and pretty much pay whatever they are asking. If I was the competition I would be pouring money into coming up with something free, open and good that sets customers free from this particular monopoly.
If it wasn't for CUDA I'd buy an AMD card, even though they're not as good they're prices don't take the micky.
You're absolutely correct, but this isn't simply a matter of developing the API for it, adoption is the issue here.
Developers need to implement it, and at the moment the only good reason to pour time into that would be if the performance was better than CUDA.
The nvidia market share is big enough that brand interoperability is not a good enough reason to implement and optimise a different API.
This has to be harder than it seems, seeing how picky applications can be just operating on slightly different CUDA versions.
And don't underestimate how much money nvidia is pumping into CUDA, this is a key component to keeping their monopoly.
AMD Rocm is slowly getting there but if people keep buying cuda just because it’s easier, then it’ll take longer to be better. Rocm needs open source community/dev support - yes , it’’s free, but that also means that it’s not as ‘sexy’ in the short term because don’t have the billions that nvidia has to hire developers. The price of freedom isn’t always ‘free’ as they say.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
