Nvidia beats market expectations again, but for how long? Nvidia has turned in another set of sizzling results on the back of AI-driven demand for its products, yet industry watchers are increasingly wondering how long this substantial growth can continue, or whether the bubble is going to burst. The GPU giant reported revenue for its fiscal Q1 2025 (ended April 28, 2024) of $26 …
The money being spent by companies on buying AI will never be fully recovered. They have spent probably trillions now in total, very few will make any money back on what they have spent.
AI might well be worth money to deploy at some stage, but that's not now. Currently they are a mess and costing their deployees in time money and damages when it says something very stupid.
The truly scary horizon is the implosion of IT when AI in its current form goes bust. People are catching on that its a novelty, with very few real world uses. When the realities really hit the rug is doing to hit supersonic speeds.
“with very few real world uses.”
Off the top of my head: 90% of all graphic design (sorry, that’s just true); MidJourney is currently destroying 99% of stock and product photo industry; fashion photography = photorealistic rendered fabric + AI Metahuman model + MidJourney for the scene. Movie work: save 90% of studio time and costs by pre-rendering all directorial options in SORA *first* to decide what to film, before scheduling the shooting locations let alone getting a single actor on set, and that’s without even considering any CGI for the film itself. Legal: chugging through 50k pages of disclosure documents to find “smoking guns” - which is what most junior corporate lawyers spend much of their time doing for several years of their career after graduating from the top law schools at obscene hourly rates - remember, it doesn’t have to be *correct*, it just has to *highlight 1 paragraph in 50 pages which might be worth reading* and identify any references which the disclosed document implies must exist and yet aren’t in the disclosure pack - a lot of that work is just “pulling on threads”; accountancy (spending 5 years to know 2600 pages of tax law by heart, and getting paid £400ph to regurgitate it, does that sound like a good use of human brain cells to you?). Architecture: we’ve already reached the point that for the *average* job of building a house, a trained architect can get GPT to draw up plans to a few- paragraph prompt. We’re only one GPT version away from an app which allows a trades builder to draw up plans, a BOM, and a project plan, based on a client brief. Architects charge a lot less to review plans, than to draw them up.
And this is what GPT4, MidJourney et al can *already* do.
While I completely agree that "AI" as it exists today is mostly hype and greed/speculation, I don't believe IT as a profession or an industry will suffer long-term because of it.
In part, because while "AI" is not actually "I", regardless of the "A", the demand for compute capacity of all kinds is never ending.
There are still many many (!!) problems in the world which would be tackled by throwing CPU (GPU, whatever) resources at them. I can't imagine that will ever change. And that pile of gear will need wrangling by skilled hands and brains of IT types. I don't expect that will ever change very much either.
Unfortunately despite management fees to Big Pension providers like L&G and Standard Life ‘fund managers/experts in investment’ they won’t be exiting so look forward to a huge losses soon.
‘Over the long term stocks and shares outperform …’. When pressured on this.
Same weasel talk as why pensions shat the bed during COVID. They literally didn’t see it coming.. despite what played out Jan-Mar 2020 across the media.
So long as they are limited by TSMC's packaging capacity / HBM3E production capacity rather than customer demand. TSMC said they are doubling their capacity of the type of packaging Nvidia uses but it will not satisfy the demand they are seeing. Multiple DRAM OEMs are reporting their HBM production capacity is sold out for the next year or beyond. So there are a few more quarters at least of Nvidia being unable to satisfy demand.
In a normal market this would result in people turning to competitors, but as of now Nvidia has no real competition in the AI space. Again leading one to believe there are at least a few more quarters of Nvidia being unable to satisfy demand.
At some point production will no longer be a constraint, and/or some competition will emerge. Until then I'm holding onto the Nvidia I bought a while back and wishing like hell I had bought more!
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
