You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
If this guy (Altman) has the $1.4tn to spend he claims to have (and there's booming demand for his company's stuff), might I suggest that no US taxpayer funding is needed !
And if he doesn't, that there might be some trust issues worth investigating before handing over any large taxpayer cheques.
OpenAI is the obviously unprofitable one. Apparently losing $8.5bn on around $12bn sales for 2025. And with profitability not forecast until 2019 earliest. Assuming the reported values mean what we think they do (a glance at OpenAI's corporate structure chart suggests caution may be in order)...
This feels like a circular argument (not unlike the circular AI funding). He's arguing that if China isn't allowed to buy nVidia GPUs, then China will develop their own GPUs and ecosystem and race ahead and beat the US. But how does that make any sense ? If China could do already do that, why...
Thanks, that's very interesting data. Looking at TSMC in isolation, it seems to have quite stable net profit margins over the past 20 years. Does this give the lie to the periodic suggestion that TSMC is exploiting an effective monopoly to overcharge their customers ? If I read the data...
ICF was Intel Custom Foundry - the previous iteration of Intel attempting to establish an external foundry business. Believe it ran from around 2010 to 2018. In its later/final stage it was using Intel 10nm (probably the most delayed process in Intel history). Believe they had a small number...
Right, so taking this at face value (it's directly from Intel), the key questions that occur to me are:
1. How does this 18A yield ramp compare to the ramp on the nearest equivalent TSMC process ?
2. Do we expect the 2nm yield ramp to be significantly faster than that for 18A ? If so, why and...
Hold on a moment. What's actually changed here ? The Intel share price has increased. And that's about it so far. It seems rather premature to be unfurling the "Mission Accomplished" banner (as a previous US president once did). Share prices go up and down. Sales and profits are what really...
Just reading the first part of your comment about purchase rebates reminded me of something - this all sounds remarkably like what Intel used to call "contra revenue" ! Though nVidia has different motivations for doing it (far less competitive pressure), the mechanism appears to be much the same.
"OpenAI's latest deals with the three companies include an ambitious promise to deliver 26 gigawatts worth of computing capacity using their chips"
Is the gigawatt the new unit of computing capacity then ? Leaving aside the dimensional analysis fail here, I'm curious exactly who dreamed up this...
I do hope the official history (if one ever exists) doesn't record Pat Gelsinger's claim of 5N4Y as fact. If they'd really achieved that (with real, full nodes all going into full production), they wouldn't be where they are today. I'd happily never read about the supposed 5N4Y ever gain.
Anyone know if this claim from the McKinsey report is correct ?
“About a third of the population in the semi industry is over the age of 55. You’re starting to see signals that some of that population is becoming less satisfied.”
That does seem very high. May be correct, but doesn't feel like...
Never trust a slide with unmarked axes. Instant lab fail when I was at school and university. Also, why do they use the wording "at similar performance/power" ? The normal wording would be "higher performance at the same power".
Probably so. But this is probably real cash (and extracted from taxpayers) rather than some whizzy, complex deal which few really understand and may be just a 21st century form of barter.
Not confident, based on past experience, that this is really going to move the needle for Europe. But...
Is it just me or do some of these creative financing arrangements (the CoreWeave/nVidia circular relationship being just one example and this feeling in some ways similar) display symptoms of a positive feedback system ? [OpenAI can't have all the $100bns they claim to be...
I'd be far more impressed if the goal was to reach definite targets in sales, margins and profits rather than a $1Tn stock market valuation. Fix the business and let the stock price take care of itself.
People have been saying this for some years now and it seems to make obvious sense. But that hasn't stopped nVidia dominance. Or is that just the training side ? So why hasn't that happened ? Is it simply the ecosystem (CUDA etc) constraints ?
For some reason, the first law of thermodynamics comes to mind as I read this proposal. After an injection of a mere $100bn only 3 years later (actually less since "by 2028") we are led to expect that "Taxpayers could make hundreds of billions of dollars". This, of course, is in addition to the...
Indeed, there is some quite staggering zero sum game "thinking" going on here with this policy.
There may well be some abuse of the H1B visa system, but there's just no recognition here of the fact that targeted skilled immigration has made the US the tech superpower it is today.
I'm reading...
