Until 2007 I would've agreed with you. But judging by recent history, most rating changes seem to me much more reactive in nature than proactive. Or at least they can be explained easily this way, while the reasoning that 'they know more than we do' is a bit harder to support after they got blindsided by the global recession back then.
-
We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.
AMD shut out of the debt market
Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
There is a spectrum of grades of "clean room" status. Clean-room equip are usually built in a "mfg grade" cleanroom environment, which is way better than say the quality of a Walmart or Amazon warehouse but no where near the leve of cleanliness that is required for the end-use production environment.
When a tool is installed it goes through a series of stages of clean up procedures, so the tool is cleaned up without dirtying up the production environment in the process. (think air-locks on a space ship)
One question that begs a deeper analysis is that the models used in CDO valuation worked right until the crisis, so while the rating agencies deserve a just criticism, it isn't just a matter of throw everything they said/did away.
One of their criticism is that they are too late to change a rating, just like they did with AMD, OTOH, there isn't much wrong with their baseline scenario. You may question one or other assumption but there isn't anything fundamentally broken.
They will always seem reactive because they will never attempt to call the top or the bottom, so the ratings changes will always occur after a peak or a trough.
The difference is the market builds in lots of superfluous peaks and troughs, we never know which ones are the materially significant (long-term trend setting) tops or bottoms. Whereas the analysts are paid on a daily basis to know when things are just doing the normal sideways range trading versus actually setting upon a new multi-year trajectory of higher highs or lower lows.
I disagree here. The problem is not that they are reactive, but that they issue downgrades or upgrades long after drastic changes happened, and we are talking about normal conditions, not black swan events.
Look at AMD. AMD is suffering since october September but in December, it was clear that Q4 would be hell and they would have to pay another WSA charge, but only now, two months later, they are putting the continuity of the operations in question.
To be fair with them they would have to rerun the models on each quarter to capture the changes, but that's the only way for them to become a credible and reliable mechanism of evaluating a company's creditworthiness.
Not so sure on that one. Part of the recession in Europe boiled down to the problem that essentially half the companies and states in the world were grossly overvalued if you only looked at their ratings. At times we had two or more articles a day in our newspapers about reevaluations and outlook changes. Their unannounced and infrequent changes could very well have done more financial damage than the actual crisis which wasn't all that hard in northern Europe. European politicians even thought about creating an european rating agency (not convinced that this is such a good idea though).
@Idontcare, it just strikes me as odd when you run with full speed into a giant concrete wall when you're paid to find out years in advance where such a wall might exist, how sturdy it is and which color it will have. Reactive corrections are such a simple explanation, it should at least raise an eyebrow of your inner engineer self.
I'm not sure what you are disagreeing with.
They issue changes long after a market tops or bottoms, they don't attempt to call the bottom or the top until they feel they have confirmation of it.
No different than standard technical analysis trading. You always wait for confirmation from other indicators before you have confidence in making a change to your existing standing in the market.
Trying to gauge the long term (5-10yrs) fiscal viability of a company on the basis of 90 days worth of revenue activity is a bit challenging. Going to be reacting to a lot of noise if that is your threshold for changing ratings.
I mean surely you must agree there is a good reason why nearly everyone who makes a living being a ratings house or analyst approaches market volatility in basically the same way (100 or 250 day trailing average approach, never a 10 day trailing average approach).
NTMBK
Lifer
Don't forget about total platform costs- and upgrades.
A few years back Opterons had a big advantage over Xeons, and lots of new computer installations were using AMD parts. Move on a bit, and they want to update these computer installations to use more modern parts. But they have a very large number of AMD motherboards installed already- so it works out far cheaper to simply drop new Piledriver based CPUs into the existing systems then to replace the whole thing. It's a complicated equation to balance.
Historically AMD had a larger than average advantage over Intel in markets requiring 4 or more sockets per system, as AMD's Hypertransport point-to-point topology scaled better to high numbers of sockets than Intel's FSB. But Intel brought out their own point-to-point system with QPI, combined with the excellent improvements of their Core architecture.
ShintaiDK
Lifer
AMD never reached the same amount of sockets like Intel. Simply because an 8 socket Opteron partly destroyed itself with coherency. Thats why 8 sockets+ was usually Intel. Even the FSB Xeons scaled to 32 sockets.
Newisys did however design Horus to fix this issues up to 32 sockets.
NTMBK
Lifer
Huh, I was always under the impression that memory bandwidth would kill the FSB designs beyond 4 sockets? I'm just basing this on retroactively read Anandtech articles, so I'm happy to learn. 🙂
Of course, it could well be a situation which is very dependent on what algorithm is being run. If there is lots of cache thrashing and working on overlapping portions of data, I can see that definitely being the case, whereas if you're most of the time working on separate portions of memory then you shouldn't have such a high synchronisation cost and point to point should work much better.
EDIT: That Horus is pretty cool to read about, cheers for the pointer.
A rating agency, that makes a living rating credit bonds, has the luxury of not calling at a top at a bottom, because they get paid the moment they release the report.
A bank or fund, whose money depends on being able to redeem the bond, cannot have the luxury of not calling at a top or at a bottom without being absolutely sure of it.
And once you go to CCC and below ratings, the situation is very fluid. It didn't happen one or two times, but it is almost SOP of the rating companies to downgrade companies to C when they are almost going into chapter 11 or chapter 7, even after months of a tough situation unraveling. When you are dealing with assets in which their very existence is in question, you *have* to measure the long term prospects of a company using 90-days data, even less if you have access. There will be no long term prospect if the CCC company doesn't deal with their short term issues in a timely manner.
That's why the rate issue by those agencies aren't a reliable measure of creditworthiness of a company.
The irony is, the more volatile the asset is, the more you would need help in assessing creditworthiness, and the less you can count the rate as a reliable measure.
No not at all but in the space/market they're targeting value consumer it's all about price. IMO, AMD is literally giving those mobile parts away because unlike Kabini, it's their APU / volume vehicle to help fulfill their end (is there any doubt that it's the short one?) of the WSA.
By the way, not that anybody cares except me :whiste:, intel capex per CPU shipped > $40, AMD's $1B commitment to Global Foundries is sub $20 per unit.
VirtualLarry
No Lifer
So you are saying that CapEx per CPU is that high? Given the volume that Intel does, I thought that it would be reversed, that Intel was spending lower capex per CPU than AMD, just because they do some multiple of AMD's volume.
If capex portion of the CPU is so high, and there is mfg, test, package, ship, plus profit in there, I don't see how that they can sell any CPU for $42.
I think that your numbers are wrong.
Because they sell other chips for $289.
His AMD numbers are wrong though. There is also the amount paid to TSMC to consider. (plus, probably other hidden costs at GFL)
CAPEX =! annual depreciation + amortization. You are assuming that Intel CAPEX levels are here to stay, or at least that revenue levels will be linear for in the future. Those are not given at the current moment.
That said, there is another issue. You are taking GLF commitments and dividing by them by the entire shipment number. This is wrong, not all their chips are manufactured at GLF. Using Q312 as a base, it's not too far fetched to see some 30 million GLF chips this year. assuming they will take the minimum commitment only, this brings COGS to around 36 per chip, which takes us to the gross margin issue. Gross margin for GLF chips will be in the high 20's or low 30's range.
And here I have to ask. When you have a COGS of 36 on top of your R&D expenses, how much worth is AMD fabless strategy compared to Intel foundry strategy?
No its not given, the current trend is even higher capex levels! Intels Business model is just plain bad for the future consolidating market, and they are going to share those investments or drown, excactly take the steps they do, but they have to rapidly speed up the process. Pablos point is very clear and valid there os no need to try to talk it down.
And then you are exactly bringing the arguments why Richland is here, to fullfill the minimum commitment at lowest loss. And as you know there is an end to it. We have been over this. Richland only makes sense in that context.
Are you seriously proposing AMD should be having fabs or what? Looking at how things have evolved with accellerating cost exiting the fab business was the only possible choice. They have nowhere the financial muscles to be there, now or in the future,- if there is a future for them.
This sounds harsh, but i dont think it matters if AMD is shut out of the debt market. Under alle circumstances AMD is living on the mercy of mubadalas wet dreams about a fab in the middle of the dry hot desert, and the continuing funding when they are near death.
How people dare to invest in such a untransparant mess is beyond me. How do you analyze your risk here 🙂?
How people dare to invest in such a untransparant mess is beyond me. How do you analyze your risk here 🙂?
Intel made a choice, and this choice is to build up capacity for the mobile market. If they can get a share of the mobile market for them with their 14nm chips, then they will a lot of extra revenue and CAPEX will become a non-issue. But if they don't get a share, yes, Pablo doomsday scenario will bite them.
Yes, I agree with that. I just pointed out that Richland might be even worse to GLF than Trinity was, and with that, this may offset some if not all Kabini gains.
No, I'm not saying that AMD should have retained their fabs. What I'm saying is that AMD is paying CAPEX burden through other means.
TBH,the whole world economy is the pan and we are going to be mired in debt for decades,with higher taxes and less services every year. More companies failing means more people out of work,a lower number of people paying taxes,more benefits claimed,even more public debt gained and only sadistic dorks would like that. Some people are full of glee at the prospects of certain companies doing "worse" like Apple for example, but its like seeing someone being hit by car , and thinking you had a lucky escape,but a train is heading right for you.
Now in a CPU section instead of actually talking about technology,people are so bored they seem to be talking less about tech, and more about the companies which comprise the world banking/investment industry, whose creative maths has helped push us into this black hole. Even healthy,very profitable companies making no losses are being screwed over, by the same people, because they don't reach whatever pie in the sky estimates they give out. Ultimately that leads to a few things,like job cuts/reduced recruitment drives,less company benefits and more constrained pay for people even in those profitable companies.
Now in a CPU section instead of actually talking about technology,people are so bored they seem to be talking less about tech, and more about the companies which comprise the world banking/investment industry, whose creative maths has helped push us into this black hole. Even healthy,very profitable companies making no losses are being screwed over, by the same people, because they don't reach whatever pie in the sky estimates they give out. Ultimately that leads to a few things,like job cuts/reduced recruitment drives,less company benefits and more constrained pay for people even in those profitable companies.
Last edited:
You were not supposed to reply
Intel 2013 capex = $13B per CFO. 2012 = $11.5B, 2011 $10.7B so it is fair to say that ~$12B is the new normal. Intel ships 300MM processors per year, thus capex is $40 per processor.
Now some would argue that capex <> depreciation but in the absence of substantial growth*, I would argue that it should be - that movie has been seen time and time again, most recently in oil and gas, and it catches up to most companies.
* intel's main business is x86 processors isn't it and that market is shrinking, isn't it?
I treated the $1B AMD has to pay Gloflo as an expense akin to Intel's capex burden.
You are right, AMD has the further advantage that part of the $1B includes factory operating costs, and the disadvantage of having to pay for wafers to TSMC on Kabini. But even adding a generous $10 per Kabini x 20MM units = $200MM which leaves a total of $1.2B over 60MM processors or $20 each. which is half of Intel capex alone, not including Intel's factory operating costs.
As another poster pointed out, what keeps Intel rolling is their high ASP. It's certainly not their low factory costs..
Ratings agencies are just covering their you know what. The market, which is far more rational as it involves hundreds of decision-makers, has hardly budged.
Though I hope it does so I can make a killing buying AMD bonds for 50 cents on the dollar...can you see the problem with these agencies? They're trying to influence a perfectly functional market.
Though I hope it does so I can make a killing buying AMD bonds for 50 cents on the dollar...can you see the problem with these agencies? They're trying to influence a perfectly functional market.
Why would you burden Kabini? The WSA does not burden Kabini.
Are you estimating 60MM Kabini processors per year? That's roughly what AMD shipped of everything in 2012, and more than twice what it shipped of Brazos.
The COGS numbers I brought here are for big core only, not for Kabini.
Foundry or fabless are also financial management strategies. I just pointed out that there isn't any inherent advantage to AMD because they are fabless. They don't pay CAPEX, but they pay in COGS.
Last edited:
I am quite sure kabini and temash is positioned correct on the market, and considering its growth market for both, shipping the same as the total of 2012 is not out of reach - far grom, - its going on the emerging markets. It all depends on the new atom, but i think its safe to assume its targeted to battle on the phones. If so, there will be an enormous market for both temash and kabini, because the slot between haswell and atom is huge, and exactly where the growth is comming.
It is what happens after haswell the real problems is starting for AMD, and they need to pull a rabbit from the hat 🙂 Intels slow and inefficient gfx part can not continue forewer, and when Intel gets the size down in half, they are getting at the powerlevels and size that can start to fight mid and high end tablets. All the 7w IB, and Haswell is just BS imho, its both to big and way to expensive, but in an generation or two Intel will get there. And its just a huge grinding machine.
I dont know how AMD is supposed to compete in that market, because Intel will sell all the way to the low end, because of marginal cost of all their future huge expensive capacity. Intel will get pressed because fighting ARM A7 and little-big with Atom is going to cost way to much, and they are fighting an ecosystem with far lower costs, and therefore - when they realize Atom is going to be the same success as Itanic, they will push AMD using the excessive capacity.
And mubadalas "ambitious" plan to move 10.000 german technicians to the middle of nowhere is going to look like a dream 100 years before the financial crisis.
It is what happens after haswell the real problems is starting for AMD, and they need to pull a rabbit from the hat 🙂 Intels slow and inefficient gfx part can not continue forewer, and when Intel gets the size down in half, they are getting at the powerlevels and size that can start to fight mid and high end tablets. All the 7w IB, and Haswell is just BS imho, its both to big and way to expensive, but in an generation or two Intel will get there. And its just a huge grinding machine.
I dont know how AMD is supposed to compete in that market, because Intel will sell all the way to the low end, because of marginal cost of all their future huge expensive capacity. Intel will get pressed because fighting ARM A7 and little-big with Atom is going to cost way to much, and they are fighting an ecosystem with far lower costs, and therefore - when they realize Atom is going to be the same success as Itanic, they will push AMD using the excessive capacity.
And mubadalas "ambitious" plan to move 10.000 german technicians to the middle of nowhere is going to look like a dream 100 years before the financial crisis.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 25K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 24K
-
-
