[Raqia]

Originally Posted by Ethatron

Yeah, it's like saying the chinese population shrinks because of a baby-wonder in India, which is ridiculous. And it'd be rather pointless to make a growth program to secure your relative rank - there is really nothing constructive and meaningfull you can do to "stop" relative contraction - and I'm not sure there exist a sane motivation to do so (just wanting to be the biggest seller in X belongs into my quote category).

My point here is that there's opportunity cost for traditional PC makers to keep investing in what they're doing over supplying parts for more disruptive gadgets. This is obviously better for AMD since it just can't compete in the PC supply chain w/ Intel and can leverage its superior mobile core design and low idle wattage GPUs doing something besides just laptops. It's also better for Intel since it's well past the point of diminishing returns for its investment traditional PC CPU designs, and it could put its fabs to better use making higher margin parts.

[Alexko]

Originally Posted by Raqia

My point here is that there's opportunity cost for traditional PC makers to keep investing in what they're doing over supplying parts for more disruptive gadgets. This is obviously better for AMD since it just can't compete in the PC supply chain w/ Intel and can leverage its superior mobile core design and low idle wattage GPUs doing something besides just laptops. It's also better for Intel since it's well past the point of diminishing returns for its investment traditional PC CPU designs, and it could put its fabs to better use making higher margin parts.

Both AMD and Intel are investing in mobile architectures (Bobcat/Jaguar, Atom, Intel's recent digital radio announcements, etc.) but as far as I'm aware, margins are much lower in the handheld market than they are in the PC world, let alone servers. And competition is so rough that I don't see them increasing any time soon.

[fellix]

AMD today announced their Third Quarter Results, revenue for the third quarter ended September 29, 2012 and is expected to decrease approximately 10 percent compared to Q2. The company previously forecasted third quarter 2012 revenue to decrease 1 percent, plus or minus 3 percent, sequentially.

The fall in revenue is directly related towards a number of issues but primarily a “challenging macroeconomic environment,” as the company states in his report.

The challenging macroeconomic environment that AMD denoted obviously are product demand, low average selling prices for Trinity APUs which have been launched cheaper than expected (again due to demand). Next to all that there has been an inventory write-down of roughly $100 million, due to lower anticipated future demand for certain products.

Simply put, it is not going well with AMD, the pending Microsoft Windows 8 OS release has a negative effect on demand for PCs and related products as well. Very simply put, lots of you guys are waiting and thus delaying the purchase of a new PC until Microsoft Windows 8 debuts. We can only hope that things will pick up for them and with the pending releases within the A-Series, the FX series update and Hondo APUs with hopefully a big hit for them on the Sea Islands GPU release in early 2013.

Source

[silent_guy]

Ouch, painful numbers. If the general market weakness is true, it will be hard for Nvidia to escape some of that as well.

Markets are not reacting very friendly today. At a round $2B market cap, Nvidia can easily buy AMD with cash/short term investments on-hand.

[UniversalTruth]

Very simply put, lots of you guys are waiting and thus delaying the purchase of a new PC until Microsoft Windows 8 debuts. We can only hope that things will pick up for them and with the pending releases within the A-Series, the FX series update and Hondo APUs with hopefully a big hit for them on the Sea Islands GPU release in early 2013.

Yup, sure, we are waiting for 200$ Radeon HD 8970. If that happens and they are able to produce enough to satisfy the demand, then everything will be fine.

[Alexko]

Originally Posted by silent_guy

Ouch, painful numbers. If the general market weakness is true, it will be hard for Nvidia to escape some of that as well.

Markets are not reacting very friendly today. At a round $2B market cap, Nvidia can easily buy AMD with cash/short term investments on-hand.

I'm not sure how profitable it may be, but Tegra should at least bring in a good bit of revenue, and so far that market seems to be growing fast.

Plus, NVIDIA's quarter will end a few days after the launch of Windows 8, so they might see a small uplift from that; but perhaps not enough to make much of a difference.

[sheepdogexpress]

http://semiaccurate.com/2012/10/12/a...t-engineering/

This is too much. I think AMD is done.

[jimbo75]

Tbh it's time AMD died. It's clear that boardroom decisions are a joke - they are so bad that I've often wondered if they are actually intel double-agents.

They are too small to be capable of reaching the heights of the likes of Intel, and too big (bloated) to compete with smaller, nimbler companies.

The good news is x86 doesn't have the stranglehold it used to have, and Intel's position is much weaker because of that.

AMD should basically lay-off 75% of their workforce right now (all from cpu and marketing) and continue as a discrete graphics company. Or maybe the old ATI guys should just start again.

[Alexko]

Originally Posted by sheepdogexpress

http://semiaccurate.com/2012/10/12/a...t-engineering/

This is too much. I think AMD is done.

AMD laying off 30% of their engineers sounds absurd. Sadly, Charlie has been predicting these things pretty accurately. I would still like to believe that they're not this crazy.

[UniversalTruth]

And also this from the other thread.

Originally Posted by iMacmatician

From ExtremeTech: "AMD executives have forgotten how to run a company."

The company’s old CFO and interim CEO, Thomas Seifert, left in September. Bob Feldstein, the man behind AMD’s Xbox 360 and Xbox Durango (720), left in July. Ben Williams, one of the original Opteron developers, left at the same time. Demers, CTO of the graphics business, left in February. All of this comes after AMD’s decision to fire 12% of its workforce last year, including some of the engineers that were critical to its success in the turbulent 2007-2010 period. And those are just the biggest names; our sources across the industry tell us that AMD resumes have poured in at every level.

Because folks, that’s what we’re down to. Executives, including executives in charge of long-standing projects and major relationships, have fled. The core of engineers and designers who worked in the trenches on K7, Opteron, and ATI cards are mostly gone. Read has publicly given up on keeping pace with Intel. AMD and GlobalFoundries have dissolved the bonds of ownership at a time when Intel claims its IDM status is what allows it to build cutting-edge hardware. Most ominously, AMD has begun acting like a company that’s terrified of its own investors rather than partnered with them.

Originally Posted by jimbo75

Tbh it's time AMD died. It's clear that boardroom decisions are a joke - they are so bad that I've often wondered if they are actually intel double-agents.

There is something dirty going on there. When AMD acquired ATi, the joint company didn't grow but instead shrunk very rapidly. Like someone manipulated its value, perhaps I don't understand very well financial/ economic side of things but the united company should have been stronger not weaker. Even with the new debt. This debt (or ATi itself) was the initial point for all present troubles.

Originally Posted by jimbo75

AMD should basically lay-off 75% of their workforce right now (all from cpu and marketing) and continue as a discrete graphics company. Or maybe the old ATI guys should just start again.

Originally Posted by jimbo75

They are too small to be capable of reaching the heights of the likes of Intel, and too big (bloated) to compete with smaller, nimbler companies.

Sadly true.

Originally Posted by jimbo75

The good news is x86 doesn't have the stranglehold it used to have, and Intel's position is much weaker because of that.

I see in neither of the articles Charlie or Joel Hruska mentioning that someone wants to buy AMD. Will they simply allow it to bankrupt, or simply exist but as a small company about which we hear once in a year?
What about the competition law? Will they allow intel sit alone in the business?

[3dilettante]

Most ominously, AMD has begun acting like a company that’s terrified of its own investors rather than partnered with them.

This part of the article isn't totally fair. AMD has a long history of being slimy to investors.

Originally Posted by UniversalTruth

There is something dirty going on there. When AMD acquired ATi, the joint company didn't grow but instead shrunk very rapidly. Like someone manipulated its value, perhaps I don't understand very well financial/ economic side of things but the united company should have been stronger not weaker. Even with the new debt. This debt (or ATi itself) was the initial point for all present troubles.

AMD bought ATI for far too much, right before both companies launched embarassing next gen products. We saw some of of the same investor crap being called out in this article as new going on even then, and if my memory were better I could probably point out examples from before that.

I see in neither of the articles Charlie or Joel Hruska mentioning that someone wants to buy AMD.

Engineers can be poached and (some) patents bought. The organization that is known as AMD is not an asset as we can see.

Will they simply allow it to bankrupt, or simply exist but as a small company about which we hear once in a year?

Read hopes for the latter. He pins his hopes on being a nimble design company that can help customers create leading SoC products. I'm not sure if I follow what exactly AMD is outsourcing.
If the company is supposed to help design and build products with outsourced engineering, why have the middle man?

However, there isn't much sign at this point that AMD's design chops are as compelling as they need to be to support it at its current size.

What about the competition law? Will they allow intel sit alone in the business?

Intel will point to ARM and companies like Apple and Qualcomm, et al.

[eastmen]

MS should buy amd. They can use them for their devices division

[AlexV]

Originally Posted by 3dilettante

AMD bought ATI for far too much.

Ridiculously so, and this may be part of the core of their issues. The deal was dolt-ish (far too cash heavy), and it was clearly a result of either panic (why?) or interesting pressures being exerted. I'm sure that people will point out that the GPG is the only group that's sort of performing within AMD, but that's really not saying much when factoring in just how overvalued ATI at the time of acquisition (the trillion or so goodwill write-offs hint to it strongly).

Also I'm not convinced that if they'd have taken their business to a different, much cheaper, shop (think S3) it would have affected their core model too much. AMD hardly needed an expensive high-end GPU maker just to be mini-Intel and have its own "eco-system". Also, moving into a cutthroat competitive situation with NV didn't help.

Fundamentally, a big part of AMD's wreckage is due to its managing staff being rather awful at anything but myopic, reactive stuff. Ruiz would've been a good chap to manage fabs, but just that (having a preeminently Ops man decide upon acquisitions is unwise, given that he'll want to get everything in-house and he'll want the expensive best), and Dirk would've been better off left to manage small / medium-sized teams of engineers as opposed to figuring out where to place the company's chops (he was truly awful at that, if we look at the current situation which is strongly influenced by the calls he made). It's unclear where Read fits in, although I honestly think that him doing some cobweb cleaning in the upper tiers of the hierarchy and the super-underperforming areas (see the Marketing dept. cuts) was a good call. It remains to be seen if he's anything more than a one trick pony because if you slice up everybody you end up alone.

[Alexko]

Originally Posted by eastmen

MS should buy amd. They can use them for their devices division

MS could use AMD, I suppose. But AMD's interests aren't perfectly aligned with Microsoft's, so they might end up gutting the company.

That said, AMD dying is in no one's interest, except Intel and Nvidia. So in the event of really severe financial trouble, I think AMD could turn to a few big players for investments. OEMs certainly come to mind, as they really need competition in the CPU, chipset and graphics markets to get good prices. And here Rory Read's ties to Lenovo make the latter a good candidate, especially since it's recently become the world's largest OEM.

Microsoft would be a good candidate too, they've got a lot of money and need the PC space to remain healthy, could use AMD's mobile APUs in their Windows tablets, not to mention AMD hardware in the Xbox.

Samsung and IBM both make AMD machines (servers for IBM) and benefit from the silicon volume generated by AMD in GlobalFoundries, which is part of the IBM Alliance to which Samsung also belongs, so some money could come from them too.



But actually, apart from the recent contraction of the PC market, AMD is doing OK. Trinity is better than Llano, Vishera is an improvement on Bulldozer, and GCN is great. The HSA foundation has reached critical mass (almost all the big ARM guys, Apple excepted) and even their Gaming Evolved program is finally yielding results. Their software tools may still be somewhat immature, but they have their first really HPC-worthy architecture, with all the right features (easy-to-compile-for, scalable architecture, ECC in all internal and external memories, fast DP, etc.).

Their roadmap looks quite good too. Not "earth-shattering, oh my god Intel and Nvidia are doomed" good, but definitely "back to black" good. Steamroller looks good enough to significantly improve AMD's competitiveness and margins, Temash checks all the right boxes for the Windows tablet market, to the point that I'd almost want to buy one. And Papermaster's SOC approach with HSA is arguably the smartest way AMD could go for the booming mobile market.

But if the rumors are true, Kaveri is massively late and AMD's about to fire 30% of the people who actually make stuff, none of that will matter. I just find it surprising because apart from a few recent missteps—I think their marketing department really really fucked up with Southern Islands—they've been steadily improving in terms of execution for the last few years.

[jimbo75]

Too much rot has set in, that's the problem. It's going to have ramifications for the entire industry. All those OEM's and tech press who were surviving on intel + nvidia bribes just found out that is no longer an income stream. Hell mend them. With no competition there will be even less incentive to improve, and even less incentive to upgrade for buyers with those two commanding what prices they want and segmenting at will.

This is the end of desktop. Pick your favourite games console, that's where I'll be going.

[AlexV]

Originally Posted by Alexko

But actually, apart from the recent contraction of the PC market, AMD is doing OK. Trinity is better than Llano, Vishera is an improvement on Bulldozer, and GCN is great. The HSA foundation has reached critical mass (almost all the big ARM guys, Apple excepted) and even their Gaming Evolved program is finally yielding results. Their software tools may still be somewhat immature, but they have their first really HPC-worthy architecture, with all the right features (easy-to-compile-for, scalable architecture, ECC in all internal and external memories, fast DP, etc.).

Their roadmap looks quite good too. Not "earth-shattering, oh my god Intel and Nvidia are doomed" good, but definitely "back to black" good. Steamroller looks good enough to significantly improve AMD's competitiveness and margins, Temash checks all the right boxes for the Windows tablet market, to the point that I'd almost want to buy one. And Papermaster's SOC approach with HSA is arguably the smartest way AMD could go for the booming mobile market.

Haven't we constantly been performing this exercise throughout the years? Their roadmap almost always looks sort of good (to some people), they're always on the verge of getting back into the black, and they always have some really cool products that are really just one cycle away (Phenom ring a bell? How about BD?)...albeit on this occasion Steamroller looks like nothing special already, whereas prior efforts at least seemed neat from a distance. The sad reality is that they've fallen back to irrelevant server market share, are constrained to scraping whatever scrap falls from Intel's table in desktop and are sort of looking good in the dying space of netbooks.

They've shown no evidence of being capable of solving their operational / productive woes, and are therefore viewed as a pretty wobbly supplier. HSA is a composed of entities who have never, in their entire lifetime, managed to setup a single successful software eco-system...which is a good match for AMD who for its part has been rather poor, to say the least, in what regards software, but is an awful match for the underlying idea of HSA which is intrinsically tied to setting up such an eco-system. "If you build the hw the software will come" really doesn't cut it in the HSA context, and neither of those companies has shown any capacity to do anything but that. Also I'm at a loss with regards to where this critical mass is, beyond logotypes showing up on the HSA Foundation site.

We're now roughly a quarter after its introduction and they've yet to present any single novel documentation, they still link to a decrepit AMD tool-chain, and are still constrained to posting videos of Phil Rogers (a very bright and neat fellow, by the way) and saying that "HSA will be great, really". At AFDS they weren't willing to answer in any proper detail any of the interesting questions, and the most often uttered line pretty much amounted to "It's magical therefore it'll word". It's also unclear if by the time they manage to deliver anything but the above that will translate into any noteworthy source of competitive advantage, or as a means to accrue revenue. Take a peek at Ivy Bridge, it's already more integrated than anything AMD offers (seriously, communicating through cache >> onion and garlic salads). Haswell takes that further.

There's no guarantee that by the time AMD gets an achievement unlocked with the a pointer-is-a-pointer line they're pushing (a rather bad paradigm to begin with, in the context of wide parallel programming, but I digress), Intel won't have already been there and done that. They may toast NVIDIA, who lacks a pretty important ingredient in this whole soup, but ultimately that's somewhat Pyrrhic, if it ever comes to be.

[AlexV]

Originally Posted by jimbo75

This is the end of desktop. Pick your favourite games console, that's where I'll be going.

No it's not, AMD has been a fringe-player, trend-follower for a long time now and the desktop is alive and well. Even if they were to implode (which is hardly a given, and somewhat unlikely to be honest), the desktop would still be fine. People need to take a long hard look at Intel's business model, their margin model and their OpEx and the realization will come that any trouble that AMD can cause in its current state (or any state, for that matter, given their always constrained production capacities), is mere background noise compared to the challenge of keeping that particular behemoth going. They're not innovating and implementing a particular price structure out of fear of what AMD can do (the latter have been way behind for years now), they do it because otherwise they'd implode themselves. So as beautiful and motivating as hero doing battle against evil empire stories are, it's really not AMD's fortunes that's keeping the market going and alive.

[jimbo75]

Originally Posted by AlexV

No it's not, AMD has been a fringe-player, trend-follower for a long time now and the desktop is alive and well. Even if they were to implode (which is hardly a given, and somewhat unlikely to be honest), the desktop would still be fine. People need to take a long hard look at Intel's business model, their margin model and their OpEx and the realization will come that any trouble that AMD can cause in its current state (or any state, for that matter, given their always constrained production capacities), is mere background noise compared to the challenge of keeping that particular behemoth going. They're not innovating and implementing a particular price structure out of fear of what AMD can do (the latter have been way behind for years now), they do it because otherwise they'd implode themselves. So as beautiful and motivating as hero doing battle against evil empire stories are, it's really not AMD's fortunes that's keeping the market going and alive.

You can't just remove a 25-30% market player and expect nothing to happen. I'm mostly concerned by Nvidia than intel anyway.

Those people who think Trinity is bad value because of cheap Radeons are going to find out just how good value they actually were, because there might not be any more cheap Radeons. When AMD leaves the market, the entry-gaming point will probably be raised by $100. That is going to kill my business for one.

[AlexV]

Originally Posted by jimbo75

You can't just remove a 25-30% market player and expect nothing to happen. I'm mostly concerned by Nvidia than intel anyway.

Those people who think Trinity is bad value because of cheap Radeons are going to find out just how good value they actually were, because there might not be any more cheap Radeons. When AMD leaves the market, the entry-gaming point will probably be raised by $100. That is going to kill my business for one.

Yes you can, when it's just a fringe tolerated player. It will leave Intel in a conundrum, no doubt, and I'm sure they'd rather not bother with that, as evidenced by their tolerance of AMD's existence within strict boundaries, as opposed to them going and slicing their throat. Trinity is bad because it had a shoddy launch schedule (praying that people would suddenly start buying Llanos was ultimately dumb) and because it is scarcely available in the space where it's most appealing, which is laptops. AMD's throng of claimed design wins has translated into a very limited sampling of products being available. Also, let's move out of the emotional "Oh man, AMD is keeping the market honest, everybody else is evil". They're all corporations, they're not people, and they're not evil or good or whatever. AMD sells its stuff for as much as it (assumes? hopes? has established that?) can sell it for, in quantities that meet some revenue targets that are set correlated with their expenses et al. They'd sell 1000$ SKUs in a second, if they had anything for that space, but they don't.

They're not keeping the market honest, or being kind to consumers or whatnot, they merely have a product stack that constrains them to particular segments. As evidenced by recent developments, the whole "Sweetspot" strategy was a big load of gunk that was meant to put lipstick on the "we need to engage in super competitive pricing to stall and perhaps reverse share erosion; we're also far more financially constrained and can't afford risky, top-gun projects" porcine unit.

[jimbo75]

It's got nothing to do with keeping the market honest or anything else. It's simple competition and although they put not much up vs intel, it's still better than nothing. They own a very large chunk of desktop share - just about all of it under $500. It's a huge ~40 million chip market that they own that intel is going to have to fill. A lot of those people were happy with cheap AMD quads, now they are going to be asked to be happy with cheap intel dual cores. More likely they'll end up buying a console.

Like I said, I'm more concerned about Nvidia though. Every year they struggle with late cards with process issues, and the reason for that is they are forced into it by AMD. With no competition, they'll just fall back into a comfort zone with the highest yields possible.

We're looking at total stagnation.

[AlexV]

They own precisely jack, I'd have trouble identifying even a single captive customer that AMD has...mainly because they offer no ancillary services to induce that captivity. They have no software stack worth noting and they offer no specialist services. NVIDIA tries to trap you with FUDA, Intel gives you a pretty fat toolchain and the guarantee that you won't get fired, AMD gives you the address of the nearest place of worship so that you can pray that someone will make the tools that you need, and that competitor's devrel efforts haven't hamstrung their HW that you just bought.

The fringe market-share they hold (which is actually under 20% as we speak in desktop and under 5% in server) can easily be serviced by Intel, but doing so would put Intel at risk of over-commoditization, so they choose (emphasis on choose) not to pursue it. If AMD were to go kaboom, Intel would service it. And no, that has nothing to do with them pricing everything and 1000USD and laughing all the way to the bank, as some (naively) assume. Read what I posted above, Intel's main problem is keeping Intel running within optimal parameters. Their entire business model is built around the idea of high-margins and high-revenue. If the revenue stream is gutted, they're in a very tight position to say the least. By way of consequence, they cannot price themselves to the moon, there are constraints in place as a consequence of what the market will accommodate. Otherwise stated, they can only bump up the price within the space of whatever consumer surplus there may be accumulated, beyond that they will start negatively impacting the volume of sales, which has a negative impact on their business.

The same holds for NVIDIA, for obvious reasons, albeit NVIDIA isn't in a similarly dominating position by rapport with AMD. They cannot price themselves to the heavens as they are reliant on a certain volume of revenue for continued existence (within the confines of a particular margin model, of course). Whatever shuffling would occur as a consequence of AMD opting for building fishing gear would be well within acceptable limits for the market. It may ultimately not be acceptable to you, but that's hardly a herald of the desktop dying. The habit of killing the cow that perpetually gives you milk (and by extension cheese and ancillary products) just so that you can have a steak once is pretty dumb, especially when there are no other cows to be had in the land. I would operate under the assumptions that planners for multi-billion dollar companies aren't (all) dumb. In this context, the claim that the existence of limpy AMD is what dictates the currently encountered pricing structure is flawed.

[jimbo75]

Do you really think that Nvidia would continue to press themselves onward and upward with absolutely no competition? That they'd continue to struggle with yield and capacity problems, losing money, just so they could give us all faster graphics cards every year? They are making absolutely nothing in client gpu and that's obvious. If AMD goes tits-up as looks likely, Nvidia has no reason to continue with this loss making (break even at best) and will either slow down development or (more likely) lower performance to maximise yield.

As for intel, they'd rather not be releasing any HD 4000 budget chips. And with AMD gone they won't. They will take the chance to bump prices by $10 or $20 however. That is assuming they even bother taking AMD's place - with margins so low why would they bother when they can do a lot of damage to Nvidia just by leaving the lower end? What does intel have to gain from taking AMD's sub-$500 desktop share? There is no money there, that's half the reason why AMD can't make money. I half believe that intel only releases chips at this level in order to keep AMD down - if that's true then by simply pulling out of that market space once AMD is gone, they can cause a lot of damage to Nvidia.

[AlexV]

Originally Posted by jimbo75

Do you really think that Nvidia would continue to press themselves onward and upward with absolutely no competition?

In brief yes. Mainly because it's their core market, and will remain so for quite a while (JHH hopes for Tegra to come in and help, but that's the future). They are a company, they need a revenue stream. It's complete and utter gunk to claim that they're not making anything off of the consumer market or that it's a loss generating endeavour - stop reading websites that claim that and have been claiming it for years, with financial figures never even hinting at it (but you don't actually need data when you have belief, I presume).

Also, they're not giving you anything, they want to repeatedly get your money because otherwise they're screwed. In order to repeatedly get your money, they need to incentize you to give it to them, otherwise you'd just keep it in the safe or spend it on booze (they want the benefit of getting one of their GPUs to outweigh the cost of opportunity attached to this choice). Selling you a faster, shinier something, has that effect. They stop doing that, or charge you so much so as to have the benefits fall under the cost of opportunity, they're screwed, because they no longer get your money and they need it to cover OpEx, CapEx et al., whilst also generating some neat extra money for all stakeholders.

Same reasoning applies to Intel, they're looking at expanding their array of revenue sources, which is a tough proposition going forward, and you assume they'd pull out of a market that's lucrative for some definition of lucrative. Really? It's AMD keeping this whole construct together / saving us from being price-raped / innovation starved? Hardly.

[jimbo75]

How can Nvidia possibly be making money on client GPU? The figures couldn't be more clear - they are being kept afloat by Quadro and have been for years. In the past 4 years they have averaged a yearly profit of $184m. With the kind of profits they make on professional cards, there is simply no chance that they are making any on client with those figures.

Again you are assuming that the low end market is lucrative for intel. It's not lucrative for AMD so I have real doubt's that it's lucrative for intel. On top of that it is clear that mobile is their main focus so given the choice to put more wafers behind Atom, or more behind Celeron and Pentium, what do you think they'll choose?

[Alexko]

Originally Posted by jimbo75

How can Nvidia possibly be making money on client GPU? The figures couldn't be more clear - they are being kept afloat by Quadro and have been for years. In the past 4 years they have averaged a yearly profit of $184m. With the kind of profits they make on professional cards, there is simply no chance that they are making any on client with those figures.

Again you are assuming that the low end market is lucrative for intel. It's not lucrative for AMD so I have real doubt's that it's lucrative for intel. On top of that it is clear that mobile is their main focus so given the choice to put more wafers behind Atom, or more behind Celeron and Pentium, what do you think they'll choose?

Putting NVIDIA aside for now—because this has already been discussed many times—Intel generally has much smaller dies competing with equivalent AMD products, and no need to pay for a foundry's margins since they make their chips themselves.

To be sure, low-end products are less profitable than high-end ones, but I'm confident their entire lineup enjoys quite comfortable margins.