Very good sure but a must? People talk about China not catching up with the most advanced process like it's all that matters and they are lost without.
Look at TSMC, 20 and 16nm last year were 28% of revenues and this was a good year for them, share wise.
It is a huge investment, margins tend to be lower and because there are so few customers, you can easily end up with low utilization if you loose a key design. With TSMC, Samsung, GlobalFoundries and Intel, the market is overcrowded for the volumes as it is. China might be better off aiming for a long FinFET node with EUV in a few years, even if behind others. Who's gonna design on 3nm if costs keep rising, would be a highly competitive niche not worth the trouble for China. From a business perspective it certainly isn't a must but they do need to be there when a disruption that democratizes the most advanced node happens.
I know you believe in Intel but there is a real danger for them if AMD is competitive.
Intel still fully depends on PC and server.Those 2 segments were some 50 billion out of 59.4 billion revenue last year. Income wise, all else brings very little at best.
In PC , Intel has been pushing ASPs up to offset the decline in units, you can see that in their financial reports where they provide minimal details on ASPs.
AMD has some 1 billion in revenue from CPU/APU in PC while Intel has over 30 billion. If we think in terms of units share, it is misleading. If we think in terms of revenue share, it gets more interesting. AMD has some 3% revenue share in PC.
If AMD's Zen is competitive, they don't need great units share to shake up Intel in a big way. On top of that, what Intel did to ASPs, gives AMD the chance to easily undercut Intel's pricing.
We got 3 things that will impact Intel's PC revenue:
- the market is declining, lets say 5% per year for now
- AMD could compress ASPs, lets say by just 10% in 2018 vs 2016
- AMD could reach 20% units and revenue share by 2018
If we assume Intel's PC revenue at 30.7 billion in 2016 and we factor in the impact of those 3, Intel's PC revenue would decline to 20.6 billion in 2018.
It could easily be worse, the market could decline more if foldable phones take off and later on glasses will make it much worse but glasses are beyond 2018. ASPs could decline more and AMD could gain more share in certain circumstances - things like more cores in enthusiast or an Apple design win could enable better than 20% share.
All that, if AMD is competitive ofc,no debate there. It's also true that AMD could boost the PC market in 2017 and 2018 by offering more cores at reasonable prices and units could be up 7% this year with total revenues seeing a larger than 7% boost in 2017 and ASPs declining some in 2018.
In server Intel has over 99% share, a bit early to predict how much share they can lose to AMD and ARM short term and the market is growing but that segment for Intel could see flat or negative growth.
Overall it would be bad for Intel when you also consider what happens to margins. The new segments are not big enough yet to offset such declines.
As for AMD, they got 4 billions something revenue per year with some half of it from semi-custom. If competitive, they can double or triple revenue as they are very low today. Just 10% revenue share in PC, server and GPU would result in double the revenue when you add the 2 billions from semi-custom. For triple revenue this year, they need everything to go right and maybe grab some new opportunities, GPUs in server, wins in cars, some growth in semi-custom.
Fully agreed about China. Very doubtful about Amd, unfortunately not even with a better cpu and plataform (and enough capitals), Amd was able to hurt Intel much in past years. Now Amd is too little to push hard its new solution, Amd has not money to drive the OEMs to adopt the new cpu line.
Not to mention the lack of control on process is very bad in Pc space and a huge disaster in Server world.
It is a market with 1.35 billion people, hard to ignore.
There will be interesting startups from China and they could take the lead in some areas. Last month, Synaptics made an investment in http://www.oxitechnology.com/ to enable on display optical fingerprint sensing. http://www.eetimes.com/author.asp?section_id=36&doc_id=1330833 Thought it was noticeable for a global leader to invest in a China startup. We might see that more and more. Big names like Huawei are getting better every day while having a huge R&D budget, wouldn't underestimate them. And they aren't only about their own phones.
Foundry doesn't have to be at the most advanced node, even more so today when few can afford it. The push for scaling , at this point, is a bit ridiculous. A huge effort for small gains to serve a few customers. And it won't be enough for server, glasses, robots. Both manufacturing and the architecture must change. Too much data, too much power , crazy costs, the current pace of scaling can't be enough.
Memory will be tough until they get there but memory is changing too. In memory processing, neuromorphic and all that ,will change the landscape.
Auto and industrial could be segments where local firms can generate some nice revenue. China is the leader in electric car sales and local brands take it seriously, unlike others. Industrial is a big segment for them and China is pushing in this area.
Fascinating that folks rush into a conclusion on China's efforts.
It's not about the short term, if they catch up in 5 years or in 10 years, not much of a difference for them. They have a huge internal market and want a bigger chunk of the BOM. Today they spend a lot importing energy and technology enabled products. Renewables fix the energy problem while technology needs a push, so they push. It's not about the next 5 years, it's about the next 100 years.
The timing relates to China's growth trajectory in the last decades and there might be opportunities soon enough. If the world transitions to 3D ICs, NV memory, some might fall, others will rise. Good time to take a risk.
As for his year, lets see if AMD can double or triple its revenue, boost the PC market and drive Intel into a tailspin.