Market Currents
Intel (INTC) has bagged Altera (ALTR) as a chip foundry client. Altera, which has a near-duopoly...
-
Monday, February 25, 5:02 PM ETIntel (INTC) has bagged Altera (ALTR) as a chip foundry client. Altera, which has a near-duopoly on the FPGA market in tandem with Xilinx (XLNX), has hired Intel to make FPGAs using the latter's next-gen 14nm manufacturing process and 3D transistor tech. Altera says it will still rely on its existing foundry partner, TSMC (TSM), to make 20nm and older-process chips. The deal comes a month after reports Intel had struck a foundry deal with Cisco. ALTR +2.1% AH. XLNX -0.8%. INTC +0.5%. TSM +0.9% after falling 3.4% in regular trading.
Other date
TECH ETFs IN FOCUS
Latest Tech Articles
This news story has 18 comments:
http://seekingalpha.co...
Why will it be an uphill battle? The foundries are stumbling and having troubles bringing up new processes. I would think Intel is in the driver's seat on this one.
1) INTC is not going to be a pure-play foundry as they will always put Intel's IP first and thus not interest those who are more satisfied with the ARM ecosystem to make SoCs.
2) INTC process is and will be primarily geared towards their CPU products and thus will be hard for those willing to work with Intel's IP but still require custom in-house IP developed by their own engineers. Intel will need to devote resources for support for their complex processes.
The above is based on discussions with Intel and GlobalFoundry engineers.
Intel's process is geared to either high performance or low power, meaning a design can be optimized for one or the other. Customers who can now market a compelling product with ARM will likely continue to do so. They may not be able to market competitive products in the future if the foundries continue to stumble as they have. I think the technology gap will widen and ARM based designs will find themselves uncompetitive.
Yes, Intel will have to devote resources and they appear to be doing just that. Custom IP will have to be synthesized to Intel's process and mated to Intel's libraries. No doubt about it and putting that infrastructure in place takes time.
BTW, I don't know why you are including GloFo engineers in this discussion.
"I don't want to reveal anything specific of what they said but GF places themselves as a competitor of Intel although at this point they want to go after TSMC."
Thanks for the laugh. I've been following them via AMD for 30 years and they haven't met a single schedule in memory. They've had disastrous yields and even AMD has paid to get out of contracts with them. When AMD wants to get away from you, you know you're in real trouble. GloFo has the lowest credibility in this industry. That could change at some point but there are no signs yet beyond more foils, smoke and mirrors. TSMC is a step up but they're characterized by delays and yield problems as well. 20/22nm is going to be much harder and I'll believe trigate when we see volume shipping and not before. Both those guys were about 5 years late with HiK-MG and it's still only partial volume.
Bottom line, believe it when you see it.
That statement was definitely out of this world. But as always you gotta keep a straight face when talking to industry types as they're quite proud of their turf.
GF frankly has management issues. Don't know if they have totally rectified that but I guess we shall see very soon if they can back up their talk. If they can actually grab business from TSMC I guess you can call that progress.
A billion here, a billion there, pretty soon you're talking about real money.
Intel has to start somewhere and with the enormous design complexities of sub-45nm processes with double patterning and FinFET (TriGate) the pure-play foundries (e.g. TSMC and GlobalFoundries) are struggling to make the foundry ecosystem work with their own process capabilities, EDA vendors and customers.
Intel's process is geared towards their own CPU designs and tuned to improve the yields to maintain profitability. Changing this to meet the demands of SoCs will prove an enormous challenge. So what they learn with FPGA/programmable-logic chips appears pedagogical at this point.
Altera is mentioned as a one of the big clients of TSMCs but it probably is not a significant in Intels volumes
http://bit.ly/ZyuYOd