Hacker News Reader

Intel has never been a company focused on chip volume, they chose to exit the DRAM business 20+ years ago. They focus on high end high margin products, which is now primarily servers with the contraction of the PC market. This is how they justify greater than $10B yearly expenditures on fabs and R+D. They have maintained, and still do maintain a 2 generation lead on process technology.

Alternate architectures are not going to do much to dent the two generation lead in process.

IoT and Wearables mean nothing to Intel, there is not enough money there. Mostly a marketing effort to signal to their investors they aren't going to miss the next big opportunity like they did mobile. Of course that ignores the fact that Intel was never going to be much of a player in mobile. All the money in mobile is in the base stations, not in the handsets, and Qualcomm (3rd largest semi conductor company behind Intel and Samsung) completely dominates the IP in that market. What matters there is radio technology, not microprocessors.

And because of that, TSMC grows into the largest Pure Play company with market cap surpass Intel, and TSMC still has room to grow. PC market is shrinking, their Server market is healthy for now but if Zen is anything as decent as rumoured it will give Intel some competition. I dont believe ARM or PowerPC will give much trouble for Intel, at least not in the short run of 3 - 4 years time. Pretty much like Intel dont have much IP in mobile, no one will have the similar Server / Enterprise Grade IP in Storage, PCI-E, Memory, Networking and FPGA as Intel. All these matters a lot more then x86 these days.

I am not sure where the 2 Generation lead came from. It is more of a Generation lead. And it is shrinking, mainly due to Intel slowing down and other picking up pace a little. The 10nm from Samsung and TSMC next year wont be anything as good as Intel's 14nm. But their 7nm in 2018 will be better then Intel 14nm+. Of coz Intel will have 10nm by then, but it will be much smaller volume.

> IoT and Wearables mean nothing to Intel, there is not enough money there

Wearables maybe, as the market is relatively small as form factor-defined markets go, but I wouldn't be so quick to write off IoT, which describes a business model change more than a certain form factor or set of devices.

I listened to Murthy Renduchintala, President of the Intel IoTG group, speak at GE Minds+Machines earlier this week. He stressed FPGAs (acquisition of Altera) as well as SDN/NFV as key areas of interest for Intel going forward.

This is considering IoT from a semiconductor company's perspective; There will not be much revenue from the chips for this market. How much can they make per smart light bulb? You also have to note that Intel doesn't have semiconductor IP for nearly any radio protocol (BT, BTLE, Zigbee, they have wifi but not low end) necessary to build SoCs to win those markets.

IoT will be a big market, but the profits won't be going primarily to the semiconductor companies via chip sales. We won't likely see an "Intel Cloud Platform" or Intel selling IoT deployment services.

The FPGA tech is interesting, but mostly irrelevant to IoT. FPGAs can really benefit from Intels leading edge manufacturing, so that is a smart acquisition (even before considering IA integration possibilities).

Intel does not have much of a presence in networking, and there are a lot of competitors there who are more invested, and further, design their own silicon (Cisco, Juniper, Huawei, brocade maybe). Maybe if they are able to productize the silicon (as opposed to more exotic materials) optical transceivers they have spent a ton researching they would have something huge here.

All the money in mobile is in the base stations, not in the handsets, and Qualcomm (3rd largest semi conductor company behind Intel and Samsung) completely dominates the IP in that market.

Citation needed.

From the second half of your comment, I think it's possible you are confusing the base band processor[1] with base stations. Qualcomm does have a lot of IP in baseband processors, and has made a lot of money from them. However, baseband processors are on the handset not the base station.

It's also incorrect to think that mobile CPUs aren't a large market. They aren't Intel size, but they are quite substantial. The majority of Qualcomm's revenue is from CPU sales[2] (although SoCs like the Snapdragon include the baseband processor on the same chip)

[1]https://en.wikipedia.org/wiki/Baseband_processor

[2] http://www.investopedia.com/articles/markets/031116/qualcomm...

Intel really really needs a market they can dominate that consumes lots and lots of chips. IoT is one bet, Wearables (apparently was) one bet.

Yes. And Computer Vision applications will likely be that market. There were big algorithmic break-throughs in the last decade and things are now mature and robust enough to be finally useful. Plus some immediate obvious customers are starting to show up: autonomous cars, AR/VR, drones.

Intel has acquired companies in the space for quite a while now (announced and un-announced deals) and will likely continue to do so. Trying to be an important player in this space is the right strategy IMO.

If that market does play out and grow as much as predicted, Nvidia will be the company most likely to give Intel a run for their money.

Or put another way, why doesn't Intel acquire Nvidia?

The potential of being a competitor in an as of yet unproven market wouldn't be a good reason to acquire them. If that market becomes very very big AND Nvidia wins it, Intel could buy them in cash if they wanted, they certainly have it on hand.

Why they haven't acquired them in the past, Intel clearly doesn't think they are worth it and preferred to compete against them directly on graphics.

Why would they? In the bigger scheme of things computer vision is still pretty much a clean slate. Nobody's making any serious money (yet). And Nvidia is still first and foremost a graphics company. Buying Nvidia for CV tech would be like buying Airbnb for its "trips" business.

Apart from that the FTC / US Justice would never allow that merger in the first place.

Intel's IoT effort seems very strange to me. You can get Rpi or similar board for $5-25, while and Edison or Currie is a lot more expensive. I know dev boards is not where they want to make their money .. but surely the chips cannot price compare with the ARM SoCs on the cheaper boards. Fine .. you can run Windows on your embedded device with these things (Windows IoT Core ... but that also works on cheaper boards like RPi). Do they have a path to success in IoT or is it just like step 2 for the underpants gnomes?

Btw .. I feel the same way wrt Cisco and IoT. Perhaps the only large company that gets IoT IMHO is Amazon.

Intel's acquisition of Altera seems promising but I don't know how big a market that is. High-end applications are certainly price-insensitive but there is a lot that can be done with older gen hardware and the Chinese are creeping up on that already.

Well, Intel has a "product management disease" - when orders for new development come from Ivy league schooled management doctrinaires, and not obvious demands of the market. All American semi companies I know suffer this, and that's why Chinese eat their lunch.

> That said, the world could probably use a state of the art open access chip house that wasn't in Asia or controlled by the Chinese. Not sure that Intel could go there given their company history and organization.

Why should a chip house not be in Asia?

What are the issues with control by the Chinese [I assume you mean the Chinese government]?

The primary issues are trade tariffs, loss of intellectual property, and national security.

Trade tariffs for providing economic sanctions against countries you don't like, you could add $0.10 to every chip sold into the US as a tariff and pretty much instantly put the US into a recession.

IP theft is always an issue, and it depends on a local jurisdiction government for enforcement, so if the local government is lax in their enforcement you will experience more IP theft. Hence the whole Gonkai aspect of things.

National security because, as we've seen in research, inserting a capacitor in the chip that is not part of the circuit can create an undetectable back door in silicon[1] and that can compromise chips in sensitive systems. Again local governmental oversight is the lever to avoid such things but if you were in a state of cold war with the government in question?

[1] https://www.wired.com/2016/06/demonically-clever-backdoor-hi...

> Trade tariffs for providing economic sanctions against countries you don't like, you could add $0.10 to every chip sold into the US as a tariff and pretty much instantly put the US into a recession.

But that would harm China too.

For people not aligned with the US, what is the advantage of the US versus China with regards to backdoors? Especially after the Snowden leaks, it seems to be an issue of who is worse.

Of course, but who would it hurt worse? Rarely do nation-states engage in conflict and not also hurt themselves in the process, the idea is that it hurts the target a lot worse than it hurts at home. You might say that humans often start fights knowing they are going to get hurt but feeling like the fight is justified. China has economic controls that the US does not, which gives it the ability to dodge which the US doesn't have in an economic fight.

But many chips are bought in the US, so China would be hurt quite significantly (you're saying the US would be put into a recession which means the US buys lots of chips).

What is your estimation of how many chips are sold in the US versus sold in China and manufactured into products there? Tell me your understanding of what percentage of products sold in the US and sold in China use chips manufactured in China vs in the US?