Some of you speculated why there appears to be a mass exodus of the top software talent from the Q. FWIW, this isn't just happening at the Q, it's happening at all chip companies in general, including Intel, Broadcom, NXP, Marvel, Nvidia, specifically the ones that are heavily involved in mobile O/S development such as Android. A lot of this is attributed to the change that has happened in the ecosystem that is impacting everyone. A few years ago, Android was in its it infancy, and much of the need features for wireless modem and connectivity had not been developed on the android platform. Google relied heavily on a lot of the semiconductor companies to provide not only hardware but also software stack, integration, and components into the Android framework. Often times, each semiconductor conductor company provided it's on proprietary solution, with the hope that such a software+hardware/chip solution would "vendor lock" an OEM to their own solution. Google, at the time, didn't really oversee this much, because at the time, Android was in its infancy, and Google was much more concerned with gaining market share, with each supplier and OEM left to figure out how to do some of the basic wireless connectivity/modem. What eventually happened is the Android ecosystem got considerably fragmented, as each chip supplier with their own proprietary solution started to vendor lock each other out, and created numerous of integration/interop/fragmentation issues. Google, in it's initial attempt to eliminate fragmentation, attempted to standardize the Android connectivity by taking opensource donations from companies that were willing to donate their software components to the Android ecosystem. Qualcomm, Intel, NXP, Broadcom, Samsung, Marvel, Nvidia all tried to push their own solutions as "open source", with the hope that their experience would give them an edge on the technology that was adopted by google over each other, and by providing custom solutions on top of that contribution that they made. And for the first part of the Google fragmentation fix Version 1, each chip vendor that won the opernsource contribution to Google did have a short/momentarily advantage versus all the other chip vendors, if only for a short duration. Since everything was opensource, eventually each vendor caught up with each other on the software component, and each vendor started to lose whatever software advantage one had over the other. Simultaneously, the Android ecosystem started to drastically evolve, as Google increasing got tired of dealing with all the chip OEMS, partly due to their lack of response time for new features, and partly because chip company continued to attempt to vendor lock each other out and continued to cause a fragmentation issue..So Google then took the fragmentation fix Version 2 to the next level.. It started to hire the top software talents from the chip companies that were responsible for the vendor-specific software opensource contributions. Google spent considerable resources and money on software engineers early in 2014 to start developing all the wireless/modem connectivity software needed for android in house, emphasizing new features and standardization across all chips and phone OEMs. From Google's perspective, chips and hardware are a commodity that can be easily swapped in and out, and for them software is the key asset. And as part of each subsequent release (K, L, and even M), it has amassed enough talent to develop so features in house, while at the same time spread the remaining feature development roadmap across multiple OEMs so none of them had an advantage over the other. And with all these features developed in the open source community, it ended up being a more turn-key solution for other chip companies that don't even need to make any software contributions to Android (a lot of the chip companies from Asia). So while companies like Qualcomm, Broadcom, NXP, Nvidia, etc did the initial heavy lifting of figuring out out the software, it quickly became commoditized and centralized at Google, and low cost chip players (MediaTek, Spreadtrum, Huawei,etc), don't need to spend so much resources developing their connectivity software, since what comes from Google and the opensource community works "good enough" out of the box. Phone manufacturers, seeing that the default AOSP solution from Google was now "good enough" no longer needed all the proprietary software/vendor locked solutions from any of the chip manufacturers. This is creating a numerous of issues on chip companies. (1) Phone OEM manufacturers are no longer vendor locked to a chip company's proprietary software+hardware solutions, making it harder for each chip company to differentiate from each other with software as a differentiator. (2)With Google now assuming a much bigger software role that more phone OEM's are willing to adopt out of the box without any chip vendor specific software "enhancements", there is less of a need for chip companies to provide proprietary software solutions that OEM's are moving away from. and (3)Chip companies end up having to compete more and more on cost and cost alone. Software engineers at chip companies working closely with the mobile O/S see the writing on the wall for proprietary software solutions as it becomes less and less relevant as Android O/S matures, which is probably why you see a lot of the software engineers in the Google connectivity group are from all over the place (Qualcomm, Broadcom, NXP, Nvidia,etc,etc,etc) and why there is an exodus of some of these software talent from chip companies.
6 replies (most recent on top)
This is a great thread. Thanks for sharing your insights here.
It comes right down to this. A software engineer from a top-flight school like National Taiwan University makes $12k annually + a fairly large stock bonus that can some day be worth $20k or $30k per year if they are lucky. And now MediaTek is no longer making 3G parts, they are actually shipping their first 4G parts with this tremendous advantage in lower costs of design, and state-of-the-art ARM CPUs included in the part, which are no worse than Qualcomm's CPU's because Paul Jacobs chickened out of making a 64-bit CPU in-house about 18-24 months ago, and both parts (Qualcomm 810 and MediaTek) are practically identical on the inside.
++++
Moreover the almost-2-decades Samsung/Qualcomm partnership is getting sour because Samsung is no longer printing money now that it no longer has a monopoly on CDMA or 5"-6" smartphones - the iPhone 6 has really nuked the Samsung Galaxy Product. Samsung is scrambling for leftovers in the cellphone market, and Qualcomm is beginning to scramble, too.
I don't really agree that this is all Google's fault. MediaTek has been chasing Qualcomm for half a decade, and execs have been worrying and talking about when they might have to start laying off some of their semi-retired 500+ VPs at Qualcomm in order to make a competitive product. Meanwhile, Samsung has been trying to "cut out the middleman" (Qualcomm) for half a decade, bringing modem and CPU design in-house. They have been using Qualcomm chips in the US market where customers are very demanding but shipping the worse-performing Samsung parts to overseas customers where customers don't know the difference. Now that Apple has finally gotten a little desperate and pulled its last sales trick from its sleeve - the 5" iPhone, Samsung (which is using cellphone profits to subsidize all its other divisions) is really desperate to save money as their Galaxy product line crashes in sales numbers due to iPhone 6 pressure, so they have "cut out the middleman" which is Qualcomm.
It didn't help that Qualcomm chickened out of CPU design for its 1st generation 64-bit part, the 810, and ARM sold them a lemon of a CPU core that overheats.
It's true that the software engineers in QCT just write drivers and do a really inadequate port of Android to get the handset makers started but the handset makers always wanted to put their lipstick on the final product so they could pretend they were not just Qualcomm minions. Now google is offering to adapt android to phones (Google Play Edition Phones) and still doing about 2-3 nexus devices a year, the handset makers realize they really don't need Qualcomm software to make their device successful, it was always commodity engineers doing commodity jobs for the highest possible markup by Qualcomm.
Look, its really pretty simple. The "smart" people saw the macro trends (increased commoditization, decreased work place quality and job satisfaction, QCT slave shop, no plan forward out of this, feels like what happened to other major companies who got fat and dumb) and just decided that it was better to get up and go before the hordes of others, to maximize their options. Everyone has a choice in this world. And if you choose not to choose and just sit it out and hope things improve, well thats your choice. Maybe they will. Who knows. I don't see the downside to leveraging your experience and trying out something new like @illuminaBoy. You could very well find something that has less corporate bloat and makes work fun again, instead of what its become @ the Q (well, specifically, QCT).
very insightful post, thanks. Google is smart to do this, this is what the Q is always going on about "changing the world", with costs going down everyone in the world can afford a premium smartphone
And so... Without a key (software) feature differentiator among chip companies that one could upcharge for, this will eventually be one ugly race to the bottom among chip companies, based on who can do it at the barebone cheapest cost alone. I think almost all U.S. chip companies are pretty much screwed, unless some thing comes along to allow them to differentiate themselves from the low cost Asian chip companies.