digits 12/5/2012 | 5:44:55 PM
re: Is There a Better Way to Fund Chip Startups?

Sorry, I just wanted to write that....

Pete Baldwin 12/5/2012 | 5:44:54 PM
re: Is There a Better Way to Fund Chip Startups?

On a more serious note... We touched on some of these issues in the 2010 Drew Lanza interview, about investment money shunning hardware in favor of services and apps.

It's a problem that I don't think will go away by itself (i.e., it's not part of a cycle), and of course there's a political element: If the U.S. doesn't produce the chip innovations, someone else will. (China.)

This article just scratches the surface by pointing to a couple of possible ways to alter chip funding. The GSA's CapLite* group will end up putting some more specific ideas into practice, so I'm hoping to keep track of what they end up doing.

What other ideas might work? Or, do we all even agree that there's a problem?

(* I'm not 100% sure that anybody besides me calls it that. Catchy, though, isn't it?)

Pete Baldwin 12/5/2012 | 5:44:54 PM
re: Is There a Better Way to Fund Chip Startups?

Oof.  Don't know if that's a good one or a groaner. :)  Either way, i'll plan to steal it for a future headline...

tmmarvel 12/5/2012 | 5:44:51 PM
re: Is There a Better Way to Fund Chip Startups?

VC funding going to application software and services rather than enabling hardware innovations? Of course so, so long as it still somehow works, though there will be more and more economic and physical constraints for apps, as they get more and more dynamic, interactive, bandwidth hungry etc while the hardware and fundamental system architecture concepts get increasingly inefficient and outdated vs the prevailing application demands. And it seems like a systemic problem; any individual VC investor would seem to be better of by investing exclusively into startups that make use of (as opposed to create) the IT infrastructures in place, eg Internet, common place OSs, processing hardware. However, collectively we're depleting the reservoir fueling the growth in app software and services.

I'd suggest that the solution should include that the new gen of startups involving also the vital hardware level innovation consider:

1) Focusing on architectural advantages, which should be strong enough to provide benefits even when implemented on standard programmable hardware (FPGAs), to eliminate custom chip startup costs. This can take the pre-revenue startup cost to less than $10M.

2) Applying X-as-a-Service business models. Eg if your chip is excellent as infrastructure for a given application field, then sell that infrastructure-as-a-service to the dynamic app service providers, by-passing the legacy-tied system hardware vendors, and addressing a faster growing market.

Yes, this type of XaaS startups with value-add derived from hardware level innovation may take more time and money to become profitable than the very lightweight web apps (hey, but then how much funding has eg Twitter taken?), but otoh, they also provide a couple of noteworthy advantages:

- More predictable financial performance, as this can be approximated based on the given XaaS vendors cost-efficiency gain over the existing supply chain (which often can be quite dramatic!) and the more forecastable market size of the existing market for given class of (infrastructure) service.

- Better predictability of the success factors for a XaaS vendor A vs XaaS vendor B (try getting that consistently right in consumer space -- there will be thousands of unsuccessful web apps in any given popular space for each big winner; hardware based ventures, having a meaningful barrier of entry, will not have this problem).

- There's a reason why it was the hardware level innovation based ventures that largely created the VC industry. With appropriate business models for the given times (eg XaaS vs plain chips), they are the enablers of the modernization and sustained growth of the entire economy, and offer the highest eventual investment return prospects.    

paolo.franzoi 12/5/2012 | 5:44:50 PM
re: Is There a Better Way to Fund Chip Startups?


I think the notion of trying to sell hardware to application companies is completely wrong headed.  The whole point of those software models is to abstract the notion of the hardware - especially in the XaaS model.  As it stands, I don't have a full server line for pretty much anything other than Intel/AMD based systems to choose from.  I await the arrival of MIPS and ARM based server systems to compete.  That should happen well before anything as obscure as some comm chip gets rolled into a system for some odd function.

I think the gauntlet that has been laid down in this world is the cheap Intel server and what you could do to have a fundamental advantage over it  Remember, if you are building your App to be "Cloud Ready" it certainly not going to expect some fancy custom hardware.

I do think the FPGA idea is more correct but limited.  The challenge there is convincing me that I am better off buying your FPGA (that will be hardened into an ASSP) instead of developing my own that is positioned perfectly for my application.  The biggest single challenge is that this model faces is that the FPGA IP market has never really gone anywhere.  Nobody wants to certify anything and it limits the value of what you buy.  You are going to have to support it if it breaks in your application. 

I think I presented this problem to my friends at Infineon (now Lantiq) about 8 years ago.  New ICs are becoming scarcer.  Which means that systems companies are moving to building the same stuff out of the same stuff  This leads to declining margins and the notion of big systems software investments to build Telecom Equipment has to go away.  The real question is:  What happens when all Comm hardware has the same margin as Linksys?



rhr 12/5/2012 | 5:44:49 PM
re: Is There a Better Way to Fund Chip Startups? Another excellent chip article - I keep telling you, Craig, you must dust off more often your EETs' heritage.-á

As for the dip without chips (Ray's football team won again so he is to be forgiven), it is an issue but perhaps a natural consequence of Moore's Law. Chips and their gate resources are now so huge that it is asking a lot for multiple start-ups to design from scratch the latest SoC design.-á

But as mentioned start-ups can make use of FPGAs and the multiple cores, IP and high-speed interfaces that now fit on them. -á

More system vendors can become vertically integrated (the biggest such as Cisco, Juniper, ALU and Huawei already make their own key ASICs).

Intel, Broadcom and Qualcomm can do more with their own VC funds and they do understand chips.

Industry R&D organisations such as IMEC and the EC Framework 7 can play a role here.-á

And maybe the FPGA vendors themselves can do more: funding startups and benefiting from the IP.-á

The same applies to system vendors and/ or optical module vendors that can collaborate to fund start-up silicon IP.

If holes really do appear, there will be a dollar opportunity that even VCs will notice. Ultimately,-áif the US and Europe don't respond, the likes of China and India will.
Sign In