The Big Picture: Earlier this year, we were reviewing Analyst Day slides from leading semiconductor companies and an obvious theme came up. All large companies are heading in a similar direction, which poses some potential challenges for their long-term positions. More and more customers are looking for Special purpose chips, which is a mechanism for dealing with hysteresis in Moore’s Law. And all the big players are looking forward to supporting these customers.
slice design It is a complex process. First comes putting the slide in, making sure the design works, then preparing all the files required for the foundry, and when they come back from the factory there is a lot of work involved in debugging the slide and working. It’s becoming increasingly easy to do that first part – designing a chip, but the rest is still a rather tedious process, and the big chip companies think they can help others get through it all.
Every chip design company has a large workforce that handles “operations” – managing the entire process of chip production from initial setup with the foundry all the way to production planning. Companies like Google and Facebook want to design their own chips but don’t necessarily want to do all this “backend” themselves.
Guest author Jonathan Goldberg He is the founder of D2D Advisory, a cross-functional advisory firm. Jonathan develops growth strategies and alliances for companies in the mobile, networking, gaming and software industries.
If you only produce two or three chips, it usually doesn’t make economic sense to have those teams in-house. In contrast, chip companies must have these teams, adding third-party design to the workflow does not add much burden.
And there is more. Part of designing a chip is the “fun” part of setting software requirements for logic gates, but there are many more “exciting” parts to every chip design – I/O blocks, memory, SERDES, interconnects, and more. The big chip companies already have access to this IP, and many times it’s easier for the super-caliber to get that IP from the chip companies than to mess directly with licensees.
For a while, there was a lot of concern among the major chip companies that their biggest customers would and end up with their own silicon design. Rather than fight this trend, the incumbents decided to embrace it by helping these clients.
At least that’s the strategic logic. Business logic is a little more muddled.
Simply put, doing operations services to customers isn’t a great job, or at least deviates a lot from how chip companies like to operate. The revenue of this service is not repeated, each exterior design can be put up for bidding. The work involved is fairly labor intensive, so the gross margins aren’t all that exciting, and are usually lower than the company’s gross margins. On the other hand, a lot of the cost associated with this business can be amortized on a workforce that chip companies have to keep anyway, so contribution margins might make it worth the effort.
However, we believe that the incumbents have come to a reasonable conclusion that providing these operations services can bring strategic benefits in the future. Bring in the client for engineering support, and then leverage that relationship to work on their other segments in the mix.
we wrote About Qualcomm Apparently it does just that. And they are not alone. AMD dedicated several slides on this topic in its CEO’s keynote earlier this year as well.
Marvel has also devoted significant air time to the topic in its storylines.
Marvell also provided this helpful graphic that shows the whole strategy around creating semi-custom heterogeneous compute solutions.
But the strongest outreach efforts come from from BroadcomWho is the spiritual father of this whole segment. They held an investor event focused entirely on custom silicon.
That’s already more than $2 billion for Broadcom, which largely got into the business by helping Apple build A-series chips for the iPhone a decade or so ago.
When all is said and done, chip companies aren’t really going to like this business. They devalue their design teams which are their ultimate source of differentiation. Nor is it clear how viable this business is in the long term.
Many companies have tried to do this in the past, but none have lasted. LSI and eASIC are probably the best-known, and they ended up in decidedly unexciting sales to larger companies with broader product portfolios. The case for eASIC in particular shines because this company is really struggling for relevance, not to mention profitability.
However, fighting the trend seems futile at this point. So instead, all of the big companies are making the smart choice to diversify their offering, and may/maybe/we hope they use this as a way to drive customers down their catalog chip sales funnel. This is a strategy carried forward by the necessity of a rapidly changing market and therefore a viable hedge. If current trends continue and more non-chip companies build more of their own, the semi-final incumbents will be well positioned to support them.