To answer this question, I will share the results about Interface IP, more precisely the Top 5. The Top 5 protocols, USB, PCIe, Ethernet, DDRn and MIPI, are part of the interface IP market and each of them has been characterized by very strong growth rate. If you compute actual numbers for 2010 to 2014, it results to a Cumulated Annual Growth Rate (CAGR) of 14%.
What does it means for IP vendors? Simply that the revenue generated by license sales has doubled, passing from $180M in 2010 to $360M in 2014. Please note that these standard based IP business is based on a model calling for upfront license fees and excluding royalties, with the notable exception of Rambus, not included in the research.
Should we expect this IP business to stay flat or even decline as it was claimed during DAC 2015 introduction speech (*)? If you look at numbers reported by EDAC for Q2’15, you notice that the IP market four-quarter moving average increased by 22.1%, when the same metric for the CAE market was… minus 0.3%.
Such a percentage already indicates strong growth trend, on line with the 2014-2010 CAGR of 14% calculated on actual license revenues generated by these Top five interfaces protocols (USB, DDRn, Ethernet, PCIe, MIPI). You can see the detailed graphic below, extracted from the Interface IP Survey 2010-2020 issued this month by IPnest.
But the question we want to address is: What is the interface IP market behavior for 2015 to 2020?
To answer, we will research in four directions: IP commoditization, design start including interface IP evolution, IP Externalization and total ASIC/ASSP design start decline.
Interface IP is becoming commodity
Some people think that this type of IP is becoming commodities. In fact I have made a presentation during IP-SoC in 2009 addressing this question, specifically about PCI Express, USB, HDMI or SATA. The answer was negative, let me explain why. These IP are designed against a protocol standard. If you consider PCI Express Gen-1 in 2005, it was specified at 2.5 Gbps per lane. The PCIe PHY was designed on 90 nm or even 130 nm, the digital logic (PIPE and Controller) could run at 125 or 250 MHz (depending on the number of lanes). When Gen-2 was defined, the PHY speed had doubled and the target technology was 65 or even 40 nm. Instead of decreasing like you would expect for a commodity product, the price was increasing. This was also true for the controller: more functionality added in the specification, like Single Root I/O Virtualization, logic running at 2X the frequency. Today, PCIe gen-4 is to be released and we can see some vendors advertising 16 Gbps PHY designed on 16FF. The PHY license price is expected to be in the 8X to 12X range in 2015 compared with gen-1 PHY in 2005. Did you say commodity?
Is number of design start including Interface IP declining?
Looking at this market since 2005, first as Marketing Director for an IP vendor, then as an analyst I can tell for sure that this is not the case. The adoption rate for a (successful) standard is usually ramping very fast like for SATA between 2003 and 2007 or for HDMI (in 2012 almost 100% TV sold in the US where equipped). PCIe or USB standards are different: on top of the fast adoption for certain market segment, we see pervasion occurring in other segments. USB protocol is now ubiquitous; USB LS, FS, HS or SuperSpeed (USB 3.x) can be found in every market segment. PCIe is integrated into PC, Servers, Networking, Computing, Test equipment, etc. and the number of design starts is in progress, year after year. The addressable market size is still growing. Will the TAM stabilize, sooner or later? Some protocols will disappear like SATA, some will emerge like Mobile Express, and we expect the interface IP $TAM to continue to grow up to 2019-2020, but with a lower growth rate than during 2010-2014.
Externalization rate behavior…
If you take a certain SoC design targeting a specific application, integration capabilities has grown up to a point that no design team could fill the chip with internally designed IP. The R&D team has to concentrate on designing the few differentiating function and integrate several dozens of IP, let’s call it commercial IP (externally sourced). If you took a snapshot of the semi industry in 2010, only a few companies were targeting 28 nm with a SoC integrating 80 IP or more. In 2015, these companies are designing SoC in 16FF or even 10FF, and they integrate even more IP than in 2010. But the growth reservoir for IP vendors are the semi companies designing in 130 or 90nm in 2010 and eventually moving to 65, 40 or even 28nm in 2015. They used to buy a few IP only (CPU, foundation IP and so on), but they will have to integrate commercial IP, mostly for Time-To-Market reasons (and because a design team can’t grow as fast as technology node decrease). This externalization rate is not easy to define, it varies depending on the chip maker, the protocol, the application… But for sure it grows year after year.
ASIC or ASSP design starts is declining!
This assertion is true, and many ASIC designs in the 2000’s have become FPGA designs in 2015. But is it really true for SoC designs? We know that a SoC is integrating many IP, including SDRAM Memory Controller and various interfaces like USB, PCIe and so on. IPnest assumption is that the number of SoC design starts is growing. Targeting certain application with FPGA instead of ASIC SoC is not a viable solution, for performance, power and probably the most important, for cost consideration.
Let’s call this the SoC paradox: SoC design starts count is growing, even if global ASIC/ASSP design starts is declining.
Yes, 2015-2020 IP market will grow!
This is the conclusion derived from the above mentioned consideration, comforted by precise evaluation made year after year since 2008 of the interface IP market, with a fine granularity (by protocol). Building a forecast is far to be easy. Building it year after year give you a “feedback loop” allowing to fine tune certain parameters, like externalization rate growth factor, for example. Here is the result for the Top 5 protocols:
Last point: you may question IPnest methodology and, more important, the validity of such forecast. Some facts:
- In Q1 2009, IPnest has built the first 5 years forecast; the expected revenue for interface IP was $415M in 2013.
- In 2014, IPnest has computed the actual results for 2013 revenue generated by interface IP, $425M. Less than 5% accuracy for 5 years forecast…
You can see below the forecast made in 2009 about the Interface IP market evolution (forecast) from 2009 to 2013:
Want to know more about the Interface IP Survey? Access the Table of Content here.