Is The IP Industry Healthy?

First of two parts: IP has grown to become the largest segment of EDA revenue, but is it sustainable?

popularity

The semiconductor industry has been through many changes, each designed to reduce the total cost associated with the design and manufacture of chips. Twenty years ago, most companies had their own fabs and designed all of the circuitry on each chip. Today, only a handful of companies still own a fab and outsourcing design, in the form of intellectual property (IP), has become commonplace.

IP has become the largest segment of the EDA industry, and most people expect it to continue growing for the foreseeable future. But how healthy is the industry and what headwinds is it facing?

MarketsandMarkets forecast the global semiconductor IP market will reach $7.01 billion by 2022, up from $3.09 billion in 2015, and growing at a CAGR of 10.55% between 2016 and 2022.


Fig. 1: Quarterly EDA, IP and services revenue by category, Q1 1996-Q1 2017. Source ESD MSS (Non-reporting companies include ARM, CEVA and Rambus. ARM Physical IP is estimated starting Q4 2016.)

But there is a lot of uncertainty building in the semiconductor industry, and a lot that could change in the next few years. Does this create doubt about the industry’s longevity?

A lone warning comes from Sanjive Agarwala, senior director of worldwide silicon development and fellow at Texas Instruments. He is working on chip targeting automotive and Industry 4.0, and they are characterized as having large processors and a number of sensors. “We are trying to find the best technology that is either inside or outside of TI,” says Agarwala. “This is where IP has evolved over the years. That is why we used Arm cores, because they were the best out there. Similarly with interfaces as they have evolved.”

When asked about the future, though, Agarwala sees many headwinds. “There is a feeling that the IP business is growing, but I don’t think it will continue to grow, or if it does for only another year or two. The reason is simple. The number of nodes will slow down, industry consolidation, standardization of IP—these all add margin pressure, and R&D costs are proportional to what you can afford. In my mind this means that in the next few years it will become very tough to stay in this race and you will need to consolidate the IP business itself. There was a lot of shift from internal IP development to external and that made sense. There is a lot that we don’t do inside anymore, but that shift is no longer happening.”

How healthy are IP companies? “The IP industry is growing at a double-digit pace, according to the ESD Alliance, and so it must be healthy,” says Graham Bell, vice president of marketing for Uniquify, Inc.. “However, we can see that not all IP companies are sharing the wealth equally. ARM is the market share leader with a recent growth rate of 22%. Most IP companies experience growth in the single digits. Competition is fierce, with many Tier-2 and Tier-3 providers having overlapping portfolios.”

But revenue and growth do not define the bottom line. “There are two dimensions to health,” says Hugh Durdan, vice president of strategy and products for eSilicon. “One is revenue. Is it growing? It looks quite healthy today. The second dimension is profitability. That is where the real challenge lays for the IP business.”

Industry evolution
The IP industry has been evolving ever since its creation. “We see companies that used to have captive IP groups beginning to use third-party IP,” says Tom Wong, director of marketing for design IP at Cadence. “This trend is accelerating. The make-versus-buy decision and total cost of ownership conclusion finally shifted in favor of 3rd party commercial IP.”

There are several reasons why companies buy IP. “There are some devices, such as memories, where it makes sense to obtain them from a company that will make sure they can be fabricated,” explains Dave Kelf, vice president of marketing for OneSpin Solutions. “It just isn’t worthwhile developing them because of the cost. Moving up a level, do we build a team to do the standard interfaces or is there a consultant who is an expert in the standard who can sell that IP? If it is not our core competency then it is probably not worth developing it. Finally, there are pieces such as processors, which require a lot of stuff in addition to the hardware, such as compilers. Do you want to get involved with that?”

And a company’s core competence has changed over time. “A couple of decades ago, the processor was considered to be a main point of differentiation,” points out Anush Mohandass, vice president of business development for NetSpeed Systems. “It was not considered that the processor could be outsourced. The mobile industry showed the fallacy of this. A similar thing is going to happen with different segments of the markets taking off in different directions. In automotive the focus is heterogeneous computing, and functional safety, and IoT where the focus is about a trillion devices. These all require IP of some form, and I don’t see one company making the trillion devices. A whole slew of companies will be making them and IP is the foundation on which this growth is built upon.”

And the rate of change may be accelerating. “IP, unlike EDA, is infinitely elastic as far as the types of products that are needed,” adds Warren Savage, general manager of the IP Division of Silvaco. “For EDA, once you have a full flow, you are mostly optimizing that and new flows are on a pretty slow lifecycle compared to semiconductor design which is driven by the consumer market.”

And there have always been a large number of IP startups. “It is not just EDA companies who supply IP, and indeed, we see more IP design houses who can offer specialist IP,” says Oliver King, CTO for Moortec. “Those vendors can optimize their IP solutions in ways that a generalist IP vendor cannot. Companies work with IP vendors who they trust to provide solutions which will work, and which are proven. This is the benefit over in-house designs and also open source solutions. The cost of that IP is relative against the risk of the alternative solutions.”

Still, the headwinds pointed out by TI’s Agarwala are real, even though the end result may not be bad.

Impact of Moore’s Law slowdown
If the rate at which new nodes were being developed or adopted was slowing, it could be argued that IP would become more stable. It appears that all aspects of that statement are counter to what is actually happening.

First, new nodes are being developed. “Node-to-node migration is shorter from 10nm to 7nm than it was from 28nm to 16nm,” points out Cadence’s Wong. “The pace of proliferation outstrips the ability of the IP vendors to support each and every new process flavor. There is now bulk CMOS, finFET and FD-SOI. We have 16nm, 14nm, 12nm, 10nm, 7nm all deploying in the same time frame and co-existing.”

Even the older nodes are getting refreshed. “In the past, at 40nm, you may have had a low power version and a high performance version,” says Durdan. “When 40 was done, everyone moved to the next node. But at 28nm, that continues to evolve today, and there are the original 28 HPP and HPM and HPL. And now they have an HPC and an HPC+.”

This puts pressure on the IP developer’s margins. “It is a double-edged sword because it does provide an opportunity to differentiate by attacking those nodes, but at the same time it is a burden because as these node variants come out, it becomes harder to recoup the investment and make a profit before the next variant comes along,” continues Durdan. “This has caused difficulty for the IP suppliers in the amount they have to invest to keep up with each variant of each processes node at each of the foundries.”

If customers are not migrating to the new smaller nodes, that triggers a new dynamic. “How will we increase the performance of next generation systems,” asks Rick O’Connor, executive director of the RISC-V foundation. “The consensus view is purpose-built heterogeneous cores, and many of them. If you are building an IoT device you don’t need all of the instructions that would go into a server-class machine. You don’t need an x86 device to power an IoT device that powers up every few minutes to process a few packets. RISC-V allows you to build these multi-core heterogeneous platforms or single core deeply embedded tailored specified and implemented for the exact application you are trying to run.”

In fact, the industry is looking toward increasing diversification. “The IP market has a lot of room to grow but most of it is specialized,” says Laurie Balch, a chief analyst with Gary Smith EDA. “Tackling the industry is about developing a portfolio of diverse solutions rather than one platform that many people can use. It is a customized product for a niche group of users and can change more rapidly than tools within the EDA industry.

The rapid development of new processes does create a potential problem for IP users and foundries. “There is a large class of customers who do not have their own IP and are heavily reliant on 3rd party IP ecosystem,” points out Durdan. “If the foundries cannot maintain a viable 3rd party IP ecosystem for their nodes, then it limits the customers that they can target as being the ones with their own IP.”

Evolving standards
Standards may or may not help.

“The nice thing about standards is that you have so many to choose from,” wrote Andrew Tanenbaum, a computer scientist, in 1981. “Furthermore, if you do not like any of them, you can just wait for next year’s model.”

In line with that thinking, standards are not standing still. “The standards keep evolving and the processes keep getting more complex,” says John Koeter, vice president of marketing for Synopsys’ Solutions Group. “So more complex protocols combined with more difficult implementation of that protocol in a given process node not only keep the price flat, but increasing. PCI express is complex, USB is complex, most companies just want to hand it over to an IP provider and they want it configured for their SoC requirements.”

These standards also have a significant physical IP component that requires specialized design skills.

“There are two things that drive the IP suppliers to have to continue to invest. One is the process technology and the rate at which foundries come out with new variants of the process technology. The other dimension is the new versions of the standard. There was LPDDR4, the LPDDR4x which is the low voltage version and then LPDDR5 – it just continues to evolve.”

Industry consolidation
Another potential headwind for the industry involves consolidation within the semiconductor industry. Will that change the economies of scale for larger companies?

“If you are a small to midsize company doing three chips a year, there is no way that developing IP from scratch would be cost-effective,” says Koeter. “But if you are a large company doing 10 or 20 chips a year, then that might change the make-versus-buy economics, and it may make sense to pull some IP back in house. There could be other benefits of doing that, including more control and flexibility. Also, more optimization of the IP for a particular chip versus that for the general market.”

For some companies, there could be a financial incentive. “If the business model is royalty-based, and if you had two companies paying royalties on the same IP and they consolidate, they are paying double the royalty for the same investment that it would take to create that IP,” explains One Spin’s Kelf.

But this appears not to have been what has happened so far. “There is an interesting counter-dynamic,” says Durdan. “Some of those companies have a view that each business unit needs to stand on its own and they do not believe in a centralized business organization that develops IP and provides it to multiple business groups. In the past they had a large central engineering group, but today they are a lot more decentralized structure.”

It also presumes the same needs in each group. “The products may be very different, so what one business may require may be different that another group,” continues Durdan. “So even though the company gets larger and they can scale in terms of wafer purchases, it does not mean that they can scale in terms of the IP they purchase.”

Koeter also ties this back to the reasons why IP was outsourced to begin with. “Nobody has enough engineers to keep up with the task they have. Just because it has to be done correctly does not mean that it differentiates the chip – so that fundamental equation has not changed. Secondly, with the explosion of process technologies, internal teams would not just have to design it for one process variant, but may have to do it for 3, 4 or 5 variants. At that point it becomes more effective to outsource it to someone who is investing in 40 processing variants.”

NetSpeed’s Mohandass agrees that nothing has fundamentally changed. “Yes, there is consolidation within the semiconductor industry and this will actually help the IP industry. The IP industry is a buy-versus-make decision, and people are going to look for IP that can be bought off the shelf and to leverage the know-how and keep the differentiation in-house.”

IP is here to stay
There are few danger signs on the horizon at this time. “All companies will buy IP over time because it leads to higher ROI and nobody can afford to have things get screwed up,” says Geoffrey Tate, CEO of Flex Logix. “It is better to work with a small number of high quality, proven IP suppliers who have reasonable prices than to take the risk of having an internal team screw it up. In 10 years, I would be amazed if any company would be developing IP internally if they can buy it off the shelf.”

“The reports of my death are greatly exaggerated.” – Mark Twain

“We are happy to report that the State of the Union is strong,” says Wong. “The IP industry is healthy and continues to innovate.”

Coming in part two: Changes to the IP business model, and the impact of open source and new packaging technologies.

Related Stories
Re-Using IP In Packaging
The biggest advantage in advanced packaging is still missing, but that’s expected to change.
Challenges Grow For IP Reuse
Methodologies for integration become a competitive tool as complexity and possible options skyrocket.
No More Easy IP Money (July 2016)
Revenues for semiconductor IP are expected to grow, but the easy money may have already been made. Developing IP is getting tougher on several fronts.