I agree that the purchase is intriguing, but there has been little discussion, here or on ARM boards throughout cyberspace, of its implications and possible motivations. Become a Complete Fool
Absolutely. There have been some ill-informed knee-jerk reactions in the press about ARM entering the fabless market, but surprisingly little of substance so far. There has been a lot of noise about "ARM entering the microprocessor market" and "competing with licensees", but little careful thought about how the Triscend acquisition might play a role in ARM's strategic vision.
The PLD article you linked is an excellent starting-point, I think, although we have to remember it is over four years old. For clarity, I think the sort of chips Triscend designs and sells are what the article terms hybrids, because they combine a dedicated processor core with an area of programmable logic. Cor's link to the discussion article on the relative advantages of ASICs and FPGAs is also very useful background.
Unfortunately, we know relatively little detailed information about Triscend, because they are privately owned.
Here's my take on the acquisition, based on what I have read since the announcement and a healthy dose of speculation.
I'm going to put aside for the time being the question of whether this means ARM will compete directly with their own licensees, but I do intend to return to it (someone remind me if I forget).
Let's look instead at what the problem is, for ARM, to which this acquisition is the answer.
ARM has cited microcontrollers as one of their eight sectors for several years now (previously they used to refer to "industrial" applications, I think). They have also explained repeatedly the scope of the microcontroller market, and the reasons they - and others - anticipate a trend toward replacement of 4-bit and 8-bit microcontrollers by 32-bit models. Yet in ARM's presentations, the microcontroller slice of the shipment pie remains stubbornly narrow, at around 1%. That still implies growth, of course, but at that level the rounding errors must be significant. ARM have a core specifically designed for the microcontroller sector in development, and several licensees have already created platforms for the sector. ARM have even co-branded a development kit with Oki to help increase adoption. The bottom line is that this sector may be more difficult to penetrate than most of the others - there is considerable inertia behind existing lower-level platforms, and the advantages of 32-bit platforms have yet to be demonstrated convincingly to decision-makers in the sector.
If that is the problem, how does Triscend help? I think I see at least two ways.
First, PLDs are often used as part of the chip development process. Prototypes created in PLDs (or indeed FPGAs) can be used to test and demonstrate the flow of logic before the high-level design is implemented in ASIC form. This is becoming increasingly important as mask costs escalate with new process advances, meaning that designs have to be correct, first time - otherwise the non-recurring engineering costs distort the economics of the project. I think ARM alluded to this last week when they referred to Triscend effectively forming part of the development tools side of the business. Certainly there is some valid comparison with development boards, for example. Of course, there are PLDs and FPGAs available on the open market, but one-stop shopping must be helpful, especially when backed by ARM's support structures.
The other area of significance also derives from those escalating NRE costs. With set-up costs already high and rising further, the threshold of ASIC profitability is pushed further out. In a competitive market, pricing is not really flexible, so shipment numbers have to be higher before a return is created on the development investment. That is problematic, since shipment numbers in this sector are typically rather lower than others (at present), and even for those product lines that achieve respectable shipment levels, those shipments are certainly not assured at the design phase.
So, the way I see this acquisition working is that ARM will be able to sell Triscend PLDs as part of a strategy of reducing risk for end-user clients in the microcontroller sector. Now, these may well be a completely different set of clients to those ARM normally deal with, but ARM already sell development kits mainly to their licensees' clients, and there may be sales help available from an unusual source (cryptic, eh?). ARM's pitch is that the Triscend platform enables rapid and cost-effective prototyping, accelerating development and reducing risk for the development of ASICs. The corollary is that the client can choose to continue with the PLD implementation for as long as they choose, in case production rates do not justify the NRE costs of an ASIC. In either case, ARM will be able to wean the client onto an ASIC implementation, either their own or a commercial development with another licensee such as Sharp, Oki, Philips, or any other of the microcontroller sector licensees. These are the "unusual sources" of sales assistance: it is in the interests of these clients to see the Triscend architecture proliferate, because of the potential leads into later sales these can provide.
I think this view is largely consistent with most of ARM's comments and remarks on the acquisition. The press release referred to ARM being "able to proliferate the ARM architecture by selling configurable MCUs to OEMs for emerging applications" (my emphasis), and mentioned a new program to "offer a seamless migration path to take matured MCU designs into ARM partners' standard products and ASIC programs for high-volume applications".
In a Silicon Strategies interview with Peter Clarke, Warren East provided some further perspective:
He [East] added that ARM's intention is to start emerging applications off on the Triscend platform and then migrate them to licensees' products either in the microcontroller space or the ASIC space.
"An SoC customer might have a quarter of a million unit requirement per year. A microcontroller customer might only want 20,000 units," said East.
So, I anticipate that the new program will involve the microcontroller licensees actively selling Triscend cores, in exchange for pass-through of the sales opportunity if/when the implementation has matured beyond the scope of PLD shipments.
This is a significant change for Triscend. Until now, Triscend's business model has depended solely on maximizing the number of PLD chips they sell. Naturally, they would be reluctant to see any client switch from using their PLDs to using an ASIC supplied by another vendor, especially if the implementation of the latter is based on that of the former. Now, however, the greater good of ARM may prevail, and encourage Triscend to forfeit chip sales in reasonable assurance that ARM will still benefit from royalties from the new vendor.
The effect on the client is that the migration path to 32-bit ARM implementation becomes both simpler and smoother.
The effect on ARM is increased penetration of the MCU sector, using both chips (sold at relatively low margin) and royalties (100% gross margin).
The effect on ARM's microcontroller licensees is increased sales, through the use of Triscend PLDs as an intermediate step to reduce risk to the client.
I think this may be an ingenious piece of innovation by ARM. The smartest thing about the company has always been they have created and refined the IP business model, and now it seems they have added a new dimension to their doctrine of "co-opertition".
ARM have said that they intend to continue selling and supporting all Triscend products, although I think we may well see greater emphasis given now to the ARM7 hybrids. That does not prevent some further adjustment to the business model. Apparently Triscend has enjoyed some competitive advantage in rapid development, and it has been suggested that this is due to the secret sauce in their configurable system interconnect (CSI) technology. There may be some potential for this to be offered on an IP basis, although I think I would prefer to see the Triscend product range expanded successfully first.
CSI IP, assuming it is adequately protected and viable as IP, could add to the return on ARM's investment, but I think there is another factor that effectively boosts the potential return on the deal rather more (cryptic again, eh?).
We on this board are, I am sure, all familiar with the argument that ARM is effectively immune from takeover by any of the large licensees because the very act of takeover, by, say, Intel, would almost certainly mean the end of the architecture as an open de facto standard, causing licensees to seek alternatives, and so to diminish ARM's future returns on investment, and the intrinsic value. That is, as soon as anyone buys it, the value drops. While we are all ruefully familiar with this phenomenon in terms of new cars and share purchases (d'oh!), it does have some strategic value to ARM.
On the other hand, I think there is an opposite version operating in this case with Triscend. They already held at least one ARM7 license, I believe, but now they have the opportunity to pick whatever cores they want from ARM's current and future portfolio (I wouldn't expect internal cross-sales). This means that the company effectively has expanded the range of cores they can offer on PLD hybrids, simply by being taken over. If Triscend were still available on the market, this would almost certainly increase the intrinsic value of the company. This is not what is conventionally meant by "operational synergies", but there is no doubt in my mind now that Triscend is worth more as part of ARM than they were as an independent company.
I think the acquisition makes excellent strategic and financial sense. More than this, though, I am delighted to see ARM taking the business model in new and creative directions. With all due respect to ARM's technical people, it is the engineering of the business that has impressed me most consistently over the years, now more than ever.
Competition with licensees? I should coco. I expect they will be surprisingly eager to embrace the new program.
You can wake up now, it's over.
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.
I agree that the purchase is intriguing, but there has been little discussion, here or on ARM boards throughout cyberspace, of its implications and possible motivations.
Become a Complete Fool