The End Of The Learning Curve Is Nigh, says Scott McGregor

28nm is the first process shrink which doesn’t deliver a cheaper chip, says Scott McGregor, CEO of Broadcom.

According to McGregor, not only is 28nm more expensive than 40nm now, as might be expected at the start of a node, but that, on Broadcom’s current projections, 28nm will never be cheaper than 40nm across the whole lifetime of the node.

“What we’re seeing in the industry is the cost of next-generation nodes rising exponentially,” says McGregor, “what this means is, unless you need the advanced process because of performance reasons or die-size reasons, you’re not going to get a cost benefit from converting to the new node.”

McGregor went on to say that the same would be true at 20nm.

The consequences will be dramatic: companies will stick longer with a node, foundries won’t be so aggressive in their process development and the old learning curve of chips getting cheaper by 30% a year will decelerate.

Apple, reportedly, has stuck with 40nm and 65nm technology rather than risk its products on the availability of 28nm parts.

Broadcom’s strategy on 28nm was to wait on the HPM process which, says McGregor, “tends to have a better figure of merit than the HP or the Poly/SiON processes that some of our competitors are using. So when we do come out with 28nm products, they should have better specs, better power performance and whatnot.”

So, McGregor points out, Broadcom has not suffered initial ramp issues of its competitors and its business has not been limited by access to wafers.

However TSMC is currently pulling out all the stops to get 20nm into production as soon as possible to give its mobile customers a competitive process to Intel’s 22nm finfet process. It was reported last week that TSMC is putting another $700m into bringing forward 20nm deployment.



  1. Keith I agree that they won’t, however Intel has always been weak with their graphics and did acquire a small stake in them at one time( not sure they still do).
    Take away that talent pool from the market… incorporate it into and on to the chip and leave AMD and NVDA to fight for the scraps. A chess match, if they can incorporate this technology and gain from it while forcing competition to go elsewhere then so be it.
    Further more what would this do to the RISC market and ARMh?

  2. “Already Imagination PowerVR’s graphics are incorporated into Intel’s Ivy Bridge with much better performance than Intel has provided in the past and will only get better over time. (Intel should buy them) ”
    Intel won’t buy Imagination for the same reason they won’t buy ARM. If they did they would destroy its business which is as an IP provider to many other semi companies.

  3. Nothing is being said about potential cost savings with 450mm wafers. Intel said today that the fabless would not find value at these same prices for smaller chips and would have to raise prices.
    Good for someone like Intel! One of Intel’s main points was that even though prices for their chips may creep that the price per transistor was dropping for them. For each process shrink more transistors are added along with additional features such as on die, or integrated graphics, communications, even memory will provide more revenue for Intel and overall fewer needs for other companies. On die memory will be a deal breaker for the fabless to compete with.
    Already Imagination PowerVR’s graphics are incorporated into Intel’s Ivy Bridge with much better performance than Intel has provided in the past and will only get better over time. (Intel should buy them)
    So the authors’ claim maybe true for the “fabless”… doesn’t ring true for Intel.
    Also not mentioned are prices of Broadcom chips; for a smart phone… Baseband $10 – 13, Combo-chip $3-4,Touch controller $2 – 3 and GPS $1 – 2.
    Margins are tight and prices may have to go up for these Fabless companies. If any of these items can be integrated on the Intel processor that too will put more pressure on a company like Broadcom who’s manufacturing costs will go up and potential loss of revenue for any competing advances in process integration that will chip away at their bottom line.
    “Resistance is futile” these companies will have to merge in order to compete and protect their patents, and that is where their value is and Intel may just put a bull’s-eye on their back.
    With that said Broadcom is a great company, however they need to grow and will need to be creative in finding a vehicle for that growth. Following Intel’s lead with process integration for like Fabless companies may be their path. How else will they compete with rising cost(?), they must keep up to stay relavent.

  4. Spot on, Robert, IMHO

  5. Yea David the same thought occurred to me.
    BRCM was very aggressive with product process migration a couple of year back, this definitely gave them a GPM advantage over competitors that were stuck on 90/65nm nodes. Interestingly, even with fabless companies, over investing on any given node often causes under investment on the subsequent node (while you try to extract maximum value from analog and RF blocks). Unfortunately complex Analog in 28nm is much harder than equivalent functions were in 55 and 40. So maybe Scott is really saying, he’d rather milk more out of 40 than pay the engineering costs to aggressively attack 28nm.

  6. Yes indeed Robert I think the statement is not entirely unconnected with the fact that Broadcom is late to 28nm

  7. The problem with statements like this is that we have all heard them before.
    IMHO 28nm is expensive because there are so few sources for a reliable 28nm process, nothing more than a lack of viable competition is keeping prices high. If history is any guide, than this situation will change quickly when knowledge of how to overcome 28nm problems filters through the process engineering community.
    Historically it’s interesting that this view-point about the future, often proceeds that companies exit from the SOC race. (TI exited at 90nm with a similar statement, Moto also made similar noises for many years starting about 250nm (from memory)) I wonder if we will all look back at this statement and mark it as the turning point for BRCM growth.
    Having said this I’ll have to admit that 28nm has no “bang for your buck” improvement over 40nm, and in most cases, absolutely no bang improvement regardless of the price.

  8. Heard the same thing on 28 vs. 40 cost from another major this week. An industry watershed.

  9. I believe you’ll find the 32nm port of the Apple A5 processor is now shipping, but I agree there is a lot of merit in what Scott says.

Leave a Reply

Your email address will not be published. Required fields are marked *