On Saturday I attended the Chinese American Semiconductor Professional Association (CASPA) annual conference and dinner. You might have noticed that I am not Chinese 可是我会说中文，虽然我不是流利。I seem to have been a sort of honorary member since I moderated a panel for them a few years ago. The dinner keynote was by Hock Tan, the CEO of Avago. This was originally the semiconductor division of HP. When HP was split in 1999 and Agilent was spun out, it was part of Agilent. Then KKR and Siliverlake did a leveraged buyout of the division and created Avago. It has done various acquisitions and divestitures, the most notable being LSI Logic and, still pending, Broadcom. In fact the company plans to change its name to Broadcom following the acquisition. We were lucky to have Hock Tan speak. He does very little speaking and rarely appears in public. As the Wall Street Journal said: Mr. Tan has shied away from trappings like a fancy office or public appearances. He does have a fondness for expensive cars, however, recently purchasing a Tesla electric car. First thing, I think the WSJ needs to get out more. OK, a Tesla is not cheap but, to me, an "expensive car" in the context of a guy worth hundreds of millions is a Lamborghini or a McLaren. Several regular engineers I know own Teslas, too. Hock was born in Malaysia, did a degree in mechanical engineering at MIT and then an MBA at Harvard. He worked at both General Motors and Pepsi (and several other places) before becoming COO of ICS, which went private and then public before being sold to IDT for $1.7B (Cadence trivia fact of the day: building 11 used to be an IDT fab). When Avago was spun out of Agilent, he became CEO. As he admitted in his presentation, he is not a semiconductor person but he knows how to make money. He showed the chart of the Avago stock price since going public. He is also proud that over 400 millionaires have been created at Avago since the IPO. He structured his advice to the audience as a David Letterman style Top Ten List. 10. The semiconductor industry is maturing. Look at the graph below of industry revenue. You only need to glance at it to see that it is flattening. 9. Moore's Law is not dead, just middle-aged 8. SoCs are more complex and harder In a roomful of Chinese semiconductor professionals, he didn't feel any additional explanation was required. As chips move to smaller dimensions, the costs increase. They have gone up by a factor of four from 65nm to 16nm. 7. ...but users are not willing to pay Hock worked for Pepsi and a can of Pepsi costs pretty much what it did 20 years ago. Chips, not so much. That actually reminded me of a presentation from someone in HP Semiconductor (the future Avago, but this is years ago back when it was still part of HP) who showed the back page of the San Jose Mercury for several different weeks showing the price of memory chips falling dramatically (I think this was the era of the 64K DRAM) while, down in the bottom right corner, the price of a six pack of Coke remained exactly the same price. 6. Internet of Things (IoT) is not ready This will eventually be a big market. But it is now, and perhaps always will be, very fragmented. The money is probably not in the chips anyway. 5. Analog gets more respect 4. There is no #4 This is not a Monty Python Australian philosopher reference, that would be rule six, but an acknowledgement that four is a number that is considered bad luck in Chinese, since the word for four, 四, and the word for death, 死, are both "si" (although the tones are different). On the other hand, eight, 八, is considered good luck, which is why the Beijing Olympics opened at 8pm on 8/8 in 2008. 3. Software is the new hardware Future growth will mostly be in software. For example, the money in IoT is mostly in the software stack. 2. Demand is exploding in the emerging economies See the graph below showing semiconductor consumption in emerging economies compared to the developed economies (I don't think this takes account of the fact that, for example, China imports a lot of silicon that is then re-exported in finished devices such as iPhones). But the trend is clear, and China consumes a lot of silicon internally, too. 1. Despite all this, consolidation is slow There used to be dozens of PC companies in the US, now there are basically HP, Dell, Lenovo and Apple (OK, not technically a PC, I know). Semiconductor still has a very long tail. There is more consolidation to come. And we were hearing this from a man the WSJ has dubbed "the consolidator", and "a finance whiz known for overseeing lean operations". Hock Tan's parting advice, given that CASPA had just presented its scholarships to high school students, was what areas to consider for a career. No top ten list, just three items: Software Analog Emerging countries For those of us on the EDA side of the fence, consolidation is a headwind. When a company like Avago buys a company like Broadcom, the EDA budget does not go up, even if no duplicate programs are canceled. It is one of the paradoxes of semiconductor and EDA that no matter how few chips are designed by how few companies, roughly the same EDA R&D budget is required to produce tools to create those designs. Or as Lip-Bu Tan, Cadence's CEO, put it on this week's earnings call in reply to a question: I think clearly the ongoing consolidation in our customer base will continue. That potentially will have challenge in the longer term but we do not expect any material impact in the short term.![]()
![]()
↧