So, I'm snake bit.
About an hour after my "Saying Goodbye To Paul Otellini" article went live, IDC comes out with its "count" of PC units being down 14% for the first quarter. That would be a decline from 88 million units to 76 million units. Gartner says the number was 79 million units. Gartner threw a new wrinkle into the numbers in its 2013 forecast. It has a new category called "ultramobiles." Ultramobiles, presumably, are Ultrabooks and Macbook Airs. Depending upon the reporter (as you can see in the headline in the above article) this ultramobile category is left out of the PC numbers. If there were only five million Macbook Airs and Ultrabooks shipped during the quarter, the Gartner numbers could be 84 million units, down 4 million units or 4.5%.
There is no question that the number of PCs shipped was down. If I were the entire market, it would have been down 100% because I am in the market for a replacement notebook for my Dell (DELL) boat anchor and I can't find a decent Ultrabook convertible from any manufacturer. Hewlett-Packard (HPQ) and Dell have gone brain dead and really let the market and Intel (INTC) down.
Both IDC and Gartner have an awful record on forecasting and counting sales of PCs. The errors are usually on the low side because they have no good way to count all the emerging market "cottage" assemblers of PCs.
I received the following message from Paul McWilliams who writes NextInning and is another old semiconductor guy who could teach me a thing or two:
" Q) Does the severity of the fall off in PC sales surprise you? Does it change the time table for when Intel will be a growth story once again?
A) The data doesn't surprise me at all. PC sales have been soft for quite a while now, and even as we were entering Q1 expectations were the quarter would be softer than usual. However, that doesn't stop financial writers from leveraging the data for sensational headlines.
So IDC says PC shipments in the first quarter were down a record 13.9% sequentially to 76.3M units. Gartner disagrees and says PC unit volume was 79.2M, which represents an 11.2% decline. According to Gartner, this is the first time unit volume has been below 80M units since Q2 2009. Both IDC and Gartner say Hewlett-Packard was the big loser, and I believe that. Both also say Lenovo gained the most market share, and I believe that too. What I don't believe is the notion that IDC or Gartner can accurately gauge the number of PCs sold in a quarter.
I can remember analysts on INTC conference calls not that long ago questioning Paul Otellini as to how his numbers could be as good as they were in light of the research reports published just ahead of INTC's quarterly releases. I forget exactly what Otellini said in response, but it was something along the lines, "I don't know, you'll have to ask them about their numbers. We're the ones selling the processors."
Remember what it was like in the U.S. in the early 1990s when Michael Dell was building PCs in his dorm room? There were hundreds of small PC builders in virtually every major city in the U.S. - it was a highly fragmented industry. That's what it's like in emerging markets today. In China there are far more consumer electronics companies most U.S. citizens have never heard of than there are brand names you might recognize."
According to these two market research companies, PCs were down somewhere between 4 and 12 million units. If we use an average of eight million units and an Intel ASP (Average Selling Price) of $100, the total impact is $800 million in revenue to Intel. $800 million is within the guidance given by Intel last quarter of "$12.7 billion plus or minus $500 million." I'm not going to get spun up about an Intel miss next Wednesday. Intel has always proved the forecasting companies wrong and I think it will prove them wrong again.
The important thing is: Where is the future Intel growth going to come from?
I won't spend much time on mobile since I think that by the time Intel ramps 14nm in 2015, it will be well on the way to domination of that market. As big and as fast growing as that market is, it is only (only?) $16 billion. That would be only if Intel sold every single mobile chip in the market, including Apple (AAPL) and Samsung.
More from Paul McWilliams:
"Wall Street is also missing the point - form factor doesn't matter - people want processing power and the ARMH processors used today in mobile devices would be overloaded like a Yugo trying to pull a trailer up a steep hill with the AC running if they tried to run Norton or McAfee virus software or the plethora of crap-ware that commonly comes with a new PC. My recent PC purchase came with zero crap-ware and I still can't believe how blazing fast it is. After spending some time with Windows 8 (with classicshell installed), I really like the OS - best I've ever seen from MSFT."
The consumer confusion about notebooks, big and small tablets and smartphones is probably causing the slow PC sales more than anything else. With a good convertible PC like North Cape and a little bigger smartphone, stand-alone tablets will join the CB radios of the 1970s.
What else is there for Intel growth?
Intel makes SSDs (Solid State Drives). SSDs give near instant boot time and improve performance substantially while reducing weight and power a great deal. Intel has also begun to embed DRAM in the CPU package with Crystalwell in some versions of the upcoming Haswell CPU. Both of these technologies dramatically improve performance while reducing system power consumption.
Without getting into a technical presentation, I will just make the statement that simple close proximity of The CPU, SSD and DRAM has the potential to dramatically increase performance and reduce power.
An example of this is the Intel/Micron (MU) developed Hybrid Memory Cube. Just stacking the DRAM chips gives a 15 times increase in bandwidth (speed) and a 70% per bit reduction in power.
OK, let's knit some of this together. Since the Haswell CPU is such lower power, it will not heat up memory mounted physically close to it (heat on memory is a bad thing). Imagine an Intel "Compute Module" that would be about a 2.5'X2.5' square that mounts the CPU, with the stick-of-gum sized 128GB SSD along one side of the CPU and 4GB of DRAM mounted along the other side. The total value of this Compute Module would be: $100 - CPU + $128 - SSD + $30 - DRAM + $20 - power control chip embedded with the Haswell = $278 of Intel content per CONVERTIBLE PC. The extra $178 multiplied by 300 million PCs is another $53 billion in high margin revenue to Intel. That level of business, which seems a no-brainer for Intel, makes the mobile business look like just an attractive desert.
If Intel (or the Intel/Micron joint venture) doesn't do the SSD business, and since no non-NAND manufacturer can be long-term successful in SSDs, the list of companies to take the $38 billion in SSD business would be Samsung, Hynix or Toshiba/Sandisk (SNDK), none of which do a thing for Intel. The DRAM for the compute module can be obtained through the Intel/Micron joint venture.
Of course all of this is speculation since I am not privy to the Intel "Big Picture," but all the pieces are in place, working and cost effective today.
The plays are Intel, of course, Micron, and an overlooked LSI Corporation (LSI) which supplies the secret sauce controller chip for the Intel SSDs.