Intel (INTC) reported earnings Tuesday after the close. The results were within Intel's own guidance given a quarter ago.
First-quarter revenue was down $100 million from the midpoint of guidance, gross margin was at the low end of guidance.
Second-quarter guidance is marginally better than first quarter results, which was as expected. The first quarter could have been better if two major customers, Dell (DELL) and Hewlett-Packard (HPQ) weren't completely distracted from the PC business.
Intel has maintained the last half of 2013 will begin a recovery with a 2% increase in revenue for the full year and gross margin recovery to 60% for the full year.
Given that, the second half should be ~$29 billion and 63% gross margins. A couple of $14.5 billion quarters would be new records, so the Intel world isn't falling apart as some pundits are wont to declare.
PC client units were down 6% vs. third party estimates of down 14%. Someday the third parties will surrender and simply wait for the quarterly reports in order to get the numbers right. Why announce gross errors only days before the earnings announcement? I agree with the author of this article that the Osborne Effect has been playing a little mischief in the PC market.
Takeaways from the Conference Call:
Mr. Otellini told us that the company is better positioned to compete across all segment of computing. He said the technology lead between Intel and the next best in the industry is actually increasing.
Mr. Otellini is not much of a chest thumper, and I believe he is incapable of telling an untruth. So the company is positioned well. I was disappointed that he didn't throw us a couple of specific bones on the future of the company.
I have to say that the tone of the conference call was significantly muted relative to, say, the 4th quarter of 2010 when Mr. Otellini proudly informed listeners that Intel would move, "to a four fab model from a three fab model." This change in connection with the 32nm to 22nm conversion would have provided nearly twice the capacity to build transistors. On the Tuesday conference call, there was no Intel originated discussion of manufacturing capacity.
Some of the analysts actually tried to get at the capacity issues and were rebuffed by Stacy Smith, whose responses to analyst's capacity question were absolutely indecipherable to me. James Covelo of Goldman Sachs, no friend of Intel, asked some very specific questions related to apparent excessive capacity. Stacy Smith answered in Martian; a collection of words that made noise, but not sense. Mr. Otellini, the truthful, stayed away from that exchange.
On the topic of the foundry initiative, Mr. Otellini, when asked about making ARM based products, said, "The second is that we would not enable a chip competitor." Most would interpret that as a hard "no," but the two words "chip competitor," while not admitting to doing ASIC chips for Apple (AAPL), did not exclude that possibility. What was excluded was foundry work for Qualcomm (QCOM), Nvidia (NVDA), and, of course, Samsung (SSNLF.PK). A very careful and clever answer from Mr. Otellini.
In an answer to the very last question Mr. Otellini mentioned non-core (I guess Atom based) notebooks at a price point of as low as $200. I found that to be a "stiff-arm" to the doubters about Intel's willingness to compete from the top to the bottom of the computing market.
Overall, I was left with the impression of less transparency than previous conference calls, as though there are major initiatives that are not ready to be revealed.
Capacity -- My Favorite Subject:
The baseline for CapEx from 2006-2009 inclusive, was about $5.5 billion per year. This was enough CapEx to ramp 65nm with strained silicon, convert to 45nm with HKMG, and ramp 32nm HKMG. CapEx increases to $7 billion in 2010, $10.8 billion in 2011, $12.5 billion in 2012, and $13 billion in 2013 including a $2 billion prepayment to ASML (ASML), presumably for early access to 450mm equipment. The fourth 22nm fab would be a ground up fab that would cost $5 billion to build and equip. $1.5 billion in 2010 and $3.5 billion of the 2011 CapEx would cover the cost of the new fab. That would leave $17 billion total of CapEx for 2011-2013 over the baseline of $5.5 billion that worked just fine for the Tick-Tock process until 2010. $17 billion would build, from dirt, four more fabs over the four discussed in the fourth quarter of 2010.
What's Going On?
If Apple is going to Intel, a couple of things have to happen first. A secure source of mobile DRAM needs to be in hand. It is not logical to expect Samsung to support Intel or Apple after they have just been stripped of the Apple A chip business. Micron (MU) wasn't a solution. For some reason Micron can't build the low voltage, low power memories required for mobile. Their mobile business is a perennial money loser. Elpida solves the memory problem and Apple is loading them with mobile DRAM. My personal opinion is that Apple, Intel, and Micron are silently collaborating on the Elpida acquisition since all three of those companies view Samsung as a significant competitor for different reasons.
The second impediment to an Apple/Intel relationship is Intel's lack of a multi-mode voice and data LTE baseband solution. Again it is not reasonable to expect Qualcomm/TSMC to support Apple/Intel on baseband chips since Intel's success means Qualcomm/TSMC's decline to some extent. The Intel solution to the LTE issue should be here already, but it seems to be delayed. Mr. Otellini said as much in an investor presentation during the first quarter.
All of these things come together by the middle of the year so that Apple can say goodbye to Samsung. With a need for as many as 300 million sets of A chips and baseband chips, the Apple business is about $40 per phone or $12 billion. It would require about a million wafers, which would account for most of two of those extra four fabs. This is a really touchy thing for Apple. All the pieces have to be in place with no room for Snafus because even a burp in supply of those A chips and baseband chips will put Apple out of business...period.
These two problems, lack of LTE, and a secure mobile memory source have also inhibited a larger general Intel presence in smartphones.
Solid State Drives -- My Second Most Favorite Subject:
After listening to the SanDisk (SNDK) Conference Call, I am more convinced than ever that the NAND Flash memory capacity to support SSDs in Haswell based Ultrabooks simply doesn't exist anywhere in the semiconductor industry. The North Cape reference design from Intel could turn the PC business around very quickly. That is unless there are no NAND chips to build SSDs for them. I don't mean the kludgy hybrid drive of 20GB NAND + 500GB of spinning storage. That device just defeats the weight, power, and performance targets of Ultrabooks and convertibles. Some experts are predicting allocation of NAND by this summer. Where will the NAND Flash come from? Samsung? Samsung is reportedly buying NAND chips from Toshiba.
Intel builds SSDs today, so they must consider it a viable business. As Haswell rolls out, a high proportion of them will be used in thin, light, low power PCs and, therefore, use SSDs. Even if the SSDs were available, why would Intel allow Samsung, Hynix, or Toshiba take the SSD business from them? There is no question that Intel or the Intel/Micron joint venture can build SSDs at a high gross margin.
My feeling is that the Chandler fab shell could be destined to be a giant, dedicated NAND fab to ensure enough SSDs are available for the Haswell ramp.
Intel has much more capacity than required for the base x86 business of today. I believe the company has plans for this capacity that simply can't be discussed today. Foundry, Apple, and Solid State Drives could double Intel very quickly once started. I believe the starter's gun is cocked and ready to be fired.