Tuesday, February 27, 2007

Marvell (MRVL) Recovery?

Today Marvell (MRVL) was the one stock in the NASDAQ 100 to go up. I own the stock. The stock went up because Marvell reported its quarterly results after the market closed yesterday. In the analyst conference (my summary) it became clear that to a large extent Marvell bucked the downward sales trend that its rivals had experienced.

Marvell had a rough year in 2006, so a bad quarterly report would not be unexpected. They have the usual stock-option investigation going on. They bought the division of Intel that makes the XScale processor, among other things. Some analysts thought they paid to much money for it, since it was a money losing division for Intel. But they bought it for far less than what Intel had invested and it was not losing very much money. By combining the Intel intellectual property with their own circuits Marvell thought they could race to the lead in mobile processor technology. They were sure they could cut costs in the division, and yesterday said that process was going better than they expected.

While revenues from their other lines were about flat in the latest quarter (November 2006 to January 2007) this was better than the typical 5 to 10% downturns at competitors. Marvell thinks its numerous recent design wins mean that the normally seasonally down quarter we are now in will show a modest increase in revenues.

In 2006 Marvell's stock price peaked at $33.12 before sales to hard disk drive makers slumped. It has a high P/E ratio of 38, so it still has to grow quickly to justify even today's closing price of $20.46.

There are a lot of smart people in the semiconductor industry, but even among such a thoroughbred lot the Marvell crowd stands out. Whenever they enter a field they jump right to the cutting edge, often making products that companies who have been in the field for years cannot match. They are doing this now with video post processors for Blu-ray and HDTV.

I would not want to minimize the risk of owning a stock like Marvell, but the upside looks a lot better than the down side. One reason for caution is that while we know what revenues they have reported, we won't know how profitable the company really was until their stock-option accounting is complete and they release the results. Again, the information is there in the analyst conference (my summary) to make a good educated guess.

Tuesday, February 20, 2007

Search Wars Not Over

Last weekend I made substantial additions to a Web site for my wife's non-profit project, www.TapestryOfTheCommons.org. A lesser version of the site had been up for months. My wife complained today that the site did not come up on Google. I assured her that since there were links from other sites into her site, Google would have picked it up, but maybe not prioritized it yet. I was wrong.

At Yahoo search "tapestry of the commons" came up 3rd in the results. At MSN search the site came up number 1. Google truly did not list the site at all. Since I have an account at Google (for AdSense and this blog site) I was able to submit my wife's site. But I should not have needed to.

So I don't think the search wars are over. Like many people I tried Google early on, based on word of mouth. I was mostly using Yahoo and AltaVista to do my searching. Google, back then, was free from annoying ads. That, I guess, was the cost of entry. Yet more often than not Google gave better results; it is still my first line search engine.

But to my taste Yahoo gives better results for shopping and for investment related searches. I have gotten some good results from Ask.com when Yahoo and Google failed me. Now that Microsoft showed me it can be number 1 in results I want, I'm going to try it more.

What does this mean for investors? Some caution and hedging of bets seems smart. I own Microsoft stock based on its ability to sell operating systems, Visual Studio (which I use and love), and SQL Server. If its Web search division starts making more money, Microsoft stock will look cheap in retrospect. And if Yahoo or Ask make any gains against Google, the high Google price-to-earnings multiplier will look almost as silly as Year 2000 stock prices based on "Web page hits" rather than on earnings.

Don't get me wrong, Google is a great company. It employs lots of smart people; there is plenty of upside potential in its earnings. Google, so far, has excelled at monetizing Web searches and ads.

See also my summaries of the latest quarterly results and analyst conferences for GOOG, MSFT, and YHOO.

Wednesday, February 14, 2007

Applied Materials on 45 nm

Intel, locked in a battle-to-the-death with AMD, recently announced it had won the 45 nm war. Within hours the same day IBM announced that Intel's accomplishments, even if real, were no big deal because IBM has the same capabilities. That is good for AMD, which is partnering with IBM.

45nm means that semiconductors are constructed on a grid with cells 45 nanometers wide. Right now the most advanced silicon actually being sold is 65 nm, which Intel started selling mid-2006 and AMD started selling in January 2007. Smaller sizes mean more gates and computing power in a given area. You can, of course, waste those extra gates with a bad design.

Applied Materials executives, in the February 13th analyst conference (see my summary), gave their perspective on the transition to 45nm. They make much of the equipment used in fabrication plants to make semiconductors. They sell that equipment to just about everyone, including AMD, Intel, and IBM. They should know a thing or two about it.

They claim to have the best process. Whatever process they have they will license to anyone who will pay for it. They have worked with both IBM and Intel on the issue.

What does this mean for investors? It probably means 45 nm won't be an issue. One company or another might get an edge, but that is unpredictable right now and won't be the main determining variable. Every time process size shrinks there is a lot of handwringing; some day further shrinking either will be prohibitively expensive or just won't work.

Winners will be those who use 45 nm most effectively and are most effective at marketing. Intel is the long-reigning champion, but looks punch drunk and ready to fall. Intel fired thousands of engineers last year while AMD was hiring.

Meanwhile consumers benefit from the intense price competition. Short term investors probably want to avoid Intel and AMD, despite historically low stock prices. Long term, any letting up in the price wars will mean big bounces for both the companies.

Applied Materials on 45 nm

Intel, locked in a battle-to-the-death with AMD, recently announced it had won the 45 nm war. Within hours the same day IBM announced that Intel's accomplishments, even if real, were no big deal because IBM has the same capabilities. That is good for AMD, which is partnering with IBM.

45nm means that semiconductors are constructed on a grid with cells 45 nanometers wide. Right now the most advanced silicon actually being sold is 65 nm, which Intel started selling mid-2006 and AMD started selling in January 2007. Smaller sizes mean more gates and computing power in a given area. You can, of course, waste those extra gates with a bad design.

Applied Materials executives, in the February 13th analyst conference (see my summary), which makes equipment that any (with billions of dollars) can buy to create a semiconductor fabrication plant, said they are well-prepared for 45nm and have done their own research.

All this says that 45nm processes are mainly a red herring. One process may be better than the other on one parameter or another, but any company wanting to move into 45 nm production will be able to license the technology.

Intel is much better at announcing that it has won battles before they have begun, and at taking out ads in publications that actually review products, than it is at putting out products designed to meet its customers needs. No wonder people continue to shift to AMD processors.

Monday, February 12, 2007

Understanding Akamai

After the Akamai analyst conference the other day I was pretty enthusiastic and thought I might buy a bit of Akamai (AKAM) stock. The price shot up in after hours trading after the conference, so others were thinking the same thing. Friday the stock dropped 5%, not because of any fundamental change but because the hot money crowd that bought on Thursday found themselves in a generally down tech market.

I realized I did not really understand what Akamai did. Speed up the Internet for their customers, sure, but how?

Apparently Akamai realized that while the Internet is supposed to function automatically, in reality that functionality is not always very efficiently. Simplifying somewhat, when you enter a URL in your browser, say www.YouTube.com, the browser sends out a request for information from the URL. Out in the internet a "name server" translates the YouTube into a number, the IP (Internet Protocol) number of YouTube, and the first router (more than likely made by Cisco) uses its map of the Internet to either (1) forward the request to another router or (2) if it is connected to YouTube, actually forward the request to the YouTube server farm. There the request is processed and sent back to the originating address, again through at least one but usually many routers.

The existence of many routers means there are usually multiple paths between end points. Some paths have more routers in them than others, and routers produce delays, usually very slight ones. But an overloaded router can be a problem.

Akamai has an array of servers that monitor the state of Internet traffic. They use that information to do two things for their customers. One is that they are able to optimize routes - choose the fastest routes for their customers. The fastest route may shift from second to second or on any other time scales. The other is they allow content to be held near the end points that are frequently requesting it. Thus if a particular YouTube video is being watched by large numbers of people on a given day, instead of going to the YouTube servers every time a request is made, the request can be answered from a cache that is time-wise closer to the end user. Akamai claims to handle 20% of Internet traffic.

Akamai's servers, in addition to monitoring Web traffic, can hold its customers content and server it up in an efficient manner.

This is important to Akamai's customers because there is little point to serving up content if the end consumers have gotten bored and clicked onto something new while waiting for all those routers to get their act in gear.

It is a great company with good ideas. But I have decided not to buy the shares. Why? By my criteria (but obviously not by some other investors) the price of the stock is too high. I think of it in terms of market capitalization, which for AKAM ended today near $8.6 billion. The company is predicting that revenues for 2007 would total $610 to $625 million, with adjusted (non-GAAP) net income around $200 million. Now that is a great model, getting almost 1/3 profit out of your revenues. But the high market cap to sales and market cap to profits ratios are too rich for my portfolio, given other choices available. I have seen too many companies on upward trajectories level off their growth rates to draw a straight line several years out assuming these ratios will remain this high. Three years of continued high growth could prove me wrong.

Thursday, February 8, 2007

Semiconductor Inventories and Demand

I am fascinated by technology, by the businesses that create, manufacture, and sell it, and by the human factors in management. Listening to executives from Atmel, Linear Technology Corporation, Microchip, Maxim, Texas Instruments and Xilinx at their analyst conferences explaining Q4 results, all these factors converge. Q4 2006 was not the best quarter for the semiconductor industry. While most companies had year-to-year improved revenues, most were sequentially down from Q3. Many predicted further deterioration in revenues for Q1 2007. The down trend was widely attributed not to decreased demand, but to reductions of inventories at customers. To some extent seasonality is involved, as when end-products are aimed at consumer holiday buying and had to be produced in Q3 in order to be incorporated into products and shipped back across the Pacific to sell to spendthrift Americans.

So what investors want to know is whether this is the beginning of a downward trend, or just a temporary adjustment of inventories. It is also important to consider how companies are competing and may buck any trend.

Atmel (ATML) was down 5% sequentially and up only 2% year-over-year. They have never really recovered from the 2001 bust, so don't indicate any kind of trend.

Linear Technology Corporation (LLTC) did worse, down 8% sequentially but up 1% from the year-earlier quarter. They expect revenues to fall another 4% to 7% in this March quarter.

Microchip (MCHP) was down 6.3% sequentially, but up 6.9% year-over-year. They are optimistic that Q1 2007 will be the bottom of the cycle and are predicting that revenues will be flat sequentially. [I own MCHP stock]

Maxim (MXIM) escaped relatively unscathed, with revenues down only 1% sequentially but up 11% from Q4 2005. However, they are predicting that Q1 2006 revenues will be down 3% to 6%.

Texas Instruments (TXN) revenue was down 8% sequentially but up 4% from year-earlier.

Xilinx (XLNX) revenue was down 3.5% sequentially, but flat from the year earlier. They guided to flat to down 5% sequentially for Q1.

I think there was some tightening of end-user inventory because of uncertainty over the 2007 economy. Since the 2007 economy looks rosy at the moment, I think that when demand holds up the end users (electronics device makers) will have to start rebuilding inventories. But remember that each company produces chips in multiple categories. Slack or robust cell-phone demand will not impact all companies equally. Defense spending or cutbacks hurt some companies more than others.

When we get Q1 2006 results I'll be looking back to see which executives gave good guidance. Of course unfounded optimism is now suspect: it seems like an attempt to manipulate share prices upward.

Monday, February 5, 2007

Did Intel Bribe Dell to Exclude AMD?

Did Intel bribe Dell to use only Intel chips, even during a period of time when AMD's microprocessors were generally acknowledged to be technologically superior to Intel's?

According to an AP story, "A spokesman for Intel, which was also named as a defendant, says the company has done nothing wrong and that Intel's payments to Dell are legal." Dell declined to comment.

The lawsuit making this allegation was filed by William Lerach, a San Diego lawyer. It states that Dell received rebates from Intel of about $1 billion per year for several years.

Rebates in and of themselves, of course, are not illegal. But Intel and AMD are in a peculiar relationship, or at least were until the last couple of years. Back in the early 1980's IBM chose Intel's microprocessors for its IBM PC design that became the standard design for the industry. It was not that IBM could not design or produce such a chip, or that Intel's chips were supperior to others available at the time from Motorolla or Zilog. IBM did not want to use its own chips because it had already been accused of monopoly practices and suffered through a hairy anti-trust case. IBM even let other manufacturers copy its design and sell their "clones" for a lower price.

Once Apple knocked itself out of the game by overpricing its Macs in the mid-1980's, most of the microprocessor market belonged to Intel. That meant it had obtained IBM's enviable position: it looked like a monopoly, whether it was or not. So Intel began using the classic strategy for this kind of situation: it allowed a second source of microprocessors that were compatible with its own. More than one company competed for that sliver of the market, but AMD won it.

Intel needed AMD in order to not be a monopolist, but Intel had to take it into consideration when determining its prices. Which is precisely the idea of anti-trust legislation.

But then Intel fumbled the ball when the time came to switch from 32 to 62 bit microprocessors. AMD designed processors, the Opteron for servers and the Athlon for client computers, that computer manufacturers and end users really liked. AMD's market share began to climb.

Intel needed to buy time to re-engineer its processors and re-align its road map to meet this unexpected challenge. It did everything it could to convince computer makers to either not use AMD chips.

Dell is not known for its design innovation. It is, or was, a great business model for chugging out profits, but Dell never has had the legions of engineers available at, for instance, HP and IBM. That was one of its cost advantages, until too many IT professionals decided they needed to buy AMD Opteron and Athlon based machines. The failure to turn to AMD in a timely manner was one of several reasons Dell's profits dried up.

Whether they were bribes or legal rebates, properly accounted for or improperly accounted for, those billion dollar transfers from Intel to Dell were anti-competitive and hurt stockholders of all three companies. Intel's stockholders lost cash. Dell's got some cash, but lost more in opportunities. And of course AMD stockholders lost quite a bit of potential sales.

I don't know what the courts will decide. The lawsuit may have merit or not. But at least now we know more about how this game is played, and can adjust our portfolios accordingly.

For the record, I own AMD stock, but not Dell or Intel.

You can see my summaries of analyst conferences by following these links: