Mad Dog 21/21: Sell Phones
September 20, 2004 Hesh Wiener
Sometimes deception works, sometimes not. The Viceroy butterfly improves the odds of its survival by imitating the Monarch; predators don’t like the taste of the Monarch and avoid the tastier Viceroy. But even old Aesop pointed out that a subterfuge might not always work as expected. The wolf in a sheepskin infiltrated the flock, but when the farmer sought a sheep for the dinner table, he seized and slaughtered the disguised wolf. High technology has notable examples of misdirection, too, and, fittingly, the effects may not always be as they appear.
Mainframes are inherently very powerful computers, but most of them are sold disguised as less capable machines. This distortion is not a secret. On the contrary, IBM makes it clear to customers that all its mainframes except the most powerful models in any product family are identical to the high-end versions, but that they have been throttled back by firmware. IBM has long since used a similar strategy for the iSeries, restraining the power available to green-screen applications. Extending that strategy to encompass other performance characteristics of its midrange systems (such as memory or I/O) is, conceptually, a small step.
The practice of deception is not confined to the computer business. Wireless telephony carriers routinely pretend that the phones they offer are cheap or even free. Sign up for cell phone service, they say, and we’ll toss in a new phone at some bargain price. Like IBM offering a big processor discounted in both price and power, cellular service providers make no attempt to disguise their ploys. Like IBM, they boast about them. Every phone customer knows that some portion of each month’s phone bill goes to pay off the difference between the phone’s ostensible price and its real price. Anyone who buys a constrained processor knows that the real price IBM gets for the machine may be buried in software license fees or maintenance costs.
The clues are pretty obvious in the mainframe business. An IBM z890 mainframe running at 26 MIPS (called capacity setting 110) has a base price of about $200,000. The same box running at its full speed (called capacity setting 470) of 1,365 MIPS costs about $1.5 million. If IBM can sell the low-end hardware at a profit, even a profit of only one dollar, the big machine that has the same hardware would give IBM $1,300,001 in profit on the hardware alone.
IBM often displays a terrific knack at turning a profit, but if the company could pocket 85 percent of the price it gets for a big mainframe, its erstwhile competitors in the mainframe segment, Hitachi and Amdahl (now Fujitsu), would not have quit the business. They would have happily stayed in the game if they could have gotten even half that gross margin.
Dell happily sells servers all day long. It might, in a decent quarter, report a gross profit margin (covering all its products, including those in the low margin desktop computing segment) of around 18 percent. IBM said, in its second quarter report, that its gross margin for all hardware was 29 percent.
If this seems a bit murky, that’s because IBM has no reason to provide outsiders with the kind of detailed financial reports that would reveal its sweet spots and its struggles. Cell phone companies are no more open about the details of their activities, but the companies that make cell phones are not the ones selling the phones or related wireless phone services and the phone makers’ financial reports provide a lot of data about the handset business.
The mobile phone handset business is comparable in size to the server business. During the second quarter this year, worldwide sales of cell phones came to more than 150 million units, worth about $12 billion. This made the trade a little bigger than the total server market, which, market researcher IDC says, shipped about 1.4 million units, worth more than $11 billion, during the quarter.
The biggest maker of cell phones is Nokia, with about 30 percent of the market, a share comparable to IBM’s share of the server market. The number-two player, Motorola, has about 16 percent of the market. Nokia’s revenue from handsets was about $5 billion, so its average take was around $110 per phone. Motorola’s phone business, which has a different mix of products, reported revenue of $3.9 billion, or about $160 per phone, in the most recent quarter.
Amazon.com, which sells phones and service from most providers (but not the largest, Verizon), is just one place touting “free” phones, meaning phones whose price has been hidden in a bundle. Amazon has plenty of company. All over the world, handsets sell for less than their actual cost when they are bundled with services.
The reason for this peculiar pricing strategy is the same in the server business as it is in the cell phone game. If vendors priced every piece of the package independently, small servers in families built with all the large model hardware in the box and cell phones would give customers a terrible case of sticker shock. Most customers are apparently happier with reduced up front costs and higher costs spread over the term of the deal.
The useful lives of typical servers and mobile phones are longer than their product cycles. Mainframe product cycles seem to be about three or four years in length; midrange servers evolve at a somewhat faster pace. Typical server leases run three or four years, and many are begun in the middle of a product cycle, so lessees typically replace their machines no more often than every other product cycle.
Cell phones may only last a year before a newer model comes along, and few models are in the sales channels for more than two years. Carriers in the United States press users to sign two-year contracts, but one-year arrangements are very popular. Customers signing new contracts often get a new phone, and, as with servers, the phones customers choose may well be far into their product life cycles when they are tied to a new or renewed contract. So, for users of servers and cell phones, when a lease or a service contract comes to an end, there will always be a new generation of equipment on hand.
The end of a lease or service contract does not necessarily mean the end of the relationship between buyer and seller. Chances are, the lessee will keep the server or move to a compatible successor. Chances are, the cell phone customer will keep the phone alive.
Whatever the customer does, as long as a server remains in place, software and maintenance charges will continue. And, chances are, if the server is one of the smaller models in a processor family, the cost of software and services will include some amount that represents a hardware subsidy. The customer will continue to pay any such subsidy whether or not he brings in new hardware.
A cell phone customer is in a similar situation when his contract runs out. If he does not buy a new phone and service contract package, service charges will continue to run at the prevailing rate. The customer will, in effect, be repaying a carrier for a phone subsidy whether or not he gets a new phone.
This bothers some customers, and the bigger the subsidy carried in continuing costs, the more people paying for servers or cell phones they did not want or get. What pleases customers getting new equipment has the opposite effect on customers who want to let things rest for a while. Generally, these customers cannot persuade server vendors or cell phone carriers to reduce the continuing charges, but now that seems to be changing.
During the past few years, there have been developments in the server business and the mobile phone market that reflect the market’s response to what are perceived as excessive running costs for mature accounts.
In servers, the rise of Linux and other multi-platform software has enabled customers to delineate hardware expenditures from software costs. Linux systems running on more economical hardware have bled off enough business to give the vendors, as they say in the phone business, a wakeup call.
Every server vendor is addressing the emerging market for Linux platforms. IBM is making the most complex effort because it offers the most diverse collection of servers and it wants all its product lines, not just the machines with the X86 architecture, to support Linux. More than that, IBM will ask customers who want to run Linux on their legacy systems how much of their processor they want to devote to Linux and then sell that piece of the machine at a price that is at a significantly lower cost per engine than the rest of the server. In the case of mainframes, IBM sells only whole engines running at full speed for Linux, so the price can be less than that of a legacy engine but high enough to please Big Blue’s bean counters.
Some customers like the Linux-in-legacy concept, while others choose to run Linux applications on a dedicated X86 platform. It comes down to cases and to the rapidly changing landscape of server country.
As Intel and Advanced Micro Devices chips span more of the performance spectrum, companies building hardware with alternative architectures will continually be forced to higher ground. The shipment market share held by non-X86 systems will almost certainly continue to decline, even as the server business as a whole keeps growing. Revenue market share for RISC/Unix systems and proprietary mainframe and midrange gear could, however, stabilize–and probably will for a while, at least as server consolidation continues apace.
As it turns out, a very similar phenomenon is emerging in the mobile telephony market, and this development is already well entrenched across Europe. It’s only a matter of time before it has a big impact in the United States.
European mobile phones all use the GSM system, a communications scheme that is also used (on different carrier bands) by T-Mobile, Cingular, and AT&T in the U.S. (AT&T Wireless is going to be absorbed by Cingular in the near future.) Together, the GSM carriers have about 46 million customers. The non-GSM carriers in the U.S. (Verizon, Sprint, and Nextel) operate using a transmission scheme called CDMA. (Carriers still support some rapidly fading legacy technology, such as the analog AMPS communications method or the digital TDMA method.)
All the modulation schemes may converge as the next generation of wireless telephone is developed by the equipment industry. One possibility is an evolutionary networking concept called 3G, which is available in a limited way in Europe. In addition, WiFi, which also has properties that permit roaming and which has become immensely popular for office and home networks, could turn out to be a big winner for future cell phones. The carriers and the phone makers are aiming future services and equipment at customers who want portable access to the Internet, multimedia communications and other extensions of mobile telephony.
As the wireless technology wars unfold, all the phone makers and carriers are starting to reckon with a consumer revolution that has a lot in common with the rise of Linux in the server world.
As users of GSM phones know, their instruments include a removable circuit called a SIM (subscriber information module) card. The SIM card carries account information, such as a phone number, billing authorization data from a carrier and personal information like the owner’s phone book. Any SIM card (or at least one of the current 3.3 volt types) can work in any GSM phone, as long as the phone has not been programmed by a carrier to accept only its own SIMs. Some carriers in Europe and all carriers in the U.S. program SIM locks into the phones they sell. Many independent mobile phone dealers in Europe trade in unlocked phones (some in the U.S. are starting to), not so much because customers want them but because in Europe with its GSM standard, a dealer would like its entire phone inventory to be available to any customer regardless of the customer’s preference in carriers. Even when a phone is locked, what has been programmed can be preprogrammed. Cell phone dealers and phone repair firms routinely unlock SIM-based phones; the going rate seems to be $10 to $30.
The upshot has been the creation of a strong, growing market in SIM-only services contracts that are used in conjunction with a phone the customer already owns or which he buys separately. The cell phone market is also split another way. As an alternative to service contracts that are one or two years in length, there are also calling plans that do not involve any obligation to stay with a carrier. These pay-as-you-go plans are starting to take off in the U.S., and they are amidst a boom across Europe.
CDMA carriers’ phones don’t have SIMs. The carrier transmits billing authorization data to a phone, which is then considered activated by the carrier’s infrastructure. Each carrier locks every CDMA phone to its own service. As is the case of SIM-locked phones, CDMA phones can be unlocked, too. Once they are unlocked they could be moved among CDMA carriers, for example from Verizon to Sprint. The carriers and their dealers all offer service to customers who come in with unlocked phones, or with their own phones locked to their own networks; they just don’t make a big deal of it. Like the lower-cost SIM-only deals, the no-phone service tariffs can be less costly than phone and service bundles, with contract service providing lower per-minute rates than pay-as-you-go deals.
The market in phones and services that are not bound up with a contract differs from the contract market in a significant way: Their low cost attracts a different mix of customers than contract plans do. No-contract services appeal to tourists who want short-term arrangements, customers on a small budget, customers who want an extra phone to leave in the second car, and customers who lack strong credit histories.
A sure sign that there is an important segment of the cell phone market that prefers unbundled deals can be found on eBay. eBay is chock full of new and used phones, many unlocked, others locked and sold for use with a particular carrier. The phone makers might wish customers just threw away old phones when their contracts are up, but they can’t get their way. The phones are too darned expensive for that. It looks like the future of the mobile phone business will include a mix of advanced big-ticket phones (mostly sold bundled with hefty carrier subsidies) and less snazzy but cheaper phones, many of them unlocked.
If today’s major phone makers don’t chase the unlocked phone market, new manufacturers will undoubtedly enter the business. Some of them will produce basic, cheap phones, and a few will figure out how to make some that are best sellers. These manufacturers attack the industry giants from below. This up from under approach is the way the emerging phone makers in Korea, Samsung, Sanyo, and LG, grew to become the companies whose phones now constitute the majority of handsets sold in the CDMA market. These companies had a little more difficulty getting a foothold in European GSM country, where product design seems to have different standards than in the U.S. market, but by every indication they are on the right track.
Could the same thing happen in the server business? Of course. It’s the recurring nightmare of every company making servers for the Wintel and Lintel market: Some assembler in Asia makes its own dirt cheap, rock solid basic server and that’s all she wrote. It is this fear that drives Wintel and Lintel server makers to continually offer improved price/performance. And, as this worry ripples upward through the server market, makers of non-X86 servers will have to keep inventing ways to improve the appeal of their products.
Server makers and customers alike realize that strategies based on inertia and legacy lock-ins will suffer and eventually fail. The illusion of value is a very effective way to start conversations between vendors and customers. But as people whose calling is the cell phone business well know, ultimately success will belong to the folk who get their numbers right.