It’s no secret that the cost and availability of Intel (or compatible) chips has a major impact on the prices customers pay and the performance they get. While there are often allocations or limited supplies of some specific parts (like the 486DX2 in the third quarter of 1992), a general overall balance has always applied.
That’s about to change.
As the accompanying chart shows, overcapacity on the manufacturing side is going to cause a staggering gap between chip supplies and personal computer consumption in the next 12 months.
Of course, not all X86 processors go into PCs. Some are used in appliances and electronic gewgaws, so a difference between PC units and processors is inevitable. But the roughly 40 percent “overage” for this year isn’t going to be absorbed in these other markets.
This oversupply will cause no small amount of strange happenings. Moving from an Intel-controlled, monopolistic market to one that is highly competitive, rapidly shifting and free will make current events in the Russian economy much more understandable for us than analysis by Peter Jennings and hours of videotape ever could.
Of course, rather than approach this issue rationally, the supply side of the business will resort to that tried-and-true practice of price cutting.
No doubt some of you will remember an earlier column in which I predicted that prices would continue to decline, despite Michael Dell’s pronouncements of the end of the price wars. The X86 oversupply is a key reason for this. Nevertheless, 1993 will be very different from 1992.
The difference stems from one simple fact: In 1992, Intel had the new, rationally structured 486 product line in place, literally without competition. The company essentially could flush the pricing of the 386 and bring in the 486 as a replacement. No matter which chip sold, Intel won.
Now shoot forward to 1993. Intel doesn’t have Pentium to turn to yet, and both Cyrix and AMD will have new products that create some real confusion about which 486 fits into which performance box. Clock-triplers and quadruplers are just the start. Look for 50MHz and 66MHz 486SXs and the quick death of the 486SL.
Despite its announcement, I don’t expect Pentium chips to be available soon in any quantity (InfoCorp calls for a maximum of 100,000 to be shipped this year as systems).
Currently there are power consumption/heat problems, and the size of the die is currently so large that any real production of the P-5 would be a money-losing proposition.
Market share at steak
Without a successor to the 486 available, Intel is at the mercy of its competitors’ prices on the 386 and 486. In fact, I’ll bet a steak dinner that we see some 386SX chips down at the $10 price point before we ring in the next new year. (At least we won’t be tortured by nearly as many of those obnoxious “Intel Inside” ads that have plagued our publications and TV shows.)
Even if Intel accomplishes a Herculean engineering feat and gets the Pentium going tomorrow, it still can’t save the bacon. To get real Pentium performance, you’ll have to replace or recompile major parts of your software library. Gee, just what I’d like to do. Looks like once this price slashing starts, no one can stop it.
So what does this mean to the customer? A few things.
First, buying program purchases or fleets of PCs by processor will probably drive you crazy. One option is to pick a price point or graphics/disk/form factor combination and let the processing power float upward.
However, for those who choose vendors that change BIOSs from product to product, this will drive LAN managers crazy.
Second, look for vendors to really unbundle all the freebies they have given away up to now. My bet is that Windows, mice and on-site service either won’t be free or will start coming with more restrictions than a frequent-flyer ticket.
This situation will make 1993 another year in which prices fall, margins are thin, and the bell tolls for a number of poorly managed or underfinanced suppliers.
For customers it will mean some new economics of products and a far more complex processor/price performance paradigm than has previously been the case.