The 5Gx is nothing more and nothing less than an extremely cheap non-upgradeable motherboard. Non-upgradeable motherboards have been tried before, and failed, or at least did not achieve the degree of sucess that you envision. Remember the IBM SLC2/66?
Licensing software, like Java, the Apple Operating System, or MS/DOS is a no brainer. You want everyone to use it. Licensing the right to manufacture hardware is a separate question. What are the issues, and why are they different? Well, licensing software you establish a standard. Having the Gx be a standard would be good. So far so good. But that unfortunately is where it ends. Manufacturing cost on software is a very low portion of the total price. At the low prices of the Gx, manufacturing cost is a significant portion of cost. Software is unique and diffent. The Gx is the same as other cpus, only cheaper. It has no independent appeal - only that it is cheap. As such, it adds no value to a product, but only reduces cost.
So why would you choose to license. Licensing reduces the risk to Cyrix, but reduces the profit. By licensing the Gx to other makers, Cyrix is able to eliminate the risk of having the product not sell, or the risk of mis-guessing demand and overproducing chips. In order to compensate other makers for taking this risk, of course, Cyrix has to give the other makers most of the profit on the sale of the chips. Depending on how risky the product is, it could be wise to give up most of the profit in order to avoid the risk of being stuck with a million chips, and in order to have a small but safe profit.
Another reason for choosing licensing is if you have a shortage of production capacity. By licensing the Gx, Cyrix could free up additional capacity to make more 6x86 chips, which of course Cyrix needs badly. On the other hand, if capacity was a problem, Cyrix could always contract to have other makers make the chips on a Fab basis, so you are back where you started: By licensing you reduce risk but give up profit.
The final reason for licensing the product, and obviously the reason you favor it would be that you believe that by having additional companies out trying to sell the chips they each will have their own accounts and you get more sales people working on the problem. Arguably more companies use the 6x86 because IBM is out selling it to certain accounts, and because SGS Thompson is out selling it. Of course this must have been in Cyrix's mind when they elected to sell wafers to SGS - that SGS had close ties to certain European makers and therefore could land accounts that Cyrix acting on their own would be unable to. On the other hand another ramification of having multiple people selling the same chip to the same customers is reduced price - just look at the way IBM has driven the price down.
I think that realistically if Cyrix wanted to license the chip, they could get a license fee of about $1.50 per chip, or maybe $3 including the BIOS chip. Add in manufacturing costs of maybe $30 for the two and the cost to the licensee would be about $33. Give them a 50% markup and the cost would be about $50. Cyrix apparently is planning to sell the chips for about $80 or a little less. If they can make the chips for say $40 (allowing $10 profit for the fab), then they make $40/chip profit. If they sell 3 million, and make $40 each, that adds $120m in profit, or $6/share. If they only earn $3/chip they need to sell 40 million to earn the same amount. I refuse to believe that they could sell 10 times as many by reducing the price by $30. Even if they could, they could reduce the price themselves and make $10/chip, which is more that $3/chip.
Now you argue that people would pay $20/chip in royalties. Add in the $30 in manufacturing costs, and their costs would be $50, so they would have to sell the chips in the $75-80/range in order to make it profitable for them. If they are selling the chips for $80, do you think they can sell 10 times as many as if Cyrix alone is selling them for $80. Since Cyrix makes $40 if they sell them themselves, but only $20 if they license them (assuming anyone would pay that much, which I doubt), they would need to sell at least twice as many at the same price in order to make it worthwhile.
My suggestion would be for Cyrix to make the chips themselves, and sell them for a price in the $50-80 range. Then they should license them to other people for use as embedded processors for a fee of about $1.5 apiece. This chip could have many other uses for such purposes as disk drive controllers, in cars, etc. Who knows what all it could be used for?
Carl |
