I find it instructive to look at COMPAQ's plans about I2O, as perhaps being even more important than HP's as an indicator of what to expect. My sense from their white papers is the following:
(1) They have been working with intelligent I/O devices for servers since 1989, and use them in such things as RAID controllers. Their experience with their own efforts seems somewhat mixed in that they obviously found it difficult to get all the expected speed advantages out of the intelligent devices - due to delicate timing problems interfacing the intelligent devices and the CPU over the system bus.
(2) Nevertheless, they are members of I2O and see the advantages to a standard for intelligent I/O, and will be using I2O for ALL new intelligent devices as well as any redesigns of existing intelligent devices.
(3) Their ProLiant servers are I2O capable.
Meanwhile, since COMPAQ's white papers were written, early timing results of I2O implementations suggest that COMPAQ's past technical difficulties integrating intelligent devices are not affecting I2O implementations (because of the highly capable PCI bus?, or better design?). This would suggest that any mainstream manufacture, like COMPAQ, will quickly rethink their I2O strategy, and speed time-to-market considerably.
Incidentally, I have seen the HP announcement that specifically mentioned desktop PCs as being future recipients of I2O, but is there a concrete reference for Intel declaring their intention of putting I2O on all motherboards? I have always expected this, but where is it said, and when will it happen?
I have been projecting only about 5% of Pentium Pro/Pentium II non-NT PCs would be using I2O in 1998. That generates about 2.36 million I2O targets, less than Pentium Pro/Pentium II NT PC/Workstations. If that figure for non-NT and NT PCs were to change to say, 50%, the projected number of targets would sky rocket next year to over 40 million targets, probably yielding over $60 million in royalties, for a net addition to EPS over $1.25.
This reminds me of China. It's one thing to say the Three Gorges Dam is a huge $30 billion project, employing 15,000 workers at the building site. But how big is this project relative to China's needs? Insignificant. In order to absorb hundreds of million of workers migrating to cities in ever-increasing numbers, it would take the equivalent of 10,000 Three Gorges Dam projects. Similarly, it is natural fro all of us to toss around observations like, "I2O will be on all future motherboards" without fully realizing the enormity of that statement.
Ironically, these awesome numbers answer the question about why management has asked analysts not to include I2O in their estimates - until proven out over a couple of quarters. They are too big! You say 100% of motherboards. That's worth $2.50 in 1998 alone. What if its only 10%, for a lousy $.25. Lousy? Since when is $.25 from one project lousy? It is if you priced the stock on $2.50.
There is no conceivable way WIND management has enough information to know whether the magic number next year is 10% or 100%, and so the only intelligent thing to do is to try to keep the lid on excessive estimates. I believe this is the only reason behind management's conservative guidelines, which has nothing whatsoever to do with concern about I2O not becoming a widespread standard.
Given all this, Mark's past discussions about the cost-effectiveness of the I2O investment by WIND would only make sense to someone who would question whether the dollar in a slot machine that hit the jackpot was well-spent.
I have said it before, and I will say it again, I2O alone more than justifies WIND's current price. You get all the other thousands of projects for free.
Allen |
