Six months back we took a look at some of the historical patterns around fab and foundry investments. We dug out on old report from Jeffries analyst Mark Lapacis who laid out the history of capex required for fab operators to stay at the leading edge of Moore’s Law. His rough rule of thumb is that companies need to have annual revenue that is double (2X) the cost of a leading edge fab. Companies that fell below that level struggled to keep up and by the time revenue fell below 1X, they would abandon the race entirely.
This is an old report as evidenced by the fact that he was trying to determine if TSMC could stay in the race while Intel was the foregone winner. Times have, of course, changed since then. When we ran the numbers in February, we found that going by forecasts at the time, Intel would just barely stay above the 2X level. However, given last week’s news, forward estimates for Intel have fallen. Current consensus is looking for Intel to do $48 billion in revenue. With the current cost of a fab around $25 billion, Intel has now fallen below that 2X threshold and entered the danger zone.
This it not a good indicator. As we noted back in February, not only does Intel need to boost revenue to the point that it can support its current development, but it has to keep that pace going on into the future. Our forecast at the time was by 2028 Intel would be back in the ‘Danger Zone’, underscoring its need to win customers for Intel Foundry. And the current numbers throw even that scenario into doubt.
That all being said, it is important to remember this is just one framework for analyzing the company. And it is fairly squishy – for instance how much does a fab really cost? There is nothing deterministic about these numbers, Intel can still turn things around. It’s just that no company has ever done that before.
Leave a Reply