A recent article in Supply Chain Digest noted that Intel needed to reduce supply chain costs for its Atom chips by 80% to stay competitive and that, so far, it’s managed to shave costs by 70%. For most companies, this type of cost reduction would be unthinkable. But is it really?
Let’s think about supply chain costs and the opportunity for bloat. We have inventory. We have distribution. We have packaging. We have duties and (value added) taxes. We have inputs. We have production. Etc. Any one of these costs can be extremely bloated. You might think you need nine weeks of inventory when you only need three — there’s 66% bloat. You might think you need air freight when better planning could allow for ocean freight. There’s 50%+ bloat. You might think you need a certain type of packaging when a redesign could be just as sturdy with half the material. There’s another 50% bloat. You might think a tax is unavoidable, but a slight change could eliminate the tax. For example, under some free trade treaties, shipping the printer cartridge separate, instead of pre-loaded, eliminates a VAT. You might think that you need expensive raw material X, when a slight redesign would allow you to produce a better quality product with cheap raw material Y. A slight, lean, re-design of your manufacturing layout might double throughput. Etc. It is a possibility, especially if you have a lot of levers.
However, Intel only had one. Cost-service tradeoffs were off the table, as the chip had to work. Computer chips have about the highest value-to-weight ratios that you can get, which leaves little or no room to improve distribution costs. Intel’s packaging has been getting smaller and smaller over the years, so little room was left in packaging. Once a chip is designed, the raw material needs are locked in (and it still takes years to bring a fundamentally new chip design to market). Due to the nature of chip fabrication, once the plant is built, the process can’t be (significantly) changed. A chip is a single well-defined component, so there’s no wiggle room where duties are concerned. All that was left for Intel was the inventory lever.
For it’s traditional chips, which sold for $100 or more, as compared to the $20 or less selling point for the new Atom chip (which was being designed for the low-cost mobile device market), Intel operated on a nine-week total order cycle time. During the first seven weeks there were typically a large number of order changes — over 90% of orders were changed after initial placement. This led to significant inventory builds as factories spent considerable time optimizing and re-optimizing the factory schedule. However, if Intel could move to a true “make-to-order” model, reduce cycle time to two weeks, plan within four days and allow no changes after that time, the small hit in factory utilization that might would result would be swamped by the reduction in inventory, storage, and handling for all the chips what were currently routed to the DC.
Even though the internal perception was that you can’t truly build to order in the chip industry, an Intel factory was chosen in Asia as the pilot plant and an iterative approach that incrementally ratcheted down from nine to (a little over) two weeks was implemented. The end result was that supply chain costs were brought down from about $5.50 per chip to about $1.40 and that the minor hit that was expected to factory utilization did not materialize.
But is it enough? While Atom chips may be selling for close to $20 now, the market will very quickly force the cost down to about $10 (or less) per chip as more chip makers focus their sights on the mobile markets. Intel expects to get its supply chain costs down to under $1.00 per chip in 2010, but if disruption causes chip prices to fall rapidly, that could still be in excess of 10% per chip! And while supply chain costs of 10% would be welcome in some industries, it’s rather high in the chip industry where they have been traditionally been around the 5% mark, or less.
While I think supply chain will be Intel’s saviour, I think their work is just beginning and that they will have to pull off multiple revolutions to keep Intel at the forefront. Especially since Intel’s legal bill is going to skyrocket yet again as it just agreed to pay $1.25 Billion to settle disputes with AMD and has just been accused of bribing computer makers. This is on top of the $1.45 Billion the E.U. fined Intel in May.