I’ve always been skeptical about the added value of those many weeks spent trying to renegotiate and obtain fair prices for the electronic components we purchase. Given the number of references, it looks more like inefficient data exchange sessions than real negotiations to seek win-win situations. However, this is likely to change, and that’s truly good news.
The electronic component industry has always been prone to price discrepancies. It always strikes me how two different EMS/OEM/ODM can buy the same components with prices ranging from one to three times despite similar quantities.
While there are specific and concrete reasons behind those price discrepancies, I’m convinced that digitization will bring a new wave of transparency that will force commodity component prices to reach a worldwide equilibrium. This will lead to a new purchasing era that will be much more about speed-to-market, innovation, exclusivity, and gainsharing in supplier relationships than about harvesting three to five percent savings from contracts coming up for renewal.
Why such big discrepancies?
Before discussing digitization consequences, it’s important to understand why semiconductor prices are so different from one buyer to another.
Steep semiconductor learning curve
The learning curve is simply the long-term erosion of manufacturing costs related to production experience. The learning curve plays a critical role in the semiconductor industry since it is very, very steep.
As a former Intel executive told me: “We all disclose our fab utilization rate but the most secret metric is our yield.”
He then explained how a chipset is like a garden. To work limited-spaces as intensively as you can, you need to use space efficiently and to follow yield-maximizing strategies. In the garden, eliminate unnecessary pathways by planting crops in blocks rather than in rows, plant lettuce and spinach to grow in the shade of tomatoes, respect proper carrot spacing, plant crops that take a while — such as carrots — alongside faster-growing crops such as radishes, and so on.
The more experienced the gardener becomes, the more he knows how to design a high-yield garden.
Dies on a wafer are like small gardens. You can optimize their designs to shrink their sizes and drastically decrease production costs.
The more production experience is gained, the more you learn how to make an optimized product: designs are improved, die sizes are shrunk, process is debugged, wafer fab yields rise and even assembly yields rise. As yields rise more product is shipped per unit of cost, thus revenues rise while manufacturing costs stay constant or even decline, the average cost per device declines, leading to lower prices and higher volumes.
Price control mechanisms: Why doesn’t the learning curve help everyone?
Some customers will benefit from those lower costs: this is the case of high-volume customers on highly competitive markets (i.e. computer, communication, consumer). Those will surf down the curve and benefit from better prices.
However, some customers will only get a few percent price reduction every year, no matter what the production cost savings are. This is the case of high-mix, low-volume customers on less competitive markets (industrial, medical, military, aeronautic). Their products often have very long life cycles and huge safety requirements making it very complicated to redesign a product and change a component. As a result: manufacturers have no pressure to lower their prices since they are sure of keeping all the volumes for the next 10 or 20 years of product lifetime.
But this can only be true if manufacturers keep control over their distributors. Indeed, distributors have access to good prices thanks to high-volume customers. Manufacturers need to keep it clear that those good prices should be restricted to a few end customers only.