Component lead times exist for a reason
If you want to make a brand new set of silicon chips, start with a pile of sand.
The time has come. You’re ready to order all the parts for your board. You were certain that the parts in your design are quite popular. Ordering should be no problem. Why bother quoting early? You’re making so many of your product, you’re certain that you have to order by the reel anyway. This means you’ll be ordering from the factory (possibly via distribution). So the moment of truth comes and you get your quoted Bill of Material (BOM) back from your Contract Manufacturer (CM):
U13 has a lead time of 28 weeks.
Let’s break down what’s just happened and why you’re about to get yelled at by your boss.
Surely there’s stock somewhere?
When you’re ordering a significant amount of parts, say 50,000 of a particular component, there is a low likelihood that there will be enough stock in the world’s distributors. Yes, exceptions exist for very popular or very low cost chips. But the normal scenario is that distributors do not want to buy and hold this inventory. They make money by charging a premium on parts they sell and it’s unlikely they will be able to “turn” this inventory within a short timespan. For this example, let’s assume you’re designing in 50,000 small microcontrollers.
So there is no appreciable stock of your specific flavor of microcontroller in the world. Yes, you might be able to switch to a slightly different variant (higher memory, faster speed), but this requires lots of differentiation between boards. Not to mention a higher cost if you’re moving to a higher quality part. Read Voja’s account (below) of needing to design in multiple versions of chips. Even a small change in driver chips creates a logistical hassle that is carried forward with the product forever.
The key to building a new product is nothing more than taking the process one step at a time. Hopefully each step isn’t…medium.com
What is lead time anyway?
OK, so you’ve resigned that you truly need to order direct from the factory. You’ve received your horrific lead time, which means your hardware can’t be delivered in any less than a half a year from today (or more). What was the process that produced that lead time quote?
The important thing to know is that any time you are ordering from the factory, you’re buying parts that have not physically been created yet. It’s almost like you’re buying “futures” for chips that will exist sometime later. This is why you will very rarely get quoted a lead time of less than 6 weeks. The silicon processing, chip testing and eventually die cutting and packaging all add up to weeks of work. That is a best case scenario. Depending on the complexity of a chip, there could be over 300 steps to finish processing the silicon. Each step needs to go perfectly.
In a best case scenario, your CM quotes you a 6–8 week lead time. They are talking to someone at a distributor or directly at the manufacturer who knows the wafer starts planner. The planner normally works at the semiconductor fabrication facility (fab) and has looked ahead and seen that there is enough capacity to fulfill your order. They quoted the standard lead time back through the chain (or it is recorded somewhere and updated periodically). These numbers are sometimes viewable at online distributors. The online distributors are also getting quoted lead times to refill their stock when they run out, so they show that in case you’re relying on them alone. But the planner at a fab is the ultimate word in how soon you will get parts in-hand if buying reels at a time.
When the lead time gets longer
28 weeks. How can this even happen?
Well there a few things that will push out the delivery date and they stack up pretty quickly.
You are not alone. There are other people who also want to buy the same chip that you’re buying. Assuming the chip fab has only 1000 wafer starts per month for the part you want and customers are trying to buy enough parts to require 1500 wafer starts in a month, there will be some disappointed customers. If there is not enough capacity, the customers get in a queue and will get parts as they are available. But why don’t they just up the number of wafer starts?
Fabs can pump out many times the 1000 wafers listed above. Why not just start more? It’s because there are so many different types of chips that are being processed at the same time. Each calculation that is done for how many wafers to start is not just based around the demand for a part. It is also how much the company can make per chip. If there is a higher margin chip being made and sufficient demand for it, why wouldn’t the fab choose to make more money? This is a delicate balance. An overshoot of capacity for the higher cost chips floods the market with those chips. Supply and demand says…the price of that chip will then fall. So the people deciding wafer starts need to take a lot into account.
As mentioned above, the 6–8 week estimates are when things go well. When there are machines that go down in a fab or when there are poorly performing machines (as processes change over time), it can create bottlenecks. Now imagine those thousands of wafers moving through a fab. Each step that is delayed causes a build up of excess wafers at a particular step. Undoing this build up is much like automotive traffic after an accident has been cleared away. Yes, there is more control over properly restarting in a fab, thanks to automation. But the overall output of the fab can slow down because of this. Once again factoring in the priority of wafers, it’s possible that your particular part has been de-prioritized. Worse yet, you might be marked as a customer who is less important.
Money talks. When a large customer has an existing agreement for a certain number of parts per month coming from a fab, they will be prioritized over someone ordering just a few reels at a time. This means that your wafers which are sitting waiting for a formerly broken piece of equipment might be sitting there just a bit longer. If there are wafers that haven’t even been started yet and you’re a small customer and you’re buying a product that isn’t high margin, they might not even start your wafers. If you are lower priority, your wafers could be in line to wait for another few weeks until the backlog has been cleared. All of these things will be recorded as a longer lead time.
Fab output can also be affected by general supply and demand trends, especially when it comes to large companies. Remember how Samsung did a recall on a large swath of phones due to exploding batteries? All of the replacement phones are currently being built to send to customers. This is creating a global demand for memory that was not accounted for (as well as other components). If you’re a tiny company looking to order memory and call up a fab for a small order…who do you think they’ll prioritize? You? Or a company buying millions of units per year? Good luck with your memory purchase!
These trends are also seen after natural disasters that impact the supply chain. Back in 2011 when there was flooding in Thailand, the drop in availability of the global supply of hard drives was a prime example (though a slightly larger scale). It took many months to get production facilities back online and the most prudent purchasing agents went out and secured their supply of hard drives from the existing market nearly overnight. These world-level events can have large ripples into product availability.
It’s hard to explain just how interconnected the supply chain is. The fact that another company’s sourcing issues could impact the lead time of your project is really weird to think about. With fewer and fewer fabs in the world (even though individual fabs might have higher capacity), the supply chain can teeter on the edge of chaos from time to time. All you can do is adjust accordingly.
Around here, we recommend that you quote early and quote often. You might not be able to avoid the 28 week lead time on the part you had in your design. But you’ll be able to weigh the benefits of designing in a different part with higher availability during the beginning of your design cycle.
Have you been burned by lead times in the past? Let us know below.