Has anyone seen any interesting write ups or analysis on what has been happening with network equipment pricing and availability in the United States over the last couple of years?
Everyone (or at least I did) knew by March 2020 that what manufacturers were doing wasn’t really going to work anymore going forward and yet it’s 2 1/3 years later and we’re still looking at a 1+ year lead time on basic products.
Not to be glib but I am pretty sure I could’ve devised a process to build ethernet switches by hand with a soldering iron by now.
Not to mention how many additional supply chains could’ve been established since then.
Did Huawei actually serve an important purpose in the market?
Did we put too many eggs in the Broadcom basket?
Did the industry come together and “agree” that the time is nigh to charge $16,000 for a Dell S4148T but also that it would take 9 months to get it at that price?
I think it would be fascinating to learn more about what is happening in this market.
Please share any resources on or off-list.
Thanks and have a great weekend!
The underlying problem is silicon Fab capacity. It has nothing to do with the actual manufacturing of the products once all the components arrive. If you don’t have components for your product, a new order for components will take a year to arrive just because the factories that turn raw silicon wafers into computer chips are backlogged 12 to 18 months.
Note that all of the existing factories are running around the clock, and although additional facilities are being built it takes time for them to be completed. I’m also assuming that there has been some question about whether this is short term or long term demand which would influence whether you spend a billion dollars or more building a new fab.
The YouTube channel Asianometry has some good insights into the underlying supply chain problems:
Deposits that the issue isn’t with leading as chips, as you might think, but with so-called “trailing edge“ chips: microprocessors and support circuits that make up the bulk of electronic components, including routers and switches.
Asianometry points out that trailing edge components account for 50% of sales in the chip market, and given their much lower price of just a few dollars each, they represent many times as many units. This makes them a much larger factor in the supply chain problem.
Everything is finally keyed to “just in time“ manufacturing, with very little component supply in the pipelines. When Covid hit, those pipelines dried up and everything collapsed.
Okay so I suppose how many years are we allocating mentally for them to take any action?
If you look at the NANOG archives, this was discussed just a month or so ago, somewhere in the middle of the Juniper MX204 EOL discussion, IIRC.
I’ve been seeing a number of analysts with models which show a brief reprieve in 2023 then a regression (i.e. back to where we are today, or even worse) from there out for another 4-5yrs, and no-one’s putting much credibility in models much past 5yrs right now. Which is good, because some of those models say we’re screwed for the rest of our lifetimes.
I don’t know the details of why, although I do recall that some of it is based on queueing theory, so the ultra-short hyper-efficient JIT pipelines our society has spent the last 50yrs building likely play a part.
FWIW, I saw something about vehicle manufacturing where a factory rep was proudly proclaiming that they’d optimized the JIT process to where they needed some specific part – and engine, I think? – delivered from the part manufacturer twice daily… seemingly completely oblivious to the supply-chain risks involved.