I’ve mentioned supply chain issues in the past. Now they may be intersecting with a new problem. Let’s dive in.
One source who should know tells me that the primary supply chain issue right now is image sensors. Getting new ones not only into production but at the quantity the market might demand is apparently really tough right now. The Sony A1 II supply, for instance, is probably choking on image sensor scarcity at the moment (other low supply parts might be impacting shipments, as well).
What apparently is happening is that when you find that you need more image sensors than you projected (and originally ordered), getting additional production on fab just isn’t going to be a simple phone call asking for more.
You might have noticed that the final CIPA shipment numbers for 2024 exceeded the original projections back at the start of the year. Those projections are made from the conglomerated individual manufacturers production forecasts. And those original 2024 projections could be characterized as “another year like 2023.”
If you look at the monthly shipment numbers, though, you’ll see that 2024 started “hotter” than 2023 in January, then 2024 mostly mimicked 2023 until September, at which point 2024 suddenly showed growth that didn’t happen the previous year. I’ll bet some of that was image sensor orders finally catching up. The overall ILC bodies trended about 10% higher for the year. Thus, if you said to the fab you needed 100 image sensors at the start of last year, you found you needed 110. Overall, the Japanese makers needed about 600,000 more image sensors in 2024 than they did in 2023. That by itself takes up a lot of fab time and wafers.
Now let’s say you want to introduce a new camera with a new image sensor. Squeezing 10% more existing sensors out of the production lines is one thing, but now you want another 2%, but from sensors that have never been done before (or are special in that they are stacked and require more and better production time)?
Let’s take Nikon for example. They have basically six image sensors (20mp DX, three variations of 24mp FX, two 45mp FX). Of those, four are basically older image sensors (20mp DX, two 24mp FX, one 45mp FX) and two are “special” and newer (24mp FX partial stacked and 45mp FX stacked). Everyone keeps asking when Nikon will drop a new >45mp FX sensor, a >20mp DX sensor, or another “special” sensor (e.g. stacked). Historically, their last two “new” sensors were in 2020 (Z9) and 2024 (Z6III). So age-wise we get:
- 10 years old — 24mp FX (Z5, Z6, Z6II, Zf, the latter one with minor changes)
- 8 years old — 20mp DX (all DX Z’s)
- 7 years old — 45mp FX (Z7, Z7II)
- 4 years old — 45mp stacked FX (Z8, Z9)
- 1 year old — 24mp partial stacked FX (Z6III)
If you look at the other makers, you see similar patterns. Sony’s “newest” image sensors are 2021 (2) and 2023. Getting new sensor tech on fab is slowing down and large sensor fab utilization apparently remains at 100% (or more), so new image sensors are coming more slowly now.
So image sensors are a big part of the supply chain issues that the camera companies are struggling with.
What’s the new problem that intersects? Little or no demand for the same image sensors. Oops ;~).
Recently I've seen three different photography sites complaining that they didn’t get any affiliate commissions for newly announced cameras, because no one clicked on the link and ordered one.
We’ve had six cameras introduced so far this year, only two of which have new image sensors. Three of the older sensor cameras that were just introduced have little demand, and were part of those affiliate link failures I just mentioned. Another is a new body for an old inside (it’s an OM-3 on the outside, OM-1 II on the inside). So how does the market continue to grow? By selling more bodies with the same image sensor? I don’t think so.
It’s not that you can’t create a “better” camera with an “older” image sensor. The Sony A1 II and Nikon Z50II are good examples of re-invigorating a model without needing a new sensor. The Fujifilm X-M5 is a “different” camera than Fujifilm has done in the past with the same image sensor. But in general, it feels to me that a number of companies are not quite getting things right. The biggest part of their long-term success will be upgrading, not new customers buying older sensor products.
It doesn’t help that CMOS image sensors have pretty much climbed the hill. We’ve been stuck on the top of a plateau for a while now with dynamic range (DR). As I like to put it: “in most circumstances, current cameras record the randomness of photons accurately.” Moreover, the real challenge in DR that hasn’t been breeched is saturation, not noise floor, which is where everyone's misplaced attention has been.
The primary thing that new image sensors seem to be concentrating on is speed. Speed helps both with quantity of pixels the sensor can have, as well as things like removing the physical shutter and lowering costs (e.g. Z8, Z9). But advancements in speed are essentially driven by a form of Moore’s Law, so it’s tough to really break a new sensor out in a way that others can’t also do. Another aspect of this is that dedicated camera image sensors have traditionally been done on old, larger process fabs where you can't get "easy" speed benefits by using smaller process size. Unfortunately, fabs that can do small process are the ones that are most utilized at the moment, to the point where at the current "best" processes you have maybe four customers locking up all of production. (None of those are camera makers.)
Meanwhile, we just keep getting “more” piled into the bodies themselves via menus. I got a headache staring at the details for the just-announced Panasonic S1RII's video capabilities: over 60 different settings, which produces six different crops, has compression interdependencies, and varies in bit rate. Plus there’s some other footnotes you need to apply to fully understand a setting!
I really have to ask how many people need 60 different FullHD, 4K, 6K, and 8K settings. How many can even keep track of what all of them do and how they differ? Are all these settings actually used by someone? Is that what we really want camera makers to concentrate on: adding more and more choices?
Yes, I’m sure there’s some customer somewhere that will appreciate all that choice. It isn’t me, and I’m pretty sophisticated as a user. My number one request of Nikon—I predominately use Z8’s and Z9’s for my work—is to provide a truly useful save/restore settings function, not to give me another video option. A settings system where I can name my camera settings files (birding, wildlife, football, ice hockey, portrait, etc.) and restore them with a single command (or better yet, button press, e.g. Cycle saved settings).
Which brings us to the thing that’s really holding the camera industry back: they think they’re designing hardware, not software. If you think about what a modern mirrorless camera is, it is data coming from an image sensor processed by a SoC (System on Chip), coupled with some form of interface where the customer tells the SoC what to do with that data. “Processing” and “Interface” are software tasks, not hardware ones (though they may involve some hardware).
So while the Tokyo camera makers all wait for new image sensors and more production of them, they’re really staring at the wrong problem. They’re failing at software, not hardware.
You can think about that this way: if I told you that you had to use the existing Sony Exmor 24mp sensor (or any other currently existing image sensor) and design the perfect camera using it, what would happen? It doesn’t matter what you think “perfect” is; such a camera would produce excellent images that print really well to 19” on the long axis (because that’s where 24mp sensor state-of-the-art is). In other words, it’s not the image sensor that’s holding anything back.
A lot of people have poo-pooed Sigma’s recently announced bf camera. I don’t. Sigma is tackling exactly what I just asked in the last paragraph. Good on them. I see some design decisions in the bf that I agree with and many I don’t, but at least it’s not just another DSLR-like 24mp camera with as many features as we can cram into it. It’s a camera company challenging what “perfect” might actually be.
Likewise, I’ve found I tend to like Leica’s current modern UI (DLux-8, Q3, etc.). I’m not menu diving with these cameras, I’m pretty directly making the settings changes I need to make, and only those.
Need another example to understand my point? Look no further than the Nikon Z9. When I first was given a pre-release unit to take to Africa and test, what I found was that the Z9 was a really good camera, better than anything Nikon had done in mirrorless to date and mostly equalling or bettering what Canon and Sony were doing at the time. Then we got firmware 2.x. Then 3.x. Then 4.x. Then 5.x. Each time the already excellent camera became better. Why? Because the software was refined and added to. Amazingly, the Z9 still hasn't caught up to what Nikon knows how to do (content authentication, HEIF, Pixel shift shooting, etc.). Whether that means we get firmware 6.x or a Z9II with mostly software changes—I'd bet on an EVF upgrade, but not much else on the hardware side—doesn't particularly matter.
While many of you sit and wait for the latest and greatest hardware to finally arrive at your dealer, I'm personally waiting for the Japanese camera companies to get better at software. Way better. Because that's really what's needed to break out of the current volume doldrums and attract both new and upgrading customers.
