Nvidia Is Still Making Billions In Q4 2023 Despite a Giant Drop In PC Demand (theverge.com) 22
In its fourth quarter and full-year earnings report yesterday, Nvidia reported $6.05 billion in revenue for Q4 of its fiscal 2023 and $26.92 billion for the full year. That's "almost identical to last year, though profit was down 55 percent," notes The Verge.
"Remember: in 2021, $5 billion in revenue a quarter was a new Nvidia record. Now it's the status quo: the company says it's expecting to see $6.5 billion next quarter, too." From the report: Nvidia's data center and automotive businesses were actually up this quarter, with record revenue for automotive of $294 million; the dip was largely in Nvidia's graphics business, particularly gaming, which were each down 46 percent. That gaming decline includes "lower shipments of SOCs for game consoles," which is code for "Nintendo isn't selling as many Switches anymore" -- it's the only game console that uses an Nvidia chip. Like other chipmakers, Nvidia is shipping fewer GPUs to retailers and partners instead of slashing prices. The polite phrase is "lower sell-in to partners to help align channel inventory levels with current demand expectations." Nvidia also blamed disruptions in China due to covid and other issues.
Every PC maker is reporting that demand for computers has tanked this past quarter, with research firm Gartner calling the 28.5 percent dip in shipments "the largest quarterly shipment decline since Gartner began tracking the PC market in the mid-1990s." That was on top of the slump companies like Nvidia had already seen. And while AMD seemed optimistic this quarter that the slump won't last for long, even it suggested that client processor and gaming revenue would continue to go down in the first half of the calendar year.
"Remember: in 2021, $5 billion in revenue a quarter was a new Nvidia record. Now it's the status quo: the company says it's expecting to see $6.5 billion next quarter, too." From the report: Nvidia's data center and automotive businesses were actually up this quarter, with record revenue for automotive of $294 million; the dip was largely in Nvidia's graphics business, particularly gaming, which were each down 46 percent. That gaming decline includes "lower shipments of SOCs for game consoles," which is code for "Nintendo isn't selling as many Switches anymore" -- it's the only game console that uses an Nvidia chip. Like other chipmakers, Nvidia is shipping fewer GPUs to retailers and partners instead of slashing prices. The polite phrase is "lower sell-in to partners to help align channel inventory levels with current demand expectations." Nvidia also blamed disruptions in China due to covid and other issues.
Every PC maker is reporting that demand for computers has tanked this past quarter, with research firm Gartner calling the 28.5 percent dip in shipments "the largest quarterly shipment decline since Gartner began tracking the PC market in the mid-1990s." That was on top of the slump companies like Nvidia had already seen. And while AMD seemed optimistic this quarter that the slump won't last for long, even it suggested that client processor and gaming revenue would continue to go down in the first half of the calendar year.
They've got a defacto monopoly (Score:5, Insightful)
I think AMD's problem is board partners overvolting and underbuilding GPUs, but then AMD needs to account for that reality in their designs. And they need to stop letting Microsoft Windows Update stomp all over their drivers. These are basic things that make buying their cards a like walking into a landmine field. It's 2023, I shouldn't have to ask "is this video card going to be crash city?"
Re: (Score:2)
Cross your fingers if you want all those low-end chipset drivers to work together. It's a total mess.
This hasn't been a real thing since AMD brought out their own chipsets for Athlon, and has had them for every platform since. AMD chipsets are great, every bit as solid as the Intel ones. The drivers have not always been stellar, but this is not a problem on Linux anyway.
However, I still stick with nvidia GPUs because the software stack remains more mature. It is unfortunate that the OSS driver is so bad, and that you therefore need to use the corporate driver even on Linux if you want full functionality, b
Re: (Score:2)
What's it like still living in 2006? FYI, Windows Mobile, Blackberry, and PalmOS are all garbage that won't exist in 4 years. Start learning ObjectiveC and the Android SDK. And in late 2019 if you ever get there, start stockpiling toilet paper, hand sanitizer, and N95 surgical masks - you'll make a killing in about 6 months in high-margin aftermarket resale.
FYI, here in 2023, the problems you speak of with AMD haven't been problems for quite some time, since they started making their own chipsets that pa
Re: (Score:1)
Re: They've got a defacto monopoly (Score:1)
Well, duh... (Score:4, Interesting)
OEMs had an absolutely incredible mid-2020 through early-2021. Lots of pent-up demand, as well as manufactured demand, as well as a general public with stimulus checks and an inability to spend them on travel, means that the profits from that time span were going to come at the expense of demand over the next 2-3 years. Windows 11 being received tepidly-at-best isn't exactly the shot in the arm OEMs were hoping it'd be, either.
nVidia, on the other hand, had incredibly challenging timing on their 3000-series GPUs. It was announced to have considerably better performance than the 2000-series, at lower prices than the 2000-series. Add in the cryptocurrency boom *and* the supply chain woes *and* neural network development heating up...and suddenly there was a massive demand that nVidia simply couldn't supply.
Since nVidia seems able to generally meet supply now, a lot of that pent-up demand is paying up. Some of this is a good thing - gamers and video editors and CAD folks can finally replace their aging 900-series and 1000-series cards. Other parts of it have to do with nVidia adjusting their pricing to what the market clearly will-bear. It used to be possible to get an entry level GPU for about $150, maybe $250 for a low-midrange card. Now, the cheapest 4000-series GPU I see available at retail is $799. Sure, that might not exactly be 'entry level', but it was *definitely* possible to get a 2060 for under $500 in 2019. We can't blame Vladmir Putin or Xi Jinping for the fact that the retail price of a GTX-1650 is higher in 2023 than it was in 2019.
I don't necessarily *fault* nVidia for being opportunistic here...but we can't quite act like their strategy is unique here - capitalizing on known, pent-up demand isn't exactly rocket science, and if they were still pricing their cards at pre-pandemic levels, even giving them a bit of wiggle room for inflation, these charts would look very different.
The real question here is how long nVidia will keep these much-higher prices in place. Yes, it's way better than the scalpers, and yes, I do understand that nVidia can charge what they want to charge, but these numbers are unlikely to be sustainable through 2025, and the real question is whether they'll end up in exactly the same place as the OEMs: enjoying a massive amount of sales as a result of pent-up demand to the point where they're willing to trade future success to keep it going. Someone who pays $399 for a 4070 may well be willing to pay $449 for a 5070 in two years, but a person who pays $1,079 for a 4070 is very unlikely to buy a 5070 for quite some time.
Re: (Score:2)
Add in the cryptocurrency boom *and* the supply chain woes *and* neural network development heating up...and suddenly there was a massive demand that nVidia simply couldn't supply.
Proof-of-work cryptocurrency mining on GPUs is still a thing for certain coins. These days though, it's basically just not profitable if you actually have to pay for your electricity.
small wonder, given their time machine (Score:1)
Isn't Q4 July-September in the USA?
Re: (Score:2)
Some companies have their fiscal year offset to the calendar year because reasons/. Nvidia's fiscal Q4 ends in January.
Low demand is because your products are overpriced (Score:4, Interesting)
Five years ago the flagship Nvidia GPU, the 1080 ti, cost around $700. Today's flagship 4090, which has a smaller die than the 1080 ti, is priced at $1700. Mid-range cards have experienced similar price hikes.
There is nothing to justify that other than greed and an effort to maintain the huge margins they enjoyed during the crypto boom. I've always been a PC gamer but it's getting hard to justify when I could buy a console for a third of the cost of a GPU alone.
I only hope Intel continues development of their Arc GPUs and lights a fire under the Nvidia/AMD duopoly.
Re: (Score:2)
I get the feeling we're going to see the same thing with car prices. There is a more competition in the car market, but the price spike wasn't so extreme, and it's easier to hide the cost from Joe Average through financing.
Re: (Score:2)
Nvidia's profit margin is lower now than it was 5 years ago. Spring 2018 was also just before the release of a new generation, rather than just after.
Re: Low demand is because your products are overpr (Score:2)
There's absolutely no doubt that Nvidia GPUs are expensive, especially compared historically. Are they overpriced? It depends on one's viewpoint. Consumers are paying more, but the market is accepting the price increase because the high-end is selling out.
One other thing to keep in mind is that only part of the GPU profit goes to Nvidia. The board manufacturers and the retailers also add on their own profit margins. It's not at all clear which of the three are taking the biggest cut.
Re: (Score:2)
> It's not at all clear which of the three are taking the biggest cut.
I don't know, but EVGA getting out of GPU business recently, suggests that it might be Nvidia that is taking the biggest cut.
Re: (Score:1)
Moore's Law is dead. For the first time in 40 years transistors have gone up in price.
Re: (Score:2)
the 1080 ti, cost around $700. Today's flagship 4090
There were cards better and faster than the 1080 ti back then, the difference was they were reserved for workstations. If you're going to compare like product models than do so. Unfortunately the 4090 had no comparison since the high RAM workstation cards had artificially inflated prices, and in that respect the 4090 is quite a bit cheaper than the Quadro P6000 ;-)
On the upside it still reinforces your statement, yeah the high end GPUs have gotten quite a bit more expensive. The 4080 ti is easily $300 more
Re: (Score:2)
This is understated on Slashdot (Score:2)
For every article we hear of doom and gloom it's worth remembering that the finance world is freaking out profits not being as high as the year prior. Nothing more. Also worth noting is that the year prior needed a big arse asterisk applied to it thanks to COVID. We keep talking like it's the end of days, so it's good to see an article reminding everyone that they are still making f-you levels of money.
Just in the past 3 days alone I've heard the following:
"Facebook can't make money": - they can, even after
Of course they are. (Score:2)
Yawn. Go away NVidia. (Score:2)
Just making everything bigger, heavier, and more expensive every year isn't going to cut it, NVidia. Come up with something new or make something functional, affordable, and compatible. That's what AMD is