Worldwide Semiconductor Revenue Declined 11% in 2023, Intel Reclaims No. 1 Spot

[AleksandarK]

Worldwide semiconductor revenue in 2023 totaled $533 billion, a decrease of 11.1% from 2022, according to preliminary results by Gartner, Inc.

"While the cyclicality in the semiconductor industry was present again in 2023, the market suffered a difficult year with memory revenue recording one of its worst declines in history," said Alan Priestley, VP Analyst at Gartner. "The underperforming market also negatively impacted several semiconductor vendors. Only 9 of the top 25 semiconductor vendors posted revenue growth in 2023, with 10 experiencing double-digit declines."

The combined semiconductor revenue of the top 25 semiconductor vendors declined 14.1% in 2023, accounting for 74.4% of the market, down from 77.2% in 2022.



Intel Regained No. 1 Spot in 2023
Following the underperformance of memory vendors in 2023, the ranking of the top 10 semiconductor vendors changed year-over-year (see Table 1).

• Intel reclaimed the No.1 spot from Samsung, after two years in the No. 2 position. Intel's 2023 revenue totaled $48.7 billion while Samsung's revenue reached $39.9 billion.
• Nvidia's 2023 semiconductor revenue grew 56.4% to total $24 billion, propelling the company into the top five for the first time ever. This is due to its leading position in the artificial intelligence (AI) silicon market.
• STMicroelectronics moved up three slots to secure the No. 8 spot - the same position it held in 2019. Its revenue increased 7.7% in 2023, largely driven by a strong position in the automotive segment.

Memory Revenue Declined 37% in 2023
Revenue for memory products declined 37% in 2023, experiencing the biggest decline of all the segments in the semiconductor market. "Smartphones, PCs and servers, three of the largest segments for DRAM and NAND, faced weaker than expected demand and excess channel inventory, especially in the first half of 2023," said Joe Unsworth, VP Analyst at Gartner.

In 2023, DRAM revenue declined 38.5% to total $48.4 billion and NAND flash revenue decreased from 37.5% to $36.2 billion.

Nonmemory Revenue Declined 3% in 2023
Nonmemory revenue fared better and declined 3% in 2023. The market witnessed weaker demand and excess channel inventory negatively impacted the segment throughout the year.

"Unlike the memory vendors, most non-memory vendors experienced a relatively benign pricing environment in 2023," said Unsworth. "The demand for non-memory semiconductors for AI applications was the strongest growth driver, with the automotive sector (especially electric vehicles), defense and aerospace industries, also outperforming most other application segments."

Gartner clients can read more in "Market Share Analysis: Semiconductors, Worldwide, 2023 (Preliminary)."

View at TechPowerUp Main Site | Source

 

[las]

Intel 1st and Nvidia as the big winner is not really surprising.

Intel will probably start to pull more and more ahead over the next years, with 20A and 18A almost ready + Open for business.

Maybe AMD will use Intel fab in a few years

 

[Daven]

Intel 1st and Nvidia as the big winner is not really surprising.

Intel will probably start to pull more and more ahead over the next years, with 20A and 18A almost ready + Open for business.

Maybe AMD will use Intel fab in a few years

AMD, Apple, Qualcomm, Nvidia, etc will all switch to Intel once Intel pivots from being a chip designer to a chip maker. Right now, none of those companies will seriously use IFS as Intel would use the detailed info given to a chip manufacturer against said companies. Just imagine Intel knowing AMD yields, volumes, transistor layout, release dates, etc. Then imagine Intel having control over AMD yields, volumes, lead times, etc.

Don’t be fooled as long as Intel makes its own chips its too risky for direct competitors to use IFS as Intel would favor the manufacturing of its own chips while using less quality production controls for its competitors. Its a tricky situation since Intel has so much fab capacity but lower performing products but as long as TSMC (and partly Samsung) can keep up with demand then IFS will not grow.

 

[las]

AMD, Apple, Qualcomm, Nvidia, etc will all switch to Intel once Intel pivots from being a chip designer to a chip maker. Right now, none of those companies will seriously use IFS as Intel would use the detailed info given to a chip manufacturer against said companies. Just imagine Intel knowing AMD yields, volumes, transistor layout, release dates, etc. Then imagine Intel having control over AMD yields, volumes, lead times, etc.

Don’t be fooled as long as Intel makes its own chips its too risky for direct competitors to use IFS as Intel would favor the manufacturing of its own chips while using less quality production controls for its competitors. Its a tricky situation since Intel has so much fab capacity but lower performing products but as long as TSMC (and partly Samsung) can keep up with demand then IFS will not grow.

And this is also going to be the "downfall" of AMD as they will be forced into using TSMC or even Samsung, meaning Inferior nodes, not now (Samsung is inferior tho), but in 1-2 maybe 3 years when Intel regain leadership. The moment that Intel launches 18A before TSMC 2nm. Intel 20A is more or less equal to TSMC 3nm.

AMD had some good runs back in the days as well but Intel always came back eventually. Ryzen 3000 and forward has been great. 1000 and 2000 series not so much.

I will probably be buying 8800X3D/9800X3D later this year tho, unless Intel impress with Arrow Lake and 20A, we will see.

There's pros and cons of being fabless, AMD saw the pro during Intels sleeping CEO days while being stuck at 14nm, tables are turning after Pat Gelsinger took over the chair and replaced people + all those brand new fabs are starting to be ready for mass production.

Intel took 1st spot even without all their fabs running at full blast.

Intel Fab 52 + 62, and 27 eventually, is going to be very important

Could see Apple start using 18A as soon as possible, they wanted chip production out of Asia for years

If Trump winning, it could hurt TSMC alot going forward. TSMC is where they are today because of Apple money.

 

[Daven]

And this is also going to be the "downfall" of AMD as they will be forced into using TSMC or even Samsung, meaning Inferior nodes, not now (Samsung is inferior tho), but in 1-2 maybe 3 years when Intel regain leadership. The moment that Intel launches 18A before TSMC 2nm. Intel 20A is more or less equal to TSMC 3nm.

AMD had some good runs back in the days as well but Intel always came back eventually. Ryzen 3000 and forward has been great. 1000 and 2000 series not so much.

I will probably be buying 8800X3D/9800X3D later this year tho, unless Intel impress with Arrow Lake and 20A, we will see.

There's pros and cons of being fabless, AMD saw the pro during Intels sleeping CEO days while being stuck at 14nm, tables are turning after Pat Gelsinger took over the chair and replaced people + all those brand new fabs are starting to be ready for mass production.

Intel took 1st spot even without all their fabs running at full blast.

Intel Fab 52 + 62, and 27 eventually, is going to be very important

Unless Intel pivots to a new business model, I do not believe they will be around in three years. The difference between now and in the past are the number of players. Before it was just Intel. No one else was a serious chip designer before 2007. AMD made a play with K8 then blew it. Then the iphone happened, then ARM got serious, then the GPU exploded onto the scene. The situation is similar to when Balmer at MS tried to takeover the mobile market. They were laughably too late so Balmer was canned and they pivoted to services through Big Data. Intel is also laughably too late to the GPU and high efficiency (ARM) space. Intel must pivot from a chip designer to a chip maker before its again too late.

I know this view is extremely unpopular especially on a DIY tech site but you live with the market you have not the market you want.

 

[las]

Unless Intel pivots to a new business model, I do not believe they will be around in three years. The difference between now and in the past are the number of players. Before it was just Intel. No one else was a serious chip designer before 2007. AMD made a play with K8 then blew it. Then the iphone happened, then ARM got serious, then the GPU exploded onto the scene. The situation is similar to when Balmer at MS tried to takeover the mobile market. They were laughably too late so Balmer was canned and they pivoted to services through Big Data. Intel is also laughably too late to the GPU and high efficiency (ARM) space. Intel must pivot from a chip designer to a chip maker before its again too late.

I know this view is extremely unpopular especially on a DIY tech site but you live with the market you have not the market you want.

Intel is not really late to GPUs, they have huge iGPU marketshare and while 1st gen Arc was wonky, it was a decent first try. Drivers improved performance big time since release. With more and more games leaving DX9/10 completely, Intel GPUs are only going to get more relevant. They perform great in DX12 for the money. When Intel regain process leadership and can output GPUs here, AMD will be in big trouble in the low to mid-end market I think. Intel has mostly focus on DX12 and nothing else will matter much going forward so their GPUs are only going to get more competitive over time.

Most PC gamers are buying low to mid-end and this is exactly Intels focus for now. Also AMDs which are not really selling much high-end GPUs anyway.

AMD will be gone long before Intel thats for sure. Intels revenue is like 3 times AMDs even in times where AMD is going great and Intel spends tons of money on new fabs. Tables are turning fast. I don't expect any of them to vanish any time soon tho.

 

[R0H1T]

Arc 1st gen was wonky but decent for a first try.

Second (third?) try but yes semi decent nonetheless.

 

[Daven]

Intel is not really late to GPUs, they have huge iGPU marketshare and the Arc 1st gen was wonky but decent for a first try. With more and more games leaving DX9/10 completely, Intel GPUs are only going to get more relevant. When Intel regain process leadership and can output GPUs here, AMD will be in big trouble in the low to mid-end market I think.

AMD will be gone long before Intel thats for sure. Intels revenue is like 3 times AMDs even in times where AMD is going great and Intel spends tons of money on new fabs. Tables are turning fast. I don't expect any of them to vanish any time soon tho.

The tables are turning fast but in the opposite way. Intel revenue is just double AMDs and dropping fast from a peak of $21B in a quarter. AMD market cap is $256B, Intel’s is $190B and Nvidia’s is $1.4T.

No one is choosing Intel’s IGP. It comes with all CPUs. People are buying complete systems based on price. Systems without discrete graphics cards are cheaper.

But the reason AMD and Nvidia are accelerating is not because of what GPU is in client computers but Big Data computers. Intel is dead in this space. Ponte Vecchio is not successful. And just like the Itanium and Xeon Phi before, Intel’s compute GPUs will get cancelled.

Intel’s biggest achievement, Aurora, is way late and when initial performance numbers were finally released, only half of the potential compute capabilities were functional. Not a good sign.

The money is in compute GPUs and chip manufacturing. Last quarter TSMC made around $20B and Nvidia made $18B. Intel made $13B.

 

[R0H1T]

only half of the potential compute capabilities were functional.

Half? Surely DoD or DoE(?) would've asked for a refund then!

 

[las]

The tables are turning fast but in the opposite way. Intel revenue is just double AMDs and dropping fast from a peak of $21B in a quarter. AMD market cap is $256B, Intel’s is $190B and Nvidia’s is $1.4T.

No one is choosing Intel’s IGP. It comes with all CPUs. People are buying complete systems based on price. Systems without discrete graphics cards are cheaper.

But the reason AMD and Nvidia are accelerating is not because of what GPU is in client computers but Big Data computers. Intel is dead in this space. Ponte Vecchio is not successful. And just like the Itanium and Xeon Phi before, Intel’s compute GPUs will get cancelled.

Intel’s biggest achievement, Aurora, is way late and when initial performance numbers were finally released, only half of the potential compute capabilities were functional. Not a good sign.

Intels revenue are more like triple AMDs, AMD stock went up because of AI dreams but Nvidia is the AI king in reality, for now AMD has little to nothing to say in this market, meanwhile Nvidia earns big money.

Intel iGPU is in 98% of enterprise laptops, I work with B2B sales and AMD barely moves any machines in this segment. Too much fragmentation, AMDs mobile segment is a big mess really with mixed architectures and designs + tons of issues with networking and display drivers in combination with port reps and monitors. Many companies tried going AMD because of price but went screaming back to Intel on the next upgrade cycle.

Intel still owns most Enterprise segment too, AMD is not even close.

For CPU desktop and gaming market, AMD does well but Intel still sits at most marketshare, around 65-70%

AMD have a hard time competing with cheap Intel CPUs because Intel has cheaper motherboards and don't need expensive memory to perform well. Stuff like 12400,13400 etc, are better value than AMD for mid-end gaming. Pushed like crazy in the OEM sector too.

This post was about Intel being 1st and you act like they are struggling to stay alive? They had hard times 2018-2019ish till 2022 or so but they are slowly regaining and will soon regain process leadership as well. If 20A is on point they will start their roll and AMD will be pushed back to focussing on being "great value" again. AMD had some good runs earlier too but Intel always came back.

 

[Bwaze]

Isn't everyone and their grandma buying AI acceleration hardware? How can there be a decline then?

 

[las]

Second (third?) try but yes semi decent nonetheless.

Name those 3 tries?

Arc as we know it was a 2022 and 2023 release, all using same arch in difference scale

Isn't everyone and their grandma buying AI acceleration hardware? How can there be a decline then?

Zero decline, Nvidia is pushing mad numbers here and the biggest problem for Nvidia is actually making enough chips. They scaled back on gaming to raise enterprise and AI chip output. For now. They can adjust when needed.

 

[marios15]

And this is also going to be the "downfall" of AMD as they will be forced into using TSMC or even Samsung, meaning Inferior nodes, not now (Samsung is inferior tho), but in 1-2 maybe 3 years when Intel regain leadership. The moment that Intel launches 18A before TSMC 2nm. Intel 20A is more or less equal to TSMC 3nm.

AMD had some good runs back in the days as well but Intel always came back eventually. Ryzen 3000 and forward has been great. 1000 and 2000 series not so much.

I will probably be buying 8800X3D/9800X3D later this year tho, unless Intel impress with Arrow Lake and 20A, we will see.

There's pros and cons of being fabless, AMD saw the pro during Intels sleeping CEO days while being stuck at 14nm, tables are turning after Pat Gelsinger took over the chair and replaced people + all those brand new fabs are starting to be ready for mass production.

Intel took 1st spot even without all their fabs running at full blast.

Intel Fab 52 + 62, and 27 eventually, is going to be very important

Could see Apple start using 18A as soon as possible, they wanted chip production out of Asia for years

If Trump winning, it could hurt TSMC alot going forward. TSMC is where they are today because of Apple money.

Intel is 3-6 years behind TSMC on production of bleeding edge nodes

Just because they rebranded their 5 year 10nm delays to "Intel 7" does not mean that they will now catch up with TSMC

They were always advancing their CPUs to their latest fab nodes even with very low yields and absorbing that cost which worked well when everyone else also had their own fabs

Reached capacity? Build more fabs

The last 5 years are the slow downfall of a giant

 

[Bwaze]

Zero decline...

"Worldwide Semiconductor Revenue Declined 11% in 2023"

 

[las]

Intel is 3-6 years behind TSMC on production of bleeding edge nodes

Just because they rebranded their 5 year 10nm delays to "Intel 7" does not mean that they will now catch up with TSMC

They were always advancing their CPUs to their latest fab nodes even with very low yields and absorbing that cost which worked well when everyone else also had their own fabs

Reached capacity? Build more fabs

The last 5 years are the slow downfall of a giant

Intel 4 is already ready and producing - 20A ready later this year.

They are not 3-6 years behind TSMC haha. TSMC is stuck at 3nm for now as well, and Intel 20A is at least on par with TSMC 3nm.

Once again, Intel is 1st in the semiconductor revenue. No problems at all. They are gaining not loosing.

"Worldwide Semiconductor Revenue Declined 11% in 2023"

Talking about AI GPUs and look at Nvidia's 56.4% increase maybe

 

[ThrashZone]

Hi,
Intel 1st in e-waste
Hell they just put out a dual core lol

 

[Legacy-ZA]

Shhh, don't say that.

Isn't everyone and their grandma buying AI acceleration hardware? How can there be a decline then?

Hype/marketing, the hype has been oversold.

 

[Easo]

Intel gone in ~3 years?
People, what the hell are you even talking about???

 

[marios15]

Intel 4 is already ready and producing - 20A ready later this year.

They are not 3-6 years behind TSMC haha. TSMC is stuck at 3nm for now as well, and Intel 20A is at least on par with TSMC 3nm.

Once again, Intel is 1st in the semiconductor revenue. No problems at all. They are gaining not loosing.


Talking about AI GPUs and look at Nvidia's 56.4% increase maybe

Intel 10nm originally planned for 2016, and 3 plus(+++) later they had demo laptops with cannon lake in 2018
Intel 7nm was planned for 2-3years later

The first production chips appeared in very late 2019 with Ice Lake running on 10nm for laptops and Xeons and finally released Tiger lake, the first desktop 10nm parts in late 2020

Meanwhile TSMC had 7nm chips since 2018 (equal to Intel's 10nm)

 

[R0H1T]

Name those 3 tries?

I said this is likely the third so 1-2 before this. I'd count Larrabee as first & I guess Phi as the second?

 

[las]

Intel 10nm originally planned for 2016, and 3 plus(+++) later they had demo laptops with cannon lake in 2018
Intel 7nm was planned for 2-3years later

The first production chips appeared in very late 2019 with Ice Lake running on 10nm for laptops and Xeons and finally released Tiger lake, the first desktop 10nm parts in late 2020

Meanwhile TSMC had 7nm chips since 2018 (equal to Intel's 10nm)

Who cares what Intel planned back in the sleeping CEO days. Pat Gelsinger took over in 2021 for a reason and things are turning fast with him in the seat.

Again, whos 1st in this arcticle?

Ryzen 3000 dropped in summer 2019 on TSMC 7nm, and this is a big part of why AMD went from mediocre to good, before that they used the terrible GloFo 12nm process which gimped clockspeeds on Ryzen 1000 and 2000 series which posed no threat to Intel outside of being somewhat useful for some people, instead of absolutely terrible AMD CPUs before them.

AMD relies 100% on TSMC and when TSMC drops behind Intel, which they will eventually, AMD will have a hard time. They need to be using TSMC most advanced nodes, which is what Apple has priority on and this is going to be very expensive for AMD.

Intel has been able to compete with AMD just fine while using subpar nodes. Worst years are behind now.

Just like Nvidia beat Radeon 6000 series while using Samsung 8nm which is closer to TSMC 10nm in reality. AMD was not even able to win, while using a more advanced process node.

If AMD was forced to use an inferior process, they would not be able to compete in the high-end. We saw that with GloFo 12nm. However both Intel and Nvidia was able to do it.

 

[R0H1T]

So you're saying AMD "won" because of TSMC's 7nm process but Intel didn't, because they had a 1.5-2.5 node lead a decade back

 

[las]

So you're saying AMD "won" because of TSMC's 7nm process but Intel didn't, because they had a 1.5-2.5 node lead a decade back

AMD did not "win" anything. They went from utter crap to decent and even good. AMD had a great run with Ryzen 3000 and 5000 series, when AMD used TSMC 7nm vs Intel 14nm, yeah, like I said tho, worst years are behind for Intel. They are making chips on Intel 4 node now (mobile) and 20A later this year (desktop and enterprise) with 18A ready in 2025 or so.

A decade back tho? More like 5 years

I will probably be using a Ryzen 8800X3D/9800X3D later this year, but I expect to come crawling back to Intel for the next upgrade. I expect Intel to have process lead by 2026 or so. Their fabs will be open for business for everyone too. Intel will remain 1st in the semiconductor market for sure.

 

[Daven]

Intels revenue are more like triple AMDs, AMD stock went up because of AI dreams but Nvidia is the AI king in reality, for now AMD has little to nothing to say in this market, meanwhile Nvidia earns big money.

If you dont know the revenue figures dont post about them.

Intel Q3 2023 $14B
AMD Q3 2023 $6.1B

Again businesses are not choosing iGPUs. They are choosing complete systems. Intel is still in the majority of these systems. BoA IT buyers are not saying ‘man those Intel iGPUs are so good. We need those.’ They are saying ‘we can get 1000 dell laptops for x amount over Lenovo.’ Those laptops happen to have Intel in them. Get real.

And again the money is in Big Data hardware. All Intel’s market share is from being the leader for so long. Now they are declining.

 

[R0H1T]

A decade back tho? More like 5 years

I was talking about Intel, they always lead AMD with at least a node or so till Zen dropped so their (better)performance wasn't down to superior nodes? The last *dozer chips were out around the same time or slightly earlier & on 28nm, Intel already had 14nm chips since 2014 IIRC.