siliconbruh999
Well-known member
source:
SemiConductor R&D Spend by Company
- Thread starter siliconbruh999
- Start date
source:
Interesting to designate Qualcomm as a foundry... 
These lists are also interesting and subject to debate. (Looks like the source for this is Tech Insights.) Assuming Chinese firms are left off due to lack of verifiable data? Would have expected SMIC, YMTC, etc. to at least break $1B. Curious if companies that do in-house design (Apple, Meta, Amazon, Alphabet, etc.) don't make the cut or are just ignored for this list?
Still, interesting to see the range of R&D expenses, from SK Hynix (6.9%) and TSMC (7.1%) to Broadcom (30.3%) and Intel (33.6%). Really runs the gamut.
Also, looks like SK Hynix may be on track to overtake Intel in revenue this year? Wild times indeed.
These lists are also interesting and subject to debate. (Looks like the source for this is Tech Insights.) Assuming Chinese firms are left off due to lack of verifiable data? Would have expected SMIC, YMTC, etc. to at least break $1B. Curious if companies that do in-house design (Apple, Meta, Amazon, Alphabet, etc.) don't make the cut or are just ignored for this list?
Still, interesting to see the range of R&D expenses, from SK Hynix (6.9%) and TSMC (7.1%) to Broadcom (30.3%) and Intel (33.6%). Really runs the gamut.
Also, looks like SK Hynix may be on track to overtake Intel in revenue this year? Wild times indeed.
Good data -- The Intel spend shouldn't be surprising:
- They are an IDM (i.e. Their R&D % should be similar to Nvidia + TSMC)
- They had to spend extra to finally get past 10nm
- They brought some cutting edge technology online no one else has
- They have a full stack of products they're trying to sell (i.e. Nvidia isn't doing CPUs for the full market, AMD has one core architecture and GPUs are on life support)
- This is before they wound down some duplicate products like Gaudi
- They are an IDM (i.e. Their R&D % should be similar to Nvidia + TSMC)
- They had to spend extra to finally get past 10nm
- They brought some cutting edge technology online no one else has
- They have a full stack of products they're trying to sell (i.e. Nvidia isn't doing CPUs for the full market, AMD has one core architecture and GPUs are on life support)
- This is before they wound down some duplicate products like Gaudi
hist78
Well-known member
One thing that does surprise me is why Intel's leaders, for years, have not recognized that such huge R&D expenses, so high relative to revenue and with such poor effectiveness are not sustainable
MKWVentures
Moderator
Intel spends more on RnD and has always spent more on RnD than competitors
Intel spending in 2023 was more than TSMC and nvidia combined. despite being less than half the size.
Interestingly, only one company in the list saw its revenue drop despite all the R&D spending.
Intel did not spend too little on R&D. The just didnt develop and deliver
hist78
Well-known member
I put together this table using net profit to measure Intel’s R&D effectiveness. In 2023, Intel had to spend $9.50 on R&D to earn $1 in net profit, compared with $2.19 in 2022. Since Intel reported a net loss in 2024, R&D effectiveness for 2024 isn’t meaningful at all.
Source: Googel search.
MKWVentures
Moderator
In past 4 years intel increased R&D, Increased Capex, and dramatically lowered revenue. and the stock crashed while others doubled. That is unprecedented mismanagement... Luckily LBT will fix the disaster created by properly managing the company and focusing on success, not hype.....
No no no, what you see is exactly the case of a very "manicured" stock market diva company starting to having to till earth and do actual work.
Lowering profitability to the level of an "average semi" is very sane move, but it was received very negatively because of the prior prestige, and image are gone.
I agree this is terrible, but some context/thoughts:
- According to Google, AMD is still spending 17-25% of revenue for R&D from 2020-2024; not really lower than Intel despite not needing to R&D Foundry
- It would be interesting to see how much R&D at Intel were *completely* thrown away during this timeframe -- i.e. Intel's mobile, AI, GPU products (Larrabee)
- Intel's biggest financial liability still seems to be competition from AMD rather than missing new markets; if Intel simply kept server market revenue at 90+% 2022 and 2023 would have looked "OK", though 2024 would still be a loss.
I'm surprised that net profit became so strong in 2018-2021, despite having increasingly weaker offerings in server and desktop relative to competition. That gives an idea that even if Intel came out with very strong products tomorrow, it could still take at least 4 years for their profits to start swinging back..
MKWVentures
Moderator
If you buy silicon from TSMC before the product is ship released, is that R&D according to GAAP?
hist78
Well-known member
Those should be counted as inventory, not R&D, although they may not show up on the books immediately depending on the arrangement.
hist78
Well-known member
Intel’s R&D spending must cover both product development and manufacturing (Intel Foundry), whereas competitors such as TSMC, AMD, Nvidia, Broadcom, Qualcomm, MediaTek, Apple, Google, Meta, Microsoft, and Amazon can focus their R&D on either products or manufacturing.
Beyond efficiency concerns, Intel may simply be outgunned by a large number of highly focused rivals. Its R&D is spread too thin across many fronts to sustain the end-to-end IDM model.
MKWVentures
Moderator
that would be true IF it was shippable, but you do stepping and revisions and test chips that are not shippable and those are counted as R&D. This is why there are terms used like risk production for accounting.
On your other note, Intel spends more than Nvidia and TSMC combined (until Nvidia went up). The issue is that Intel spends a ton and is not a leader in products or processes. If you saw the project list and outcomes.... it would be clear that money was being wasted
I think they will become a leader in products, Processes are more difficult when you are not the volume leader.
[Human Thought]
I would think the rule is basically:
- Internal Samples, ES/QS = R&D
- Production/intent to sell = Not R&D
[/End Human Thought]
(Intel also accures "R&D" for depreciation of development fabs per Grok's interpretation of their 10-K)
hist78
Well-known member
With $53 billion in revenue in 2024, Intel is a leader in the semiconductor market, although it is facing challenges.
If costs are incurred during the development process and risk production, I believe there is flexibility for a company to decide whether to classify them as production costs, R&D expenses, or split them between both.
Under current IRS regulations, proposed and passed during the first Trump administration and effective since 2022, companies are required to amortize domestic R&D expenses over 5 years and foreign R&D expenses over 15 years. This may influence how companies approach cost allocation.
For example, Nvidia has generated substantial profits in recent years, and it may tend to allocate more costs to the production expense category in order to reduce its tax burden as much as possible.
Other companies, however, may take the opposite approach, allocating more expenses to the R&D category (spreading costs over 5 years) so their financial results appear more profitable, or their losses appear smaller.
But as you know, this year's R&D pays off much later (if successful, or not). Products in volume today are based on design R&D / product decisions made 5 years ago and process decisions and investments 8 years ago.
hist78
Well-known member
Yes, I totally agree. We should take a closer look at Intel’s 15 years of R&D expenses shown in my table to find broader information rather than just focusing on each single year.
We need to find a way to compare a particular year’s net profit with the R&D spending from previous years, such as the past 3 or 5 years. Do you have any suggestions?
hist78
Well-known member
In some cases, even shippable wafers or chips may never appear directly in a company’s accounting books.
For example, suppose an autonomous robot company (Trust Me Robots) designs an AI chip and asks its OEM assembly partner (Do It Corp.) to place chip orders with a foundry (Be Happy Semi Manufacturing Corp.).
When the finished robots (with the AI chips installed) arrive at Trust Me’s warehouse, Trust Me pays Do It Corp. for the full cost of production, including the AI chip expenses.
Meanwhile, Be Happy Semi receives payment from Do It Corp. for manufacturing the AI chips, according to their contract.
By using this approach, Trust Me Robots can better manage its inventory levels and the impact on its financial statements.