Intel Outspends Rivals In R&D: 28% More Than Nvidia, 156% More Than AMD 55
Intel shelled out $16.5 billion on R&D in 2024, outspending Nvidia by 28% and AMD by 156%, with much of the cash going into chip design, fabrication tech, and the upcoming Nova Lake architecture. "When you compare the R&D expenditures to the amount of revenue, though, the story takes on a very different look," notes PC Gamer. "Intel spent 31% of its net revenue, and 26% for AMD, but Nvidia and Samsung got by on just 10% and 4%, respectively." From the report: An analysis of research and development expenditure by TechInsights was reported by Korea JoongAng Daily, but you can get the numbers yourself by pulling up each company's 2024 financial results. For example, AMD declared that it spent $6.456 billion last year (pdf, page 1) on R&D, whereas Nvidia forked out $12.914 billion. It's worth noting that Nvidia's financial statements are numbered one year ahead of the actual period (FY 2026 is 2025 and so on).
Anyway, those figures pale in comparison to how much cash Intel burned through in 2024 to research and develop chip, fabrication technologies, software, and all kinds of tech stuffâ"a staggering $16.546 billion (pdf, page 25). That's 28% more than Nvidia and a frankly unbelievable 156% more than AMD. The nearest non-US semiconductor firm is Samsung Electronics, which spent a reported $9.5 billion on R&D. That would place third, comfortably ahead of AMD, and it strongly suggests that if you have your own foundries for making chips, you need to spend a lot of cash on finding ways to make better processors.
Anyway, those figures pale in comparison to how much cash Intel burned through in 2024 to research and develop chip, fabrication technologies, software, and all kinds of tech stuffâ"a staggering $16.546 billion (pdf, page 25). That's 28% more than Nvidia and a frankly unbelievable 156% more than AMD. The nearest non-US semiconductor firm is Samsung Electronics, which spent a reported $9.5 billion on R&D. That would place third, comfortably ahead of AMD, and it strongly suggests that if you have your own foundries for making chips, you need to spend a lot of cash on finding ways to make better processors.
Not apples to apples (Score:5, Insightful)
Nvidia is fabless. They can outsource the manufacturing of their designs to TSMC.
Actually, building chips is expensive. It needs lots of hardware. Nvidia designs chips. Intel wants to compete with TSMC, and they are WAY behind.
This (Score:5, Interesting)
Intel is trying to do the hardest thing and do everything, because historically vertical integration was their biggest technical advantage. AMD couldn't swing it (although that was admittedly helped along by Intel's anticompetitive acts) and wound up being better for letting it go. That's part of how they have the fastest processors around today. They also historically had to be better at design than Intel because they lacked the advantage of their then-cutting-edge processes. That design excellence combined with being on the superior process is currently unbeatable.
I genuinely wish Intel luck, because I really want them to advance technology like they used to. That relentless improvement pushed the industry forwards.
Re: (Score:3)
Re: (Score:3)
Re: (Score:2)
Because Taiwan has literal self destruction procedures for the case that they're invaded.
Re: (Score:2)
Re: (Score:2)
Well I think the overseas entities would continue operating but the destruction of their headquarters would probably be a major operational setback.
We're not watching a movie here come on, think, jeez.
Re: (Score:2)
If I as china, first think I would do would make those factories nonoperational. Those factories are needed by the US, to combat china. China has their own chip fabs.
Re: (Score:2)
I can see a set of geopolitical events where Intel ends up a clear winner and everyone else ends up decade behind
I can't. At "best" (for Intel, but not for anyone else) TSMC gets shut down in Taiwan using self-destruct protocols when China invades, and shut down here by ICE attacking their South Korean workers (which is already happening as we speak) and then Intel is an unclear winner, where what's unclear is how long it will take another fab to get fired up somewhere else with TSMC+ASML technology.
Intel has flubbed multiple process steps in a row, winding up with very poor yields on all of them. There's no obvious p
Re: (Score:2)
TSMC owns Hyundai? News to me.
Re: (Score:2)
Whoops. I did indeed conflate two stories there. I am the eggfaceman.
Though while we're here, how crazy is that shit?
Re: (Score:2)
Not very. It's what happens when you push for quick construction and none of the documentation for the custom equipment that needs to be built is in the local language, in this case English, is available. Getting custom documentation properly translated for industrial equipment takes time to ensure that everything is translated perfectly regarding construction and operation. So they brought in Koreans for the specialty construction and to train the people that would be operating the equipment. Since going
Re:Not apples to apples (Score:5, Interesting)
Intel is simultaneously a.) creating a new GPU ecosystem, b.) implementing their 1.8A process node, c.) developing the "nova lake" architecture d.) perfecting "backside" power delivery, e.) building several new fabs.
That's a lot of spending. Apparently, despite what Reddit thinks, they're not planning on going out of business.
Also, Intel is shipping their "Pro" GPUs now: the B50 is out. 16GB of VRAM and 170 INT8 TOPS at 70W for $350. Level1Techs has a review (yesterday) and it looks like a pretty good product. They're promising SR-IOV with VDI support for this GPU in Q4: so, remote VM desktops with hardware video acceleration at a non-"enterprise" price point, which would be a first.
Re: (Score:2)
They made some bets on realtime ray tracing maybe a little bit too soon. RTRT is indeed CPU bound, to some extent. It depends on the entire scene to be in memory. Pseudo ray tracing does not require that, so they approximate rays and draw not really ray traced objects. Sorry RTX, you're fake ray tracing, but it still looks good. Real time Radiosity... drools.
Re: Not apples to apples (Score:1)
Why does it matter if it's fake if it looks good? It's like mp3... If you can't hear the difference (and you can't, despite any claims you may make to the contrary - it's physically and biologically impossible), why does it matter if it's compressed?
Re: (Score:2)
Re: (Score:2)
True, but it seems like the article reached exactly the wrong conclusion. Intel does both chip design and manufacturing. AMD and Nvidia only the former. AMD spends close to what Intel does, as % revenue (31% vs. 26%). Nvidia spends less, but they've recently found a hose that money comes out of.
TSMC, a pure chip manufacturer, spends about 15%.
So the expensive thing seems to be chip design, not manufacturing.
Re: (Score:3)
The other player here is ASML which supplies TSMC's fab tech. How do they fit into this equation?
Re: (Score:3)
They fit in with Intel too, the difference is lead time. TSMC was an early customer of ASML, Intel was cautious and banked on DUV tech to push the boundaries back at the 13nm range where TSML was already placing orders for EUV machines.
ASML's High-NA EUV machines have an 18 month lead time, and can expect to take 6 months to install and commission in a facility. Intel is very much playing catch-up, the were responsible for 100% of orders from ASML last year, and ASML can only deliver low single digit units
Re: (Score:2)
ASML supplies the lithography machines to both TSMC and Intel. They seem to spend about 15% of revenue on R&D, but that might be a bit misleading since they spent 20 years working on EUV before making an actual product.
They're not the only other player though. The photomask is very important and has historically been the expertise of a few Japanese companies. Companies like Zeiss supply ASML with the optics. Advances in chip design software has enabled a lot of the progress, along with basic research on
Re: (Score:2)
Re:OK, cool, but what about your prior? (Score:5, Interesting)
Re: (Score:3)
It's not, that graph shows that Intel's R&D spend has been consistently rising, nearly triple what it was fifteen years ago. And that's consistent with the ever-rising cost of R&D on new process nodes. That never ends, you always need to be working on the next process node, and the costs to develop each node only ever go up.
It's also currently a wasted investment. Intel's been dumping enormous sums of money trying to develop new process nodes, but their 18A (1.8nm) node is suffering from unacceptabl
The joy of stock buy backs (Score:2)
Tsmc is heavily subsidized by the Taiwanese government. Intel gets the usual tax breaks but the money pumping into them from the government is a pretty recent thing and mostly response to covid era chip shortages.
Intel didn't have any serious competition for a long time so they took the money they were making and used it to pump their stock.
This is why stock buy backs are so
Re: (Score:2)
I disagree that Intel's problem was buybacks. I mean, they were dumb, yes, but Intel's problems wouldn't have been solved by that money going to more R&D. Their problems today are rooted in bad strategic bets going back to
Re: (Score:2)
I agree that buybacks are usually a stupid idea, and companies that train investors to expect them are setting themselves up for future pain
Why? I see this a lot but don't understand people's objection to stock buybacks. Are companies not supposed to ever return profits to their investors? If not, then why invest in that company? Sometimes a deep piggy bank is not in the best interest of a company -- they may end up just squandering it on something stupid like the metaverse or AI. I would much rather see a company return profit to its investors then watch it burn money while management flails and fails at something that is not in their core
Re: (Score:2)
Otherwise I think they give management too much room to play games at the expense of long-term firm health, both by opportunities for self-dealing and by training investors to expect them.
Re: (Score:3)
To start where I agree, there can be situations where a company has what amounts to free cashflow and no realistic opportunities to invest it. If that is the normal state of affairs, it should be issuing dividends.
Buybacks and dividends are the same from the company's standpoint. They are transferring $X to the shareholders with either method. But from the investor's POV, the buyback is better. If a company issues an investor a dividend they have to take that money right now. That means they have to recognize it as income even if they didn't want/need that income. By contrast, if the company increases the per-share price via a stock buyback then the investor can take that gain whenever they choose, rather than a
Re: (Score:1)
Why? I see this a lot but don't understand people's objection to stock buybacks. Are companies not supposed to ever return profits to their investors? If
No and go fuck your mother.
Re: (Score:2)
Sigh. I should set that up before october shouldn't I?
I've had it written up for over a year.
I kinda hoped we'd see something set up for him by actual family but there is literally no family, he apparently had friends but I've come to think they weren't as close as he thought or wished. Thus why we heard so much about his gross boomer work friends as if they were BFFs.
Re: (Score:2)
More specifically, the GP was speculating that R&D dropped while they were doing buybacks; my comment was refuting that.
Absolutely agree that the R&D spend has been wasted on consistently bad strategy.
Re: (Score:3)
Actually Intel 3 is not "bad" per se. It was late to the party and suffers from volume constraints. Intel has committed wafers to Sierra Forest, Granite Rapids, and Arrow Lake-U (which is really just Meteor Lake ported to Intel 3). Nobody knows the defect rate on Intel 3, but the parametric yields seem acceptable.
Had Intel 3 been ready as far back as 2023 and had the wafer volume been higher, it could have been a real winner. None of those things happened.
catching up is costly (Score:2)
Re:catching up is costly (Score:5, Insightful)
That brain drain started years and years ago, when they stopped developing their own fabrication processes in-house and started paying outside companies to develop the machines that make their chips. Those companies don't appear to have ironclad NDAs, so they then turn around and use the expertise they developed to develop fabrication machines for others as well.
Re: (Score:2)
Nobody develops all their fab tools in-house. Nobody.
Great? (Score:2)
Re:Great? (Score:5, Insightful)
The engineers likely want to make good chips. The suits and bean counters want to make maximum profit. Guess who wins.
Re: (Score:1)
+1 for an insightful comment. -1 for your dickhead personal attack. Net result = 0.
Re: (Score:2)
And you're still a festering cunt.
You allow people to live rent free in your head in *this economy*? Are you mad.
Re: (Score:2)
Yes and no. No chip is perfect. In fact any given chip family will have literally 100s of bugs in it (go look at the erratum publications from AMD or Intel). Fortunately the majority of them are usually of low impact or can be bypassed through clever software or microcode updates. Every so often you get a big one though, and Intel has been on the receiving end. With products in the market it stands to reason the bean counters will want to do anything possible to avoid a recall.
Not only are engineers not som
Who cares? Good Marketing Can Sell Anything (Score:2)
especially if they provide some cool shiny l sticker to slap on mediocre products.
Then why do Intel's products suck? (Score:4, Insightful)
Money spent is not a measure of research quality.
Re: (Score:2)
Money spent is not a measure of research quality.
You have to remember, Intel's moves are mostly about pleasing investors. As such, everything is a measure of dollars, whether spent or accumulated. So, this public statement is about showing investors that they're willing to throw their most important resource, money, at R&D, which to an investor sounds like future profits, even if they company has a history of throwing money at R&D and stiff managing to faceplant more often than not before the profit cycle begins on the output of that R&D.
"Gott
Re: (Score:2)
No argument. This number is basically a lie to please their investors. MS, incidentally, does the same. MS research does (or at least used to do) good research, and MS proper then ignores all of it and continues their crappy ways.
Re: (Score:2)
Exactly. By the same token, putting in 80 hours a week at work is meaningless if your work product is garbage.
Huawei? (Score:2)
> The nearest non-US semiconductor firm is Samsung Electronics,
I suspect huawei are up there somewhere, but it's difficult to compare. I do wonder where other Chinese companies are on this list. I suspect they're left out for some unrelated reason...
Apples and Kumquats (Score:1)
Intel makes a lot more than CPUs. Intel makes them itself.
AMD only makes CPUs. AMD pays someone else to make them.
nVidia only makes GPUs. nVidia pays someone else to make them.
So of course Intel spends a lot more on R&D. They are developing a much wider portfolio and they are making it all themselves.
Throwing good money after bad (Score:2)
AMD Treats its Customers Better than Intel (Score:2)