It would be safe to say that over the last five years, Intel’s been having a bit of a rough patch. AMD, on the other hand, has seen the complete reversal of its fortunes. The transition from the dreadful Bulldozer processors to Zen has seen the company re-enter the server and mobile markets after a long hiatus. And yet, despite all this, AMD’s R&D spending has been less than a quarter of Intel’s.
From 2017 thru 2019, Intel’s R&D spending stayed in the $13 to 13.5 billion range, while AMD’s rose from $1.8 to $2.9 billion. On one hand, this makes sense as Intel is a much larger company with its hands in a lot of pies such as client, enterprise processors, low-power, ultra-low power chips, NUCs, WiFi adapters, self-driving cars, and the newly formed Graphics and Visual Technologies Group. At the same time though, looking at Intel’s roadmap, one can’t help but wonder how so many things could have gone wrong.
The 10nm lineup is still not out in the volume segments, the Xe discrete graphics cards are more than a year away, 7nm based chips have been delayed by another 6 months and the company’s PR department is looking shadier and shadier with every new product.
This can only mean one bad decision over another because if higher funding led to better products, then Intel ought to be miles ahead of AMD and NVIDIA by now. That’s not quite the case though.
NVIDIA has spent less than both AMD and Intel on R&D, with a three-year total of $5.6 billion. In comparison, AMD invested $7.1 billion and Intel a massive $39.9 billion.
NVIDIA has been more and more invested in its Data Center business which is now nearly as large as the company’s gaming division at $1.75 billion. AMD has been spending on semi-custom solutions such as consoles, APUs, desktop and HEDT CPUs, servers, and embedded chips. Almost all these divisions have been flourishing.
Intel doesn’t really have anything to show for all that $39.9 billion spent on R&D. The only interesting product from Team Blue over the years has been Lakefield (and Alder Lake-S), and it’s hard to tell how it’ll do in the long run. The anticipation around the Xe GPUs has all but dissipated while Ice Lake and Tiger Lake are primarily the results of node shrinks with larger cache sizes. In a world where TSMC and Samsung are racing to 3nm nodes, this makes you wonder just where all that money went.