Whitebox Switching: Follow the R&D Budget

A few weeks ago HP announced that they’d start selling branded whitebox (brite-box) switches, and as expected the industry press was immediately full of opinions. As always, it makes sense to follow the money (or, in this case, the R&D budget) to understand what’s going on behind the scenes.

I did a very simple exercise: I collected the latest quarterly reports from Arista, Cisco, Dell, HP, and Juniper, and compared their R&D expenses to their revenues. Here are the results in M$ (to the best of my knowledge and understanding of accounting):

We all know that regardless of what the SDN and whitebox pundits try to tell us, we’re buying networking gear because of its software, not hardware (after all, most vendors use merchant silicon in one way or another anyway), and it’s expensive to create software – that’s why most networking vendors invest a lot in R&D.

Cisco seems to be an interesting case – they spend approximately half what Arista and Juniper do (percentage-wise), but then they do loads of acquisitions. I checked their latest annual report, and they report acquiring approximately 4G$ of intangible technology assets (read: technology they got through acquisitions) last year. When you add 1G$ per quarter to their R&D spending, they’re almost in line with Arista and Juniper.

It’s amazing how close the R&D spending (as percentage of sales) is for three major networking vendors are. Either they’re all equally stupid or there must be something you simply have to be doing once you get past the small startup stage.

The two exceptions in the above table: HP and Dell. Even when you add the occasional major acquisition by HP, their R&D spending is way below the ACJ crowd. Also keep in mind that one might do acquisitions to get interesting technology or to buy market share and customer base – one would have to check how HP reports what they got from the acquisitions (if you have time to go through yet another annual report and find the answer, please write a comment).

Long story short: it seems Dell and HP aren’t investing enough in R&D to be a viable long-term networking competitor. The best way forward for them seems to be to change the business model.

Dell is obviously going back to being an awesome logistics company – their long-term strategy seems to be Dell-branded hardware combined with third-party software (be it server or network operating systems).

HP seems to be going down the system integration (and services) path. Their latest whitebox switching announcement is a typical system integrator approach: buy hardware from vendor A, software from vendor B, and sell the result to your customer.

Does that validate the britebox model? No. We still have to see Dell and HP deliver on their promises (single support center that can actually solve your problems).

If you’re already using this model and are willing to share your experiences, please write a comment.

Does the britebox model spell problems for traditional networking vendors? Not really. Large companies preferring lean Linux-based networking software will find their way to Cumulus; those who prefer richer set of features under a unified CLI/API umbrella will stick with Arista, Cisco or Juniper – keep in mind Oracle is still selling their database products 20 years after MySQL was launched. Finally at least one of these three vendors I mentioned will eventually be smart enough to start selling software separate from hardware like Gigamon decided to do.

On a tangential note, the moves from HP and Dell will make life easier for people like myself – eventually I’ll be able to remove one or two vendors from my Data Center Fabric Update webinars, and I planned to include Cumulus in the May 2015 session anyway.


  1. I'm curious if your Dell and HP Sales and R&D numbers cover their entire business or just their networking offerings.
    1. It's their entire business. I was not able to find relevant numbers per market segment.
    2. And that makes your conclusion somewhat of a leap - the dollars Dell do spend on R&D may be disproportionately on storage and networking. Likewise a bunch of Cisco's greater spend will be on voice/conferencing/servers etc.
    3. Simon, I think you're right, however the most probable assumption would be that HP and DELL are also spreading their R&D costs among different products. And the difference isn't even 5% (which is already a lot) but rather 17-19%. Doubt that the point of this analysis was to provide some hard proof, but approximation looks interesting and even "explains" how Arista is keeping up with a game
      Just my 2c
    4. @Simon: You're absolutely right, but I wasn't able to find per-product line numbers. Cisco is reporting sales per product segment, but not the associated expenses, and most other companies aren't even that forthcoming. This back-of-the-napkin calculation was thus the best I could do, and as Anonymous pointed out, most companies would have to do R&D in most product categories they carry.

      Also, the difference is significant - Cisco (without acquisitions) is spending twice as much on R&D as HP with less than half their sales volume.
  2. HP has traditionally spent about 2.5-3% of their revenue on R&D, you can look at more data here:


    It's considered low compared to a lot of other tech companies, but again it's hard to say what that means. They make a bit deal about spending $3B/year on R&D. They are doing a lot of R&D in the area of silicon photonics which will play an increasingly important role in networking.
Add comment