Why even consider VDI? The "CapEx versus OpEx" see-saw

Before I begin, I want to thank Brian and Gabe for inviting me to blog on Brianmadden.com and especially Gabe for nudging me that I should stop procrastinating and get on with it.

Before I begin, I want to thank Brian and Gabe for inviting me to blog on Brianmadden.com and especially Gabe for nudging me that I should stop procrastinating and get on with it. Thanks guys!

For years now the debate over PC vs. VDI has raged on. I’ve watched it from the customer and vendor side of the fence and feel that the discussion too often is incomplete which causes confusion for so many. I’ll begin with reminding everybody that we used to talk about FAT CLIENTS. 

Fat clients were and still are great for many use cases, which we have understood for ages. However, we also understood many of the challenges we faced with them. Filled with crap from the IT department that you didn’t need or want to use, patching nightmare, inventory maintenance, configurations drifting, software packaging and distribution overhead amongst many other things that ultimately led to slow change in a managed environment due to so many moving parts that needed to be tested together. Broadly speaking we called this OpEx (operational expense). It’s well understood that this is the largest part of the cost of the PC lifecycle as shown in fig 1 below. 

Fig 1: Comparing the Fat Client OpEx to CapEx. 

PCs were not that expensive once they matured, i.e the CapEx (Capital Expense) was lower than the OpEx. As a result we understood that we needed to optimize our PC plant to not just reduce OpEx, but to also be able to make changes in the environment faster. To do this, we’d have to invest in better management tools and a certain level of service or settle for chaos and security risk with so many still allowing local admin rights. This was consistent with what Microsoft was also telling the world. “A better managed PC is a cheaper PC.” 

Then with much fanfare came along VDI and mistakenly over asserted claims that it would solve the ills of the fat client era. Every vendor made the mistake in the early days of leading with the promise of cost reduction, not appreciating how diverse an enterprise PC environment was. I recall the “power of one” message from Citrix and many similar promises from VMware. 

I’d submit that the reality for most is illustrated in figure 2 below. On the right of the seesaw (also known as teeter-totter) it’s the combined Fat Client story and on the left side it’s the VDI story reality for many.

Fig 2: VDI vs. Fat Client – OpEx and CapEx view. 

That reality is higher CapEx and OpEx for VDI. The higher CapEx in VDI comes from needing client and server hardware as well as storage etc. The higher OpEx comes from the reality that the marketing message of single image management from the VDI vendors is not yet feasible in an enterprise at scale with a diverse user base. Attempting to do so adds further operational complexity. I have witnessed first hand people who have tried to brut force their way to scale this model and it’s not delivered the ROI that was hoped. This has resulted in scaling back of the VDI project because the politics of the initial architecture made it very hard to save face and change course. 

As a result many stick with what they know works and end up implementing limited VDI in a 1-1 model with existing management tools. This of course further adds to CapEx. 

The CapEX challenges are well-understood and the industry is making good progress to solve these. However in my assessment the logical conversation stops here for most and they don’t think hard enough about the OpEx reality. This starts with many analysts I have spoken to who add to the customer confusion, by over rotating on CapEx and therefore poopooing the 1-1 model. I’ve argued many times that there is too much focus on just talking about the CapEx and a lot more light needs to be shone on the OpEx components.  In fact I believe that’s the smoking gun that should be front and center of the conversation. The answer I’ve got back in the past is that customers don’t believe OpEx and buy on CapEx. This line of thinking really irritates me since it’s fundamentally flawed.

Customers have understood for a long time that OpEx is all about a better-managed PC irrespective of an organization’s will to accept some of the caveats that come along with this. So overly focusing on the CapEx of VDI is the wrong answer. CapEx is a side effect of VDI that needs continued innovation as part of a VDI solution out of the box so it’s comparable to fat client CapEx. The good news here is that progress is being made.

OpEx however is a different ball game. I don’t believe the VDI vendors will, can or should fix this problem entirely. It’s a desktop management challenge, and it needs to be fixed by management vendors unless the current crop of VDI vendors, decide that they truly are going to become desktop management vendors.  I have limited expectations, as any new capabilities VDI vendors develop will be focused on future workplace concepts. Sure we’ll get small acquisitions and lots of marketing promises, but show me an established vendor who invests forward on a platform that they also say is dying as part of a previous era?

To make matters worse, I see next to zero innovation from the traditional systems management vendors to do anything about this except for stay as status quo and follow a roadmap based on incrementalism. Yawn.

Therefore, I expect more successful enterprise scale VDI deployments to be based on a 1-1 model and an increasing need for these enterprises and the partner eco system to be willing to write their own tools leveraging open source where they can. I also hope the smaller private vendors who are trying to help customers move to a shared model, gain more traction and continue to scale up their solutions for enterprise customers. However I still believe that for most, 1-1 is the path of least resistance. 

This all seems like a lot of risk and not high on the priority list of things to do. Perhaps the fat client is better, so why even consider VDI? 

I’ll share some of my experiences and observations in my next post. 

Join the conversation


Send me notifications when other members comment.

Please create a username to comment.

My favorite line from this article is:

"...show me an established vendor who invests forward on a platform that they also say is dying as part of a previous era?"

It's one of the things I'm looking for in Barcelona, frankly. Did you notice how, for the first time that I can remember, Mark T led with something other than XenApp/XenDesktop numbers in a keynote? I think it took something like 80 minutes to get to the "Classic Citrix" stuff. That, combined with Citrix and VMware's acquisitions of solutions for managing physical desktops (Virtual Computer and Wanova, respectively), indicates a shifting philosophy.


Single Instance Management is envious and contagious.

Ranging from the use of variables in code, to individual functions, to whole applications, to the entire desktop OS, to networks, to the cloud.

VDI is another management story. VDI capitalizes on hardware virtualization to accomplish easier management the same way Bromium uses it for better security.

The current situation is fragmented. We need hardware virtualization to be used for both better management AND security otherwise we will remain fragmented as an industry.


Single image is an aspiration that is very hard to achieve on the desktop. I agree tools are not there and I'd add they just bring more overhead and cost with their own set of complexity. However that doesn't mean we should stop trying as it's a better end game.

@Icelus, when has the industry ever been together. Fragmentation is the norm :-) Even developers are all over the f'ing place and since their apps require such variance in the infrastructure, it's very hard to dictate a single infrastructure configuration to them.


Fig. 2 is subjective. When you are looking at management of External an Internal computing you can add SBC OpEx and CapEx to the PC OpEx and CapEx which may tip the scale the other way.


Preaching to the choir. I am responsible for Dev, Test/QA, and Ops and tightly integrated in the business. We are currently at war with a central org of 6000 IT pros and a 1.7 billion budget that want to take ops away to dictate a single infra config which will result in crippling the business.

They like VDI, and they want VDI.


@Icelus I trying to simplify the discussion by strictly focusing on VDI in fig 2.

As for classic SBC, I'd argue it suffers from the same constraints of single image management in terms of flexibility to represent a full desktop experience. Hence not a full desktop replacement, but good for the right use cases. Will talk about it more in a follow up post post Synergy.

@appdetective Yep aspiration accurately describes single image management. I agree we've got to keep trying and also point out that at smaller scale there is success. Hope :-)



If you simplify the discussion then it's not worth having IMO. VDI provides many benefits that regular Fat PCs don't have. Simply comparing the TCO of apples and oranges isn't sufficient.

Put VDI on one end of the scale and compare it to a complex PC environment with SBC and then you have a correct comparison. You will find that a lot of your SBC costs are in your VDI costs.

BTW. I am referring to VDI as a collective of technologies: SHVD, CHVD, and RDSH.


"full desktop experience" I think needs to be defined too.

The "full desktop experience" that I am referring to has a requirement to be completely locked down. If they want a free for all then they can either build their own network or work from home on their own PC.

When you belong to an org with 2 very different mandates it's very political.

I have personally seen an entire network built for the purpose of personalization and freedom to get the job done.

Think of it as Shadow IT, but nobody can argue business requirements.


@icelus Disagree with you. There is far too much confusion in understanding lacking capabilities in the technology stack. These get masked by marketing which is a different point to talking about benefits which also get masked in marketing unless one has an understanding of the underlying base capability.

Perhaps I should have broken this into a several part series, since the other valid points that you make are follow on posts that I will get to. Appreciate the comments though.


We will agree to disagree then :).

I apologize, I should have been more specific when I said "regular Fat PC", I meant what comes out of the box.

Out of the box with VDI, we have the capability to have a single instance management of the OS config/patches, core apps, AV updates, etc. In addition to this we also get a single architecture for internal and external workers. We can acheive the same with Fat PCs by combining different technologies, but I believe the hard part is the applications and now we are talking about business requirements which is subjective.

Personally, I have automated several complex application installations in a distributed environment and it gets the job done but is not ideal in case issues arise because by nature everyone is unique.

Therefore the scales are subjective based on environment. PC OpEX is underestimated along with VDI OpEX.

Great article and discussion, I look forward in reading the rest of them.


Harry! You're still alive?

I will have to take you to task on on your position that the capex costs for VDI are greater than the cost of distributed desktop PC. 18 months ago I would have completely agreed with you. At that point the cost of a VDI desktop was significantly more expensive than the cost of a PC. However, times have moved on. It is now possible to deliver the entire endpoint, computing, and storage stack for about the same cost as a typical enterprise PC.

I'll concede that at present a VDI desktop is no cheaper than a PC, but where PC prices have been more or less flat for years, there are still further economies to be realized with VDI hardware that will allow the cost to fall even further. And the way that BYOD is moving suggests that for many organizations it will soon be assumed that the cost of the endpoint will in many cases be borne by the employee. Although, that may be pushing the argument for VDI a little too far. The only point I will concede here is that there is a big difference between " the VDI reality for many" and the VDI opportunity if the right choices are made. Early adopters are saddled with technology that was not right for the job, and even today when the right technology exists, that message still has not penetrated deep enough.

OpEx is a much harder nuts to crack. The biggest problem here is not that VDI costs more, but that VDI is easier to measure. Leading to comparison between the full cost of VDI and the visible cost of distributed desktops. The biggest problem here I think is that VDI is usually implemented in addition to existing desktop technologies meaning that no matter how little VDI costs, the overall cost of IT tends to be greater.

Still no matter how much it costs, almost every cost model ignores the business benefits that more flexible, more reliable, more ubiquitous  application delivery technologies bring. Depending on where you measure the cost of IT services from greatly impact the argument for all again any new technology. What does it matter if it doubles the cost of IT, if at the same time it doubles the bottom line form of business.




Hey I’m new to brianmadden.com as a member, but have been following the site for some time now, mostly as a knowledge source as I’m still young and learning lots in this industry.

I love this blog – definitely brings it home as far as why this virtualization wave hasn’t taken over the entire planet completely. I very much agree with the statement “there is too much focus on just talking about the CapEx and a lot more light needs to be shone on the OpEx components.”

In reality, while you’ve changed the game when it comes to security and deployment, you’ve also just managed to add new licensing costs, management tools, and some extra headache to your plate. This means it takes longer and is a bigger headache to manage everything (OpEx is increased), which is why a lot of times companies don’t see the ROI they were expecting when they were sold into this VDI tornado in the first place.

Remember: Complicated is time, and time is money.

You will save on some things, but you will spend more on others. This is why it’s not been a complete game-changer. All the pros seem to be suggesting the best environment is a 1-1 between SBC and VDI. This is the path of least resistance, and therefore the easiest and least costly, and therefore the biggest ROI. This all adds value to products that allow for seamless integration between the ideal 1-1 ratio so that you can get the best of both worlds without spending all of your time managing the beast.

With all the new technology and subsequent marketing out there, we can often times get lost in the messaging and forget what is actually being done in reality. I would like to have a better idea of what’s actually going on in the marketplace, and I need some help with some answers.

How many what percentage of the marketplace (my target is companies with 50-500 users) is using basic TS/RDS hosted desktops?

What %’age has adopted VDI? (SHVD, CHVD, and RDSH) What are they using for management tools?

What %’age is using an enterprise-level solution like Citrix?

I know some of the larger companies are getting well into it, but how about the smaller guys?

Again, I can see what all is available out there; I’m wondering who is USING it… (if anyone knows any rough %’ages or where to point me I would appreciate it!)



I've done some informal analysis of customer environments over the years and found that 100 out of 100 customers can accurately report CapEx for existing Fat PC deployments as well as upcoming RDS/VDI deployments.  Those same hundred customers can also accurately report their OpEx as a before/after analysis.  The only problem is all 100 of them are lying.  This is the simple truth.  OpEx calculations are BS 99.9% of the time.  Next argument....