Linden Lab’s Metaverse Maths: 1 + 1 = 10

by Alphaville Herald on 16/08/07 at 8:56 am

SL must grow constantly to be profitable – a sophisticated Ponzi scheme?
3000 servers for 50k simultaneous users – 1/3 of LL’s income from selling new servers each month

by Gwyneth Llewelyn, mostly harmless meta-resident

GwynethWe believe Linden Lab is a profitable company, at least since April 2007, because Jeska Linden blundered that out in a SciFi e-zine interview. Speculation on the costs of running a grid and providing the Second Life software have somewhat settled down — except, of course, in the minds of the clueless residents who swamp the Official Linden Blog with whining. For these, Linden Lab “makes millions a month” (or, who knows, a day), is about the size of Microsoft, and should be able to fix any bugs in a few days by hiring a couple of thousand programmers. In fact, the whiners make the usual mistake of mixing up income with profit — for them, earning millions is just what they see, but they have no clue on what the costs are.

On the more thoughtful blogs, however, people analyze the real statistics (at least, the ones that LL is willing to share with us!) and are quite skeptic at LL’s claims of “being profitable”. In fact, they add the little numbers up from what is LL’s income, and see that it’s impossible to run the Second Life Grid with so little money.

Between the two extremes — just looking at the income, or just adding up values for the costs — lies the very thin line of Second Life’s profitability. So, how much money is Linden Lab really making?

cHex your math is wrong because a land baron typically maintains a trading tier. He pays a flat $195 a month per sim. He buys and sells land, and tries to never let that trading tier sit there idle or be unsold before tier day. So he doesn’t pay $195 five times if he manages to buy and sell five sims that month; he buys monthly tier for $195 once, then keeps it filled and selling constantly as his trading tier. To be sure, the end users who pick up from him then open up a tier, but that parcel is now off the land baron’s tier bill and he goes to buy another piece, removing it from that person’s tier bill, etc.


First, to do some proper maths, the major issue is knowing what the variables are. On the costs side, if you never ran a distributed application over a few thousands of servers from a co-location facility (or two), it’ll be quite hard to know where to start to research for data! Then, of course, there are salary costs and office space costs. These are also notoriously hard — Linden Lab has quite a lot of people working for them, and although we have a rough idea on how much a Liaison is paid (that information is often leaked), we have no idea on how much Philip makes. Or Cory. Or Robin. Or, for that matter, qArl or Cube, who work part-time. Or Ginsu.

To make matters worse, LL attracts people because it’s a cool company, and not necessarily because they pay well. Philip had a dream job working as a consultant for Real Networks, he might very well be on LL’s roll for zero dollars — he has enough to live on his own for a few decades until Second Life really makes serious money (ie. billions of US$ per year ;) ). And finally, office space is very complex to figure out on a company where most people do telecommuting anyway, and the rest are spread around five office locations — two in California, one near Seattle, one in Boston, and one in Brighton, UK. Then you have to add on top of that running business costs, which differ in each location. Luckily for us, we know how much LL spends in advertising — mostly zero, but their public relations agency, Lewis PR, are an outsourced contract, and so are their lawyers (except for Ginsu).

On the other hand, we have the idea of the rough income that LL makes. This is not straightforward to figure out of the published statistics, but you can get a good idea on how much they make, since they publish all the data for enough months to give you a good average on the monthly income. This is actually the easiest bit. Still, at some stage, there are a few assumptions to be made. And when those are simply wrong, the results of any analysis will be completely off the mark, too.


Let’s look at the easy bits first. Peeking at the economy page, we can roughly split up LL’s major sources of income in the following way:
* Monthly fees for Premium users. This is around 100,000 residents, but many are on quarterly/yearly plans and pay less than US$9.95/month. The cheapest Premium users are the ones on yearly payments and just pay US$6/mo., so let’s take the average, for an income of roughly US$800k monthly, and growing 5-6% monthly.
* Monthly fees for private islands. For July, this means about 8700 islands. Not all are paying US$295/month, though, since many have the “old price” of US$195, and a lot are universities/schools and other non-profits which pay half the monthly fee. But let’s assume that the average is about US$200, for US$1.7M.
* Then you have tier for the mainland. This is harder to calculate, since although we know there are around 2700 sims on the mainland, we don’t know how occupied this is with land actually sold (for instance, all public land will not count towards tier, of course; and a lot of special areas — like the ones set apart for Relay for Life or other similar initiatives — is being held by LL for these cases). So let’s assume 2500 of those sims are really being used in the mainland. The worst case scenario would be if only 2500 users would pay tier for the whole sim, for US$195 per month, so this would be just US$0.5M. But I’ve read somewhere that the average land held by the Premium accounts is around 1024 m2, which would pay a tier of US$5/mo., so this is actually the same value (100,000 Premium accounts paying on average US$5/mo. of additional tier).
* But another huge chunk comes from all the setup fees of new private islands and sim auctions on the mainland. In the first case, LL rolls out almost 1000 private islands per month, or close to US$1.6M. The mainland grows much slower these days, it used to be 150-200 new sims every month, but there were less than 100 added last month. Still, their economy page says that about the equivalent of 178 sims were auctioned, and the average seems to hover between US$1700-2000 per auctioned mainland sim, so we can add another US$300k.
* And finally we have the LindeX transactions. I’m not counting on those, since the US$0.30 LL charges will very likely be eaten up in PayPal fees and other processing charges.

So, very roughly, LL has an income, by mid-2007, of around US$5M/month. Not bad, and remember, this is growing every month, so it’s not like you can predict they’ll make US$60M/year. It might be more, depending on the sustained growth of Second Life.

Infrastructure Costs

Now let’s see how much this costs in infrastructure. For some reason, LL was advised to buy their servers instead of leasing them. It has advantages and disadvantages, but one thing is for sure — they add 1 server per 4 sims, so, this means that every month they need to buy around 300 servers per month (!). Remember, we’re talking about rack-mountable top-of-the-line units here, not your US$500 “handmade” PCs. It’s very likely that they’ll pay around US$4k per server, and that’s with a good discount, based on the huge amount of servers they have to install every month. So US$1.2M will be “lost” on that.

Then we have co-location costs, the most hard thing to figure out. People tend to get used to buy their own servers and host them for as little as US$100/mo. with “unlimited” bandwidth (or at least a huge amount of TBytes/month). Well, co-location facilities have different pricing levels, and since LL apparently buys their own servers, this means that they most likely pay a fee per server cabinet (typically, that will take around 40 servers) and for the total bandwidth consumed.

Most co-location facilities do not list their prices publicly, and we can only make an educated guess at what LL’s setup truly is (a press release once mentioned they had 10GBps of total bandwidth — one link to Dallas, one to the outside world — which is consistent with the amount of simultaneous users that log in: users will, on average, require 100 Kbps per connection, and this would roughly allow up to 100k users at the same time), but here is an example from a facility in SF: they charge US$650/cabinet/month, and US$450,000/month for the total bandwidth of 10 GBps. LL requires 75 cabinets for its 3000 servers, so this would be another US$48,750/months. All together adds up roughly to US$500k — a bargain, since it gives LL around US$150/month for each server, which is not such a bad price. There are, however, several providers in San Francisco with more aggressive pricing — Linden Lab has picked Level 3 in March because of their (alleged) much higher reliability.

So infrastructure overall, at the current monthly growth, will be around US$1.7 million — not too much.

Office space

This is where my calculations usually go wrong, since there are two things that need to be calculated:
* How many Lindens require office space?
* How much do their offices really cost?

The first is purely guessing — I think that there are around 250 “active” Lindens now, and perhaps 100 of them are bound to their offices, most in SF, the rest on the other locations. The remaining Lindens are either telecommuters (Liaisons and some programmers), work part-time, or are simply “advisors” or “consultants” and very likely aren’t physically bound to LL’s HQ or branch offices.

Estimating how much office space these people take is hard. A rule of thumb is to consider 90-100 square feet per person, so, roughly, 10k square feet for all the Lindens (if you have been to their offices, let me know your impression on how big their HQ really are!). This article, however, claims that on April 1st, 2007 (yes, I know…), Linden Lab’s offices have 28.000 sq. f.! On some old (but very nice-looking!) buildings on Battery Street, SF (where LL is located), the cost is around US$20-30 per square foot, but I have no way of knowing if this is a good or bad price. Let’s assume US$30/sq. f. for a total of US$840,000/month in office space only in SF.


And finally, the hardest thing to estimate! How much do Linden Lab employees actually earn? Your guesses are as good as mine, of course, specially because Linden Lab is a “cool” company. Being a “cool” company means that people will work there for way under the usual salaries. Liaisons, for instance, often get only US$500 for a part-time job — but they become Lindens, and that’s all that counts! With 150 Liaisons or so, that would be about US$ 75k/mo. in salaries — a bargain!

The techies, developers, and the ops team will very likely not go out of their beds for less than US$5k/month. Some, indeed, might earn much more than that — since they are highly specialised professionals, and only a handful of them are available in the market for developers; everybody has been hired by Amazon, Google, eBay, PayPal, or any of the other huge companies that have thousands of developers in the same field as Linden Lab: complex, distributed systems. And they would be working on a way cooler environment — who cares about having 60k simultaneous users, when Google has possibly 60 million simultaneous users or more? :) So, we have to wildly imagine that, out of the 250 Lindens, they earn on average US$5k/month, taking into account that many will earn much more, and quite a lot are just there part-time. So this would add up to US$1.2M in salaries.

Barely Profitable

Putting it all together, Linden Lab earn perhaps 5 million dollars per month, and spend close to that in running costs.

My estimates are, obviously, quite conservative. On top of all the above will come a series of hidden costs, like maintenance of the PCs and Macs for all the team (not to mention chairs, tables, and other office supplies), license costs for Havok, Jira, and other pieces of software, and of course the phone costs, even taking into account they are using a VoIP operator to handle that. Also add the cost of supporting payment gateways, outsourcing costs for lawyers, auditors, and Lewis PR… there is not too much of a margin for “going wild”.

In essence, they have created a business model that barely allows them to survive as a profitable company. Like my many previous articles on LL’s business model, it has not changed much since, say, 2004 — the prices have risen just for private islands, and they had huge savings when introducing hardware that allows them to run 4 sims per server, but, beyond that, the only way they remain profitable is thanks to discounts from growth. If they need to upgrade their bandwidth to 20 GBps, this won’t cost twice as much. If they start buying 500 servers per month instead of just 300, they’ll get slightly better discounts. The truth is, LL is a classical type of a cost-based, telecom/infrastructure operator — it means that their prices are not speculative (like, say, web design, consultancy, or even metaverse development), but tied very closely to the number of users they need to support.


From a business plan perspective, there is a huge flaw in their model: a huge chunk — over a third of their income! — comes from new private islands being added every month, ie. LL is only a profitable company if SL grows constantly and at least linearly (that means that every month at least 1000 new private islands have to be added). There is always a danger if SL’s growth stagnates: if no new users are adding private islands like crazy, Linden Lab will lose too much income on a month and will not be able to keep it up.

You can think of it as a very sophisticated model of a Ponzi scheme :) which only works as long as it constantly grows, exponentially, with more and more users buying more and more sims. Unlike, say, web hosting services — where the setup fees are mostly speculative, and can be forfeited as a discount to long-term users — LL cannot afford to do that, or else their business model crumbles to dust.

They do, however, have several options. One, of course, is simply raising the costs of private islands, but that will make residents go wild — since they’d have to charge twice as much! The other is to lease servers instead of buying them — that means a long-term commitment, but they would very likely be able to forfeit all setup costs on private islands that way (or make them very low) at a slight increase of monthly fees (since they’d have to pay at least for a US$100 lease on each server) — but they would also cut down the monthly acquisition costs. More important than that, they would not require SL to constantly grow every month to keep the company afloat. In fact, if SL simply stagnates in its growth, they would simply have the same cost every month.

The other option is technological advance. It seems that they’re already toying around with quad-dual-core servers which would allow them to run 8 sims per server (ie. a “Class 6 server”). This would, over time, allow them to support twice the number of sims for half the running costs. But there is more. As some people installing OpenSimulator have found out, there is no real “hard” limit on the number of sims you can run per server, it’s all a question of how much performance you wish to have. OpenSimulator, of course, is a “bare-bones” sim server which doesn’t have a physics engine and doesn’t even support in-world rezzing (yet), so it just tracks down avatar positions, chat and IMs — allegedly, on a “normal” PC, you can run a hundred sims if you wish, assuming you haved enough bandwidth for 4,000 avatars (rougly 400 Mbps of upstreaming bandwidth).

But LL only manages — barely, as we all know — to run 4 sims per server. Is it all due to keeping track of 60k prims, plus 160 avatar and their attachments, and the physics engine? Where are the bottlenecks on the server code? We all know that Havok 1 is really an outdated piece of software (current version is 4.5, version 1 has been obsolete for half a decade) which slows everything down, but is it the only bottleneck? World of Warcraft has the same number of accounts as Second Life, but millions go online simultaneously every day on just a few hundred servers (as opposed to LL’s 3000 servers for barely 50k simultaneous users).

Ultimately, what LL really needs to do is to make sure they can run the same grid, using far fewer servers, and allowing an increased number of simultaneous connections. This would also allow prices of private islands to dramatically fall! So, if they can get ten million concurrent users (out of 100 million active accounts) on just 3000-5000 servers, they’d struck gold — and that’s what their ultimate goal should be.


I’m not part of Linden Lab, or their accountants. I’m not preparing due dilligence to buy Linden Lab, or help them preparing for an IPO. As a matter of fact, I’m rather messy at accounting. From the many business plans I’ve done in the past, only two of them ever were fruitful, so I can’t even claim to be good at analyzing companies. So take everything I say with a pinch of salt (or even two pounds). Really. Caveat Emptor.

37 Responses to “Linden Lab’s Metaverse Maths: 1 + 1 = 10”

  1. Jessica Holyoke

    Aug 16th, 2007

    A very intelligent article. Yay Gwyneth. And who says the Herald can’t do both business and bukkake?

  2. Anonymous

    Aug 16th, 2007

    Interesting first-guess analysis, and I’m normally a big fan of your writings however this time there are a few important issues that should be addressed:

    * Mainland region rollout estimates are wrong: As explained in blog post, current rollout is about 10/day which means about 300/month
    * You forgot an important source of revenue- classifieds (inworld + web)
    * You also ignored the exchange transactions. I believe you’re right in assuming that $0.30 fee when purchasing L$ goes to PayPal etc, but you are forgetting about the 3.5% commission when purchasing US$ for L$. According to Meta, in June US$609,187 was purchased via LindeX, for example.
    * The LL boardroom. Take Mitch Kapor for example, the board chairman. That post is his only for-profit venture..
    * I won’t even bother addressing the Ponzi scheme claim, as it’s merely an attention-grabbing, un-Gywn-like headline in my view. It’s sufficient to only mention LL’s well publicised commitment to ‘getting out of the way’ of the growth by open-sourcing the sims.

    In general, your analysis is very interesting in terms of a first-look at public LL financials, but I am glad you also recognise the fact that the numbers you have provided are only guesstimates.

  3. Tracy Welles

    Aug 16th, 2007

    While we can do the math, I think we might just be over doing it here.
    The fact is, anyone can see that Linden is charging a majority of the costs associated to lease a sim right up front. Based on your perception of server related hardware costs of around 4K (which I would typically agree), the sale of such islands pays for the servers. Charging 295 per month (not looking at the old rates), X4 on a server pays the bandwidth costs. I would wager that with mostly empty sims, it more than pays.

    While the rates climb, fewer are able to afford to be part of. This not being a part of is what will be the demise of the Linden Economy and business model. Open source the client for improved code, while yet charge the same for anyone to be able to actually be part of. Just not a sound model.

    Point being, no matter how you look at it, the business situation looks grim to me. Simply because of how it’s being done.

    I believe in the next few years things will change drastically. People want what we witness in SL, but have no way of being any real part of. The changes will take place with projects such as opensim or other topology methods for interconnect where the user themselves will be hosting their own sims and creating their own grids. Why?

    Why pay Linden for their servers and license(s) when you can operate your own sims and connect to those you choose to connect to?

    It’s going to happen. The only reason people are paying for Linden’s hardware and bandwidth is the hopes to make money. This soon too will be in decline as other more “rational” methods of making the grid(s) happens.


  4. urizenus

    Aug 16th, 2007

    Great job, but I hate it when people say ‘income’ when they mean ‘revenue’. I know that’s how we use it in the US (usually calling it ‘gross income’), but internationally, the convention is that income is what you get *after* you subtract expenses from revenue. But yeah, ok, we know what you meant and again great job.

  5. Hon Hin

    Aug 16th, 2007

    I remember the LL announcement about running 4 sims/server, but that works only for rarely populated/sea areas which have few concurrent avatars. This means that LL has to buy even more servers than you calculated.

  6. Inigo Chamerberlin

    Aug 16th, 2007

    That’s an interesting ‘back of an envelope’ analysis. It might even be pretty close to the real situation. Certainly LL are paring costs wherever possible.
    That said, the decision to buy rather than lease is one that’s always puzzled me. It might make some kind of sense in a smaller – much smaller – operation, but it makes very little sense for Linden Lab, and hasn’t done for a long time now.

    But, most worrying is, with such tight margins there is virtually no margin for error. Or for unforeseen reductions in the userbase. Linden Lab have indeed set themselves on a ‘grow or die’ course.
    One wonders if there is any contingency plan in place in the event of ‘shrinkage’?
    Somehow I doubt it. Which is kind of worrying.
    Especially with the current decrease in land values which must affect sim auction sale prices – and influence the desirability of buying an island for many potential purchasers.

  7. DaveOner

    Aug 16th, 2007

    Nothing like math from the outside looking in!

    I’m sure I could spin the numbers to make LL look profitable but that’s not internet-hip these days, is it?

    Now if Forbes came out with this kind of article everyone would put on their “anti-SL Hater” hats and talk about how they’re just frustrated because they don’t understand SL.

    At least this article was actually written well, though!

  8. Coincidental Avatar

    Aug 16th, 2007

    Thanks for the calculation, which I do not bother to read, because I already know that industry standard in this business is 30% profit margin.

    And yet another spoiler, Entropia Universe plans an IPO in LSE. They probably believe that the VR market is cooling :-)

    I believe too, that LL is profitable. The question is, how long?

    By the way, it is astonishing to me that the accounting statements of LL are not public, like the law requires them to be here. What a closed society :-)

  9. Thraxis Epsilon

    Aug 16th, 2007

    Hon Hin… you’re refering to the “low-prim” islands… which are purchased as blocks of 4 for the same cost you would pay for a normal private island sim. Which is not the same as the number of SIM’s being run per server. If you had 4 sets of those low-prim islands (16) they could all be running on one physical server.

    Each physical server runs 1 normal sim per CPU or 4 low-prim sims per CPU

  10. Tenshi Vielle

    Aug 16th, 2007

    GREAT article Gwen. Thanks. But… can someone explain to me where Phillip is getting this cash to spend weekends on Hilton Head island with Bill Clinton? I remember hearing something about that a while back.

  11. Viridian

    Aug 16th, 2007

    An interesting, if speculative, article. However, some common-sense conclusions that back up the main points can be made through observation. LL seems to want to keep both the exchange rate and the price of land fairly stable. This makes sense for a number of reasons but my gut tells me that the higher price of land allows a certain bit of padding to the bottom line. Also by now LL knows about how much they’ll take in through these sales, allowing for reasonable income/revenue predictions. I don’t think that LL is running at a loss – but what’s most interesting about thinking the line of thinking presented here through is the long term health of SL. Will SL be faster, better, bigger in 2 years, or will costs cause it to become somewhat stagnant?

  12. janeforyou Barbara

    Aug 16th, 2007

    *Uploading a pickture = 10L…not a lot lol,, but if 25.000 do 3 a day thats = 2500 UDS X 30 days = 75K USD= pays for 15 Linden workers hehe.

    * Classified, 1500 shops? 5000 L a week? ( i think its more ) 25kUSD a week x 4= 100k USD = pays for 20 Lindens workers.

    * Islands 9541, there are 4 Island on one = 2286 servers.From Februar 2006 bout 6000 new at 295 USD tier = 1,7 M/USD The old are bout 3541 Islands at 195 USD tier = 690k Usd = Total 2,4 M/ Usd… i belive but are not shure we find 690 K USD more here
    And then Total i found are 865 K USD more = 173 Lindes worker you can pay?
    I may be wrong,, but then this are just playing with numbers hehe

    This are August 16-2006.

  13. Lewis Nerd

    Aug 16th, 2007

    1+1=10 if you’re talking binary…


  14. Ananda

    Aug 16th, 2007

    Another very well thought-out analysis! In particular you pinpoint physics as the bottleneck which forces Linden Lab to invest in so many servers to begin with… I’d love to see a follow-up on just what the problems are with physics processing. It would certainly make sense for Linden Lab to come up with a way of optimizing this load, especially since the vast majority of islands are unoccupied much of the time. Or find a way to take advantage of the parallel processing capabilities some of the new GPGPU’s make available. I just have this suspicion one of the new Tesla GPU-based servers could handle many hundreds of SL simulators if the code were optimized for it.

  15. SqueezeOne Pow

    Aug 16th, 2007

    “…can someone explain to me where Phillip is getting this cash to spend weekends on Hilton Head island with Bill Clinton?”

    He was rich before SL, Tenshi.

  16. Ronin Kurosawa

    Aug 16th, 2007

    You might want to check the definition of Ponzi Scheme. Ponzi Schemes take money from new investors to pay off old investors.

    Despite what some residents might think, Second Life is not an investment vehicle.

  17. Stacey Sugar

    Aug 16th, 2007

    I like the information given, however there are considerable over-estimates on cost of office space & bandwidth costs. Just two areas which need a little more care and attention:

    1) Office space at $20-30 a year is what is quoted not per month. If office space was $360 per year per square foot, the landlord would be offering NASA quality floorspace lol (average london highest quality space as in the landmark Egg in the City is approximately £85 per square foot per year). That is a saving of $9M a year on your quote.

    2) It is unlikely LL are paying $45 per Mbit on bandwidth. They are taking a lot of bandwidth and I know that anyone can get Cogent bandwidth for a flat rate of $10 per month per Mbit. Even if they are going for premium bandwidth, which is unlikely knowing LL’s cost cutting strategies they wouldnt be paying more than $15 per Mbit per month. Therefore saving $300,000 on your quote (plus it is possible LL may have less than 10Gbps, more likely 6-8Gbps)

  18. Brennan Planer

    Aug 16th, 2007

    Gwen, Linden Lab’s is NOT paying 4K a server, otherwise Mainland would run like gold. 4K for a server is usually what is spent by an MMORPG server, and with 7K people on them there is no lag.

    The max a server in Second Life can hold is 120 Avatars (4 Sims), if they did spend 4K, I could expect them to run more.

  19. Gwyneth Llewelyn

    Aug 16th, 2007

    Thanks for all the comments :) The headline is, of course, the editor’s privilege, to catch the attention of the weary readers :) Alas, I’m not good at writing tabloid-style ;)

    Anonymous, some considerations on your comment:

    1) The figures for new islands/mainland sims I used are actually wrong, I took them from (Islands Added, July 2007 — 859) and (61.21 km2 growth from May to June, or 979.36 new islands/mainland sims). I don’t remember where I came up with the figure for 1200 total per month, and I do apologise for that; I don’t have snapshots of the figures, so I can’t tell if LL actually went back in time and changed the values :) In any case, my figures are at least 25% off — sorry about that. Exercise for the reader: go back with your calculator and correct them (the overall scenario will not change much).

    Zee Linden is just talking about regions on the mainland, I believe. In May, they launched around 250 new regions (and I’m assuming that Zee’s X/day figure means business days); in June, slightly less than a 100, according to the metrics spreadsheet. So this would actually mean less than what Zee is claiming, unless, of course, what’s on the spreadsheet is measuring something different (ie. bought land as opposed to auctioned land — in that case, however, it would mean that there would be hundreds of sims on auction!).

    2) Classifieds are in L$, so they’re a money sink, not revenue for LL. If they were clever, they’d do it in US$ and earn some interesting revenue that way :)

    3) To the best of my knowledge, the US$0.30 / 3.5% are actually PayPal charges. However, even if they’re not, 3.5% of US$600k/month is an insignificant amount, which is consistent with what LL always stated. It’ll cover the salary of the Linden that maintains the LindeX and LL’s accountant perhaps.

    4) I agree, we have no way to know how much LL’s board of directors and advisors cost.

    5) See above :)

    Tracy, the reason for interconnecting with LL is to access a booming economy and society with 9 million users, a few petaflops of assets (a billion items?), and thousands of events per day. For a long time it’ll be hard to beat that, so “new private grids” will want to give access to all of that, while still profiting from land sales (read: hosting 3D content). Obviously, companies, universities, and groups of friends might not have any interest in all that content and all those people, so the question is moot for them. However, the assumption is that Second Life will evolve “just like the web”, and although it’s fun to run a website at home just for you and your roommates, the *real* fun comes from interconnecting to the “world-wide web”… at a cost (ie. your ISP’s cable/ADSL infrastructure).

    Uri, sorry about the mixup with income and revenue :) I’ll be more careful reviewing the text for the next time. Thanks :)

    Hon Hin, you’re possibly mixing up the “void sims” with the virtual servers. LL has gone through several generations of servers, the most popular being the Class 3 servers (one region per server) and the Class 4 and 5 servers (four regions per server). In the days of the Class 3 servers (now all outdated and retired from service), LL experimented with virtual servers, and offered the “void sims”, and did in fact run four of them per server. These days things are different: *all* the grid — specially after the upgrade that made them all voice-enabled! — runs on virtual servers, each holding a region, and four per server. I imagine that if they still have old Class 3 servers, they’ll be used just for “void sims”; if not, the newer Class 4 servers will very likely be hosting *16* voids per server. In my calculations, I disregarded completely the void sims, although it might explain why LL’s figures for “new islands” and “overall area” is different.

    Inigo, that was what worried me most!! In my old days of doing business plans, one of the thing that my financial directors told me to watch out was for those “inflection points”, ie. where exactly your profitable company will suddenly run at a huge loss when a variable just changes slightly. They did some spreadsheet wizardry (“elasticity analysis”, if my memory doesn’t fail me) to figure out exactly where these points on the graphs were. A solid business plan will not show any of those “inflection points” — if they grow way beyond expectations, or way below expectations (“way below” being, say, a “shrinkage” in customers by a drastic 50% or more!), the company’s BP should still show it profitable. Now LL has a serious weakness i their plan, since it relies on a “grow or die” strategy, and these are the worst possible nightmares for managers: how to make sure that the company *constantly grows exponentially* just to make sure they’re afloat? Really, this will get nervous breakdowns for any analyst doing due dilligence on LL if they ever go public or are bought by someone else. This is also the major reason I don’t believe in either of those scenarios — unless we’re truly on a Dot.Com 2.0 Bubble (not proven!) where “ideas sell” (again…), honest and professional financial analysts will never go for a “grow or die” strategy, unless, of course, there is a Plan B that we have never heard about.

    Notice that raising the prices is a way out under this scenario :) Bad news for us poor faithful residents… (LL has, in fact, already raised the prices for private islands once)

    DaveOner, thanks for the nice words :) And yes, the margin for profit vs. loss is small enough that any accountant with a spreadsheet could easily show either scenario, depending on whom they’d present the numbers to :)

    janeforyou, as said, uploading pictures and classifieds are L$ money sinks, unless LL is selling L$ for profit on the LindeX, they don’t count :)

    Lewis, it was meant as a pun ;) Glad you spotted it, though.

    Stacey, the sq.m. cost were taken straight from real estate agencies in San Francisco, and I specifically looked up buildings in Battery Street, where LL is located. I have no idea if they’re an average, if LL is on an old building or a new one, or if they have a special deal for some reason. All could be true (or all could be wrong), so I really don’t know. The Internet is *NOT* a reliable source of information. Still, if your point is that LL could save running costs by relocating their HQ *or* making better deals than the ones announced by the real estate companies in SL, your point is well taken! Yes, keeping those astronomical costs down would certainly help LL out to continue to be profitable.

    The figure for the bandwidth costs were taken from the only co-location facility that has a Level 3 connection (we know that LL uses that) and lists their prices publicly for San Francisco. Almost all co-location facilities, including the one LL uses, have a form to get you a quote, and no prices listed publicly. I didn’t want to go through the hassle and trouble of getting in touch with a sales rep and ask them to deliver a proposal just to write an article, so I was lazy. Sure, LL might be able to negotiate a much better price for their 10 Gbps bandwidth.

    Brennan, two things. First, the *usual* price for a quad-CPU processor (or twin dual-core processors) in a 1U rack will indeed cost close to US$4k. LL sometimes posts the stats for their servers online, or talks about them in-world during the office hours, so, although we don’t have precise specifications, we can roughly estimate what “type” of machines they have. If they actually pay US$3500 or 5k, well, it’s anyone’s guess — we don’t know what brand they’re buying from these days (although we know that the Dell islands run from a Dell server :) ), and we also don’t know the discount volume they get these days. Rest assured, though, that this is the *range* of what a machine with quad-CPU/twin dual-core CPUs actually costs. Unless they’re insane and build their servers with off-the-shelf components for a quarter of the price and invest in a HUGE amount of floor space on the co-location facility, which I doubt they’ll even *allow* (it simply would take too much space).

    Your other comment is actually right on the spot: “if they did spend 4K, I could expect them to run more”. So very true! And there’s exactly where LL has to invest in their software development: get far more avatars per island, and do it soon. If they could handle ten times as more avatars per CPU, they’d be able to be *very* profitable, or, put it in another way, they could reduce the price of private islands and mainland regions *a lot* ;)

  20. Gwyneth Llewelyn

    Aug 16th, 2007

    Oh, I forgot: comparing SL’s infrastructure to a MMORPG will naturally fail. The system is completely and utterly different. For a very old article on how the grid works (it’s outdated, but it gives a rough idea), let me shamelessly plug an old article of mine:


    Aug 16th, 2007

    250 FTE? I really really doubt it. More like 125. Slash your labour costs in half.

    And it’s not revenue versus income, it’s cash flow positive versus income. I suspect they are cash flow positive, but you’re right, they might not yet be profitable.

  22. Gwyneth Llewelyn

    Aug 16th, 2007

    BTW, Stacey, the 10 Gbps bandwidth figure comes from this press release by Level 3:

    However they mix up two things: the announcement of the two co-location facilities AND “two” 10 Gbps connections. So which is which? I’m assuming that they are counting the bandwidth between both co-location facilities AND the “external” bandwidth. I might be wrong, though. As for the US$45/Mbps/month value, again, that comes from the *standard* pricing for Level 3-connected co-locations. LL might have negotiated better prices though

  23. Coincidental Avatar

    Aug 16th, 2007

    “I’m not part of Linden Lab, or their accountants”

    But probably their semi-intelligent press affiliate, who found a better place to disperse the Holy Word than her blog. Anyway, this doesn’t matter because I don’t by default believe what peeps say and this Second Wife business will be history in four years. Unlike I.

    By the way, is this Meat Linden having a feeble excuse like a holiday not to publish the economic statistics of the previous month? The statistics should not yet be bad enough to be censored, at least compared to the statistics of August.

    And there was that nice rush to purchase 1e6 m2 of land just before the announcement of reduced land supply. The insider traders will burn in hell, because the value of the land will inevitably approach zero.

    Anything positive to say about Second Lie? No, but the viewer of There sucks even more than SL client SW. That’s an amazing achievement from the competitor.

    About Opensim:

    “OpenSimulator, of course, is a “bare-bones” sim server which doesn’t have a physics engine and doesn’t even support in-world rezzing (yet)”

    OpenSim seems to be developed by half a dozen programmers as a hobby. A couple of months ago the LL opinion was that the server side is too complex for the SL community (I hate that word) to develope.

    “Where are the bottlenecks on the server code?”

    Well, let’s call it the recruiting policy. I never would employ convincing people over competent people in technical jobs. On the contrary, I dislike convincing people because I have difficulties in seeing through them what they are worth and what they are hiding.

    Or are you implying that the performance statistics provided by SIM is broken?

    Third thing is that these guys never designed the SW performance in mind. You see, “I will be gone far far away when this shit hits the customers…”, “Improving HW will take care of this” and “It will sort out itself” type of system design.

    Improving HW doesn’t help when the amount of needed HW grows faster than the HW performance.

    SW is a problem magnet. It never sort itself out.

    I’m glad to see that some basic facts of performance have creeped from WoW cooperation to LL, but they really should not have needed the cooperation with WoW to learn that.

    “This would also allow prices of private islands to dramatically fall!”

    I don’t want to take a look at what private island “owners” now have in their pants, but the message: “Land price is heading down” is coming from different directions.

    “So, if they can get ten million concurrent users (out of 100 million active accounts)”

    What about this: “The maximum concurrency of SL will never exceed 100 000″ ? Not because of technical reasons but because of competition and SL going out of fashion.

  24. guest

    Aug 16th, 2007

    Join PN today and help fight against the fascist mods and furfags!

  25. Carl Metropolitan

    Aug 16th, 2007

    I think you are missing a revenue stream for LL. They don’t only collect a US$0.30 charge on transactions, they also print and sell anywhere between 900K to 500K US$ worth of L$ a month. That’s mostly profit, and could easily be 10% or more of their income.

  26. KMeist Hax

    Aug 16th, 2007

    I do believe the whole problem is physics. If you didn’t notice, avatars are physical too. This is why sitting on a cube will prevent you from being pushed, because cubes (under normal situations) are non-physical. For the same token, wonder how /b/tards like PN manage to take down sim after sim after sim? PHYSICS! Physics makes lag, physics makes crash, physics fail. If your numbers are correct, LL has been shitting it’s time away on shit like voice and WindLight. They should be working on Havok 4 and Mono; stuff that makes the core product stable. In fact, we should petition about it sometime.

  27. Gwyneth Llewelyn

    Aug 16th, 2007

    Coincidental, “semi-intelligent press affiliate to LL” would be a lovely tag to place on a card :) Maybe I should apply for a job at LL ;) However, in spite of my optimism, the whole point of the article was NOT to “spread the word” that LL is a hugely profitable company, but to point out that they’re *barely* profitable (although likely with a positive cash flow by now, just a tiny trickle though), and that they rely *hugely* on massive growth *every month* to keep running under this particular business model. Which is sadly flawed and very dangerous — it can fall apart with another bad decision like banning casinos. What will come next? Banning mature content? Well, that’s about 2/3 of the sales, and then what? Banning “broadly offensive content”? Another one like that, and 90% of LL’s revenues would be cut down and the whole thing would immediately collapse, not in months or years, but *days*.

    So LL has to be VERY careful when making those insane policy changes at the worst time possible, AND they have to seriously review their business model to make it far less dependent on *continuous growth*. That’s just the business side of the equation, not to mention the technical side of it!

    WoW, in effect, seems to have a perfect infrastructure and a sound business model: 10 million users faithfully paying their US$19.95 every month, a few hundred servers handling 7000 players per CPU (and not 40-100!), and getting a million users simultaneously online with very likely far less demand on bandwidth (no texture streaming!!). Well, that’s a cash cow! Blizzard could even charge US$1 per month and still be immensely profitable… while LL struggles with their business model to barely stay afloat.

    Carl, I’ve read about how LL “prints money” every day to keep the L$ stable. Well, supply and demand. In the past, there has been more demand than supply for L$, so this strategy has worked out, and yes, it’s possible that LL has managed to add another half million US$ to their income. However, again, like relying on continuous growth, this is a short-term strategy that collapses as soon as people simply don’t want to buy more L$ but sell them — and in that case they need to remove their offerings on the LindeX *and* increase the money sinks. There is not much of a margin to play there.

    A friend of mine, who is much cleverer than me, suggested that LL should actually encourage banking in SL (and not forbid it, like so many have been yelling on the LL official blog) and act like the FED or the Central European Bank — loan money at a fixed interest rate to SL banks, and thus, in turn, be able to “pull out” L$ out of the economy if needed, while at the same time providing auditing to SL financial institutions. The concept is interesting and could be explored more, but I wonder how willing LL is to think about that, since it requires true interference in the economy — way more than “subtly” printing a few more L$ every month. Worth a thought, though.

    Kmeist, I totally agree. Physics is one of the major technical bottlenecks. You can add to your list something way more drastic that happens on the servers. When a sensor is fired, what actually happens is that a “spiralling” object is created on the spot where the sensor is, and it “collides” with all avatars in an area. Each collision is reported to the LSL sensor. Some have even postulated that things like figuring out the chat range works the same way. All this places an unnecessary burden on Havok (believe me, there are far better ways to deal with it *without* using Havok), specially with version 1 of Havok, which is notoriously slow at dealing with such issues. Some have described Second Life as a “front-end to Havok”, ie. ultimately the physics engine is doing almost all the work — the rest is really dealing with the asset server, IMs and group IMs, and transactions. Well, push everything into Havok, and it’s quite obvious that Havok will not handle it any more :) Definitely, a fifth attempt to port Havok 4.5 to SL should be made (as well as integrating Mono!) as quickly as possible. Who knows, we might even be able to get a thousand avatars per CPU that way, and this would certainly allow for completely new possibilities undreamed of in SL!

    Bandwidth is also an issue — 10 Gbps allow barely for 100,000 simultaneous users. If they want a million users simultaneously, that means spending at least 5 times as much on bandwidth. This is sadly a bottleneck that won’t go away easily — *except* when LL introduces the possibility of loading textures from other servers (like ActiveWorlds does — you can simply point the texture to an external URL). Textures will be loaded much slower (but then cahced on the client) since they won’t come in LL’s ultracompressed format, but the key issue is that they won’t come from the Grid at all, but from other servers, thus lowering the load on the whole grid. Why doesn’t LL implement this? Well, mostly because once you know the URL of the texture, you can easily copy it. Sure, it’s true that with the OS client and libSL you can copy *everything*, but the required know-how to do that is limited; whereas copying an URL pointing to a texture is available to *anyone*. Still, this will come; and I expect that a lot of people, well before offering a new grid, will be offering hosting services for superfast textures — at, say, L$5 per upload :)

  28. cHex Losangeles

    Aug 17th, 2007

    I believe you underestimated mainland tier; a brief example will illustrate how. Take the case of a single 512m2 parcel on a popular newly-auctioned sim. The auction winner terraforms, develops, and subdivides the sim, then sells it off. A mini-baron buys a portion of the sim, marks it up, and sells it. Someone buys our hypothetical 512m2 parcel, uses it for a couple of weeks, then finds something better and sells it again, and again it is purchased by a mini-baron, who again sells it to an end-user. In this scenario, tier for that 512m2 parcel is paid FIVE TIMES.

    Every time mainland property changes hands, an entire month’s tier on that property is charged; the monthly tier can be realized by LL several times in a month.

  29. Arthur Fermi

    Aug 17th, 2007

    Why does it always come down to the sims … I believe that the bottle necks are primarly on the networking side and not the server side. I say this, because on days with a large volume of people you can go to an empty sim and still experience lag. This isn’t a sim issue, its a networking issue. Packet loss has some really nasty side effects, every packet that is dropped, is resent, additonaly, there is a packet sent to let the server know a packet wasn’t received. As packet loss increases, so does traffic requesting the information. Now this wouldn’t be too bad on any individual sim, the real problem lies in trying to hit the server cluster that is the “asset server”, they would quickly get over loaded, affecting, rezing, teleporting, yada yada yada. LL needs to make a concerted effort to ensure that the networking comes up par with the servers.

    Additionally if they are routing significant portions of their traffic, they will see a slow down as routing is not as efficient as switching, causing bottle necks.

    They said recently they have a VPN connection between the data centers, this is an additional source of lag and over head. VPN means encryption, decription, larger packets. Most companies requiring dedicated bandwidth, and performance would look to technologies like MPLS, ATM, and direct fiber between colocation facilities.

    Arthur Fermi

  30. Hiro Pendragon

    Aug 17th, 2007

    1 + 1 does = 10 though.

    Binary. duh. :D

  31. Prokofy Neva

    Aug 17th, 2007

    I’ve got a different set of numbers, which has them both way more profitable, but also having way more expenses — so perhaps it comes to the same thing, but I don’t think their purpose is to be come profitable.

    You forgot taxes! Also travel.

    My concept of LL as a non-profit organization:

  32. It seems to me that a discussion of this sort might be well conducted using something like a Zoho spreadsheet. There’s got to be some kind of publicly editable calculation system that would allow folks to easily display the results of slight modifications in the numbers presented, allow modified bookkeeping systems to be presented, perform the math for the users, etc.

  33. Gwyneth Llewelyn

    Aug 18th, 2007

    cHex, there is a good reason for having underestimating tier on the mainland. First, unlike my previous calculations in the past, nowadays the mainland is “hardly interesting” in terms of tier, since private islands outnumber mainland regions by far. If the trend continues, the mainland will be “mostly irrelevant” in terms of revenue, except for auctions. Also, it won’t be a surprise if LL decides, once and for all, to outsource the whole mainland continents to someone else (IBM?) and simply forgets about it. In any case, granted, it *is* a source of revenue (specially because of the multiple tier payments for the same land in the some month — good point there!), and it might be the difference between being cash-flow-positive or negative.

    Arthur, I agree that bandwidth is one of the most underestimated sources of lag. As people start hosting their OpenSims at their homes, and suddenly figure out that at most 2-3 friends can log in through their ADSL/cable connection, people will realise how important “massive bandwidth” really is. An average of “100 Kbps per user” is not much, specially when that means that during peak hours you might not have much more than that available per user! Loading *several hundreds of megabytes* of textures on a crowded sim that barely struggles to give you those 100 Kbps will definitely lag things a lot — but that’s because the sim is doing *both* things, texture streaming AND figuring out avatar/prim position, all through the *same* channel. A more interesting design would be to separate assets from position tracking — a good reason why all static-content MMORPGs are WAY faster than SL, they can easily handle 7k avatars/CPU if they only have to deal with position tracking. So, yes, bandwidth is an issue, and having servers with dual purposes (asset serving and position tracking) will compete for the same bandwidth — instead of getting position tracking and seeing avatars/prims move, your computer is being bombarded with textures and sounds and animations and painfully waiting for the next bits and bytes that tell where people are moving or what they’re saying in chat…

    Hiro, Lewis spotted the pun first ;)

    And Prokofy, I agree — I haven’t figured out not only the taxes, but also missing the additional running costs like travel, etc. For instance, how much does Lewis PR cost? Or their (external) lawyers? What about accounting costs? Put that all into the “cost” section, and things look grim. Alas, the “taxes” issue is a tricky one: are you taxed on revenues or profit in California? :) If they’re only taked on profit, they might deliberately show themselves as an almost-non-profit-company by increasing their expenses (that’s a very typical trick in my country, where you’re only taxed on profits and keep a running account with the IRS on value-added tax — a good accountant will always have the company very close to be as low profit as possible, and spend as much in services as providing them, thus having the VAT accounting also as close to zero as legally possible).

    Sue, great idea :) )

  34. Guybrush Contepomi

    Aug 21st, 2007

    > And finally we have the LindeX transactions. I’m not
    > counting on those, since the US$0.30 LL charges will
    > very likely be eaten up in PayPal fees and other
    > processing charges.

    Today US$1.061.271 was spent in 2ndL. For the sake of
    simplisity, lets assume this equals the ammount users
    daily feed into the echonomy.

    Using todays exchange rates, this would buy the punters
    L$265.824.616, which in turn sells for US$860.883.

    From what I can see most of the transaction fees are
    passed on to the punters (US$0.30 when buying L$ and
    3.5% when selling), so my guess would be that Linden
    actually sees most of that US$200k/day.

    Hardly something to be ignored.

  35. Mercia McMahon

    Aug 23rd, 2007

    Gwyneth, you over-estimate Mainland tier in your 5 times over example, primarily because you underestimate how slow the property market is nowadays. The thought that land could change hands five times in one month, including a resident holding it for two weeks is simply staggering (like to buy the sim I cannot offload and try your hand at land-baroning?).
    There are other errors as well. If the non-baronial user has no other land no tier is paid at all. If fact it is *technically* possibly if unlikely to have a sim of 512m2 plots owned entirely by people with no other land and so the entire sim earns Tier of 0US. You also forget tier limits: I have bought and sold a lot of land because the sim I bought had 12000m2 of Protected Land in it and I deeded the land to a Group getting 10% bonus. So without changing my 195US tier payments I have been able to manouvere 13200m2 approx., while still being stuck with this sim unsold.

    “Every time mainland property changes hands, an entire month’s tier on that property is charged” No, every time a new tier level is triggered, the difference between that tier and the previous one is charged on top of what that Resident is normally paying.


  36. Maria LeVeaux

    Aug 25th, 2007

    Actually, that Equation is more properly expressed as 01+01=10 and would be mathamaticly Correct.
    There are only 10 kinds of people in this world, those who Understand binary, and those who don’t.

    A Question on your calculation of Office space overhead, Did you factor in the Server Locations in at least two other cities? I didn’t notice that, and there are also utilities Costs Outside rent, and those basics provided by the Building management. If those expenses are factored in then it would Appear on the Surface that LL has actually been operating at a loss for some time. A Condition i doubt would carry on more than a Few months before the various financial Trade papers “outed” their insolvency. I think, Not being privy to everything that Goes on at LL you have probably Miscalculated somewhere.


  37. Nurse practitioner jobs.

    Sep 30th, 2008

    Pain management nurse practitioner.

    Nurse practitioner uths san antonio.

Leave a Reply