Tuesday, May 3, 2011

Longtop and the remarkable capital efficiency of the Chinese cloud

A note on the comments: the comments on this post are consistently of the view that (a) the company did not build a cloud-based storage infrastructure, (b) they did not build a petabyte data storage solution for anyone, but (c) they assisted a customer build a large data set (probably not a petabyte) and (d) otherwise the press release is a marketing puff-piece but not material to the stock.  They think I have made too much of a marketing puff-piece.


I agree that it is entirely possible that the press release is a marketing puff piece of marginal importance to the stock.  It certainly has all the buzz-words.  Please do not read this post without reading the comments which are - in my view - sophisticated and probably contain some truth. At least they made me think.


Some people who I trust think this could be done within the $1.5 million of incremental capital spent this year.  I am not going to dispute that point - but whatever - it would use most of the $1.5 million of incremental capital... and that leaves the question how did they manage to to fund the rest of their growth.


The core issue remains - how is it possible that this company has grown equipment and fitting so slowly and revenue so fast? Unless there is a miracle they can't be doing it by owning cloud facilities. The company has grown revenue by almost 50 percent in the last nine months adding almost no fixed assets.


So I remain puzzled.


Now read the post - here without alteration:


----------------------------

A while back I did a lot of work on cloud computing companies. This was not because I wanted to buy the cloud companies (by and large I do not) but because I wanted to understand what they did to the existing tech giants (such as Microsoft, Google, Apple, Hewlett Packard, Dell etc). You can find those posts here, here, here and here.

This stuff is important to us. Despite the (recent) content of this blog Bronte Capital's funds are primarily long equities. We short on the side. Recently shorts have been abnormally profitable. Long term returns will be driven by well-chosen value stocks.

We have big positions in selected tech giants - and we need to understand the threats to those businesses. So we went on a self-directed course in "cloud studies". [That course involved putting our own server and storage in the cloud...]

But you never know where your research will lead. "Cloud studies" took us to a remarkable press release from Longtop Financial - what appears as a reputable Chinese IPO (not a reverse takeover) with a reputable CFO. The remarkable press release described Longtop's efforts in being a cloud provider - and more particularly a cloud-storage provider.
Longtop Financial Technologies Limited (NYSE:LFT), a leading software developer and solutions provider targeting the financial services industry in China, today announced the official launch of its new data archiving solution. 
Like many organizations worldwide, financial institutions in China are experiencing explosive data growth. They are facing big challenges in continuing to store and manage the high-volume data for compliance purposes in a cost-effective way, as well as query and retrieve them efficiently for business needs. Longtop's data achieving solution has the capability to manage petabyte (PB) level structured data at both low cost and high access efficiency. The solution has the advanced data storage and management infrastructure built on cloud storage techniques, which has unlimited horizontal scalability in both storage volume and access efficiency with a high compression ratio. As a result, it can overcome the inherent deficiencies of traditional Relational Database Management System products in high-volume data management and help customers to build the new generation of data centers. 
"Much of the strength behind Longtop's solid organic growth track record comes from our proprietary product development. Over the years, we have consistently focused on Research and Development (R&D) and going forward we will continue to invest in human resources and technology to help further strengthen our R&D capabilities," commented Weizhou Lian, Longtop's Chief Executive Officer. "I am very pleased that one of the largest banks in China has been our first customer to deploy our data archiving solution."

What struck me as peculiar about this release is that being a cloud computing company is very capital intensive and I was wondering where they purchased their storage equipment from. After all storage equipment for the cloud is a hugely market-sensitive business (that was what the 3Par spat between Dell and HP was about see here and here).

More generally cloud computing is capital intensive. The cloud companies see themselves as utilities and have similar capital needs and economics. The main knock on the cloud computing businesses - as per this otherwise laudatory article about Rackspace - is that they are hugely capital intensive:
But analysts have questioned the capital-intensive nature of Rackspace's business model, especially because the company's capital-expenditure guidance will likely be higher than its revenue trajectory for 2011. 
"Our business will always be relatively capital intensive," he [Lanham Napier] said, largely because the company regularly has to upgrade its infrastructure. "The magic is in the margin of service layer we add on top of that infrastructure. We're making investments to continue to make the business more capital efficient, but we have more work to do on it."
And of all the cloud functions the one that is most capital intensive appears to be storage. You see when I am not using Bronte's server someone else can use the spare capacity. However when I am not accessing Bronte's storage nobody else can use it. Storage is not shared and thus likely to provide less capital advantage by going to "the cloud". And that shows in pricing - as Felix Salmon has noted cloud storage is too expensive for much personal use.

And yet Longtop have done cloud storage at the petabyte level (ie akin to Rackspace or Amazon) without very much capital expenditure. At the beginning of the 2010-11 financial year Longtop had fixed assets of $26.3 million. Nine months later - the nine months in which they launched their cloud storage offering - the fixed assets had risen to $27.9 million. Revenue grew in that nine months by over 40 percent (these numbers are in my previous post).

They managed to launch a cloud storage service at the petabyte level with almost no capital expenditure.

I was puzzled.

But the above mentioned remarkable press release told us that this amazing ability to do storage at the petabyte level with enough backup for a large financial institution came from their skills in human resource management. To quote again:
Much of the strength behind Longtop's solid organic growth track record comes from our proprietary product development. Over the years, we have consistently focused on Research and Development (R&D) and going forward we will continue to invest in human resources and technology to help further strengthen our R&D capabilities..."
These are remarkable skills. You see if Longtop can do this with only human resources (no substantial capital expenditure) then the $300 million odd of capex expected at Rackspace this year is wasted. There is probably half a billion of capex at Amazon EC2. That is wasted too.

Moreover because cloud storage can be done without buying capital equipment the huge bidding war between Hewlett Packard and Dell over 3Par (a company that sells capital equipment for cloud storage) was wasted and HP wasted billions of dollars.

It is all possible of course. Longtop are about to displace Amazon, Dell, HP, Rackspace and a score of other tech giants.

These Chinese companies really are that remarkable.

Just read the press release.





John

32 comments:

  1. Hi John,

    Although I share your scepticism, one potential explanation could be that they merely wrote some form of software and are acting as systems integrators (rather than IaaS providers) for a data archiving solution?

    I haven't gone through more than the press release yet, but

    (Very big fan of your blog by the way).

    ReplyDelete
  2. John,

    I'm quite surprised at the low quality of this article and that you didn't ask managment for details of their cloud offering.

    First up, no China banks that I'm aware of have hosted their applications on 3rd party cloud platform due to security concerns.

    Secondly, it seems that Longtop data archiving solution is a piece of software rather than SaaS.

    Thirdly, the press release mentioned Longtop's solution "help customers to build the new generation of data centers." which support point number 2.

    Finally, for due diligence purpose, you can't depend on a press release. You have to read a data spec shhet for the software product. IBM does this very well except that they call thier's white paper.

    ReplyDelete
  3. I read the press release, but they have not stated anywhere that they are directly providing the physical storage infrastructure.

    They could very well be just providing the software, which would require minimal cap-ex, and which they could just as easily expense (vs. capitalize).

    Just look at carbonite.com or a slew of other backup services. They are all built on top of Amazon S3, Rackspace Cloudfiles, etc. - and they are incredibly capital efficient.

    This would cost a few hundred k to build. Max $1 mm assuming absolute gross incompetence.

    BTW check out Avnver Mandelman's book, The Sleuth Investor.

    ReplyDelete
  4. They could be system integrator for other people...

    But that is not how the press release reads...

    And the 40-50 percent increase in sales with no capital deployed also jumps out at me as - well - unlikely...

    But of course it is possible... they might just be able to grow without deploying capital.

    ReplyDelete
  5. They could be system integrator for other people...

    But that is not how the press release reads...

    And the 40-50 percent increase in sales with no capital deployed also jumps out at me as - well - unlikely...

    But of course it is possible... they might just be able to grow without deploying capital.

    ReplyDelete
  6. Dear Anon

    I agree it is unlikely that third party clouds will be used.

    But I doubt - severely doubt - that any bank in China has a petabyte of data. The internet archive - which snapshots the internet every few eeks - has 3 pb of data...

    I severely doubt that Longtop has so little capital equipment either. They have about 14 million gross of "equipment and fixtures" which will wind up - after growth this year - and assuming some people were employed for that growth - about 2000 dollars per person. (It was less than 3000 on the old employment numbers).

    That is better than a laptop and a desk - but it is hardly the amount of computer equipment you would expect at a software firm with these margins...

    That said - banks do use EDS and CSC to do their back office - and they use them quite commonly.

    John

    ReplyDelete
  7. John,

    I guess you should edit your original article to show that your concerns lie with the fact that Longtop's fixed asset is laughably low for a comapny of its size.

    In addition, the honorable thing is to retract your statments questioning Longtop's data archiving solutions as I've never seen any hedge fund guys doing due diligence on the basis of a corporate press release.

    You don't want your investors to question your competence and integrity, do you?

    ---Lim Kim San

    ReplyDelete
  8. A petabyte of data? That's not cool. You know what's cool? etc

    Writing as a techie who works vaguely in that space, that PR reads like typical PR BS, they have some new software and have added the words "built on cloud storage techniques" to make it sound impressive.

    People have been adding "cloud" into their product descriptions since it became the New Thing. I hope nobody goes short on these guys merely because media hacks write BS into PRs.

    From their web site they look like a pure-play services company so lack of capex seems entirely expected.

    ReplyDelete
  9. The Petabyte of data not being cool is the nicest thing I have heard said about this.

    --

    Archive.Org - which copies the entire internet every few weeks - is 2.9 PB of data. Guess that is not cool either.

    But there is NOT a PB of data at any financial institution in China. Its just not there...

    And given there is not a PB of data - why the hell did they talk about PB in their press release...

    I guess they made up the press release for stock marketing purposes... like cloud computing releases before them...

    You could be right - the business is real but the press release made up.

    I guess that is a way the short could be wrong...

    But I am not sure I would want to be long companies that made up press releases in China... I might be wrong being short... but it would be a strange (but possible) outcome.

    J

    ReplyDelete
  10. How do you know they are actually doing anything with the Cloud stuff? It could just be a fluffy PR piece to pump up the stock price without anything of substance behind it. The revenues come from the "core" business lines.

    ReplyDelete
  11. Just ask S1 (SONE) about how capital intensive it is to provide "cloud" (back in Internet Bubble 1.0 it was called "outsourced") software and storage to the banking industry.

    The business was (and is) far more capital intensive than anyone had anticipated in Internet Bubble 1.0

    The segregation of client data is of great importance to the banking industry. Even the use of co-location on a single physical machine may be forbidden, further reducing the "efficiencies" of going to the "cloud"...or as it was called way back before Internet Bubble 1.0, Client/Server.

    ReplyDelete
  12. I think you're off by at least two orders of magnitude when looking at the storage involved and comparing against Amazon.

    One petabyte is 1000 terabytes. A one terabyte drive costs on the order of AUD$100 retail for a single drive in Sydney. To scale up, of course you need power, rack space, controllers, cabling, etc, but we're not talking about hundreds of millions of dollars any more.

    Let's say the capital costs are two times the cost of the drive for raw storage, and then double that for redundancy (drives fail), giving $600/TB in capital cost. Let's round up a bit more and have $1MM/PB.

    http://en.wikipedia.org/wiki/Petabyte

    http://en.wikipedia.org/wiki/Orders_of_magnitude_(data)

    The Amazon elastic block store lets you walk up and create a 1TB storage volume, and they will charge $100/month, plus access fees. Once 1000 customers do that, they are managing 1PB of data.

    The discussion at:

    http://andirog.blogspot.com/2008/03/how-much-data-is-in-amazon-s3.html

    Puts the S3 storage at 22PB in 2008. There has been growth since then, and I'd be surprised if they're not managing more than 100PB of data. (No real information of course, and I haven't even looked at their financial results to reverse engineer the figures.)

    So, I agree that many Chinese companies seem to be able to do magic stuff with almost nothing, but on the face of it, $1MM for 1PB doesn't seem off too much.

    ReplyDelete
  13. If cloud storage were as simple as buying TB hard drives and linking them then there would have been no battle for 3Par.

    This is complicated stuff - and you need the box to understand WHERE everything is.

    Amazon charge more like 100 dollars a month per terrabyte connected to a server...

    http://alestic.com/2009/06/ec2-ebs-raid

    These numbers are a bit old - but the costs have probably halved since then... but no my order of magnitude is not off.

    This company would have had to be truly miraculous to increase fixed assets by only 1.6 million whilst increasing revenue by 40-50 percent.

    I dont think they could have offered any cloud services at all (no fixed assets). But they must be doing something really special.

    Just wish I understood what.

    ---

    I mean these are really truly special numbers.

    John

    ReplyDelete
  14. If cloud storage were as simple as buying TB hard drives and linking them then there would have been no battle for 3Par.

    This is complicated stuff - and you need the box to understand WHERE everything is.

    Amazon charge more like 100 dollars a month per terrabyte connected to a server...

    http://alestic.com/2009/06/ec2-ebs-raid

    These numbers are a bit old - but the costs have probably halved since then... but no my order of magnitude is not off.

    This company would have had to be truly miraculous to increase fixed assets by only 1.6 million whilst increasing revenue by 40-50 percent.

    I dont think they could have offered any cloud services at all (no fixed assets). But they must be doing something really special.

    Just wish I understood what.

    ---

    I mean these are really truly special numbers.

    John

    ReplyDelete
  15. One more thing ...

    The place where data really grows is recording CCTV surveillance video. An reasonable quality MPEG-4 stream might be ~150kB/s. A network with 100 cameras generates 15MB/s, or over 1TB/day.

    Your quote from the press release doesn't describe the data being stored. But the data coming out of banks isn't just transaction data, some of it is quite big ...

    ReplyDelete
  16. This is a completely facile analysis and you should be ashamed of yourself for publishing it. The LFT cloud solution is simple and takes unique advantage of certain labor efficiencies available in China.

    Much like another extraordinarily efficient company, China Agritech, Longtop employs a number of Chinese superhumans who simply memorize terabytes of data (which is then recited to banking clients whenever they come calling).

    And that is why labor costs are rising at this company and capex remains minimal.

    I expect a retraction any day Mr. Hempton.

    ReplyDelete
  17. You don't get to a petabyte with transaction and account data.

    You can get to a PB with videocams.

    I guess it is possible they store forever and for access photos of people at ATM machines.

    I remember a decade ago talking to McData and they had huge systems a small fraction of this size which would do all of a major Western bank...

    J

    ReplyDelete
  18. John,

    Re: "But there is NOT a PB of data at any financial institution in China. Its just not there..."

    I don't think Archive.org is a good comparison. For example, Yahoo claimed in 2008 that it then had the largest data warehouse in the world at 2 petabytes. There is an article that says Facebook currently has a 21 petabyte database. So point being that PB databases are not that special or uncommon in these days. So I don't think it is a stretch to think that there are many PRC banks with PB databases at this time.

    Also I read the press release as saying they have the theoretical capabilities of having PB storage. It is likely that this is more of a sales pitch and they will go on to actually spend capital to build up the infrastructure once they get the business from client. Or as others have suggested, that they will simply help companies build PB databases.

    But one odd thing to me is that while I can understand startups relying on cloud storage, but I find it strange that any financial institution would be looking towards having a third-party manage their storage. It is simply too sensitive of a data to rely on someone else.

    In any case, I think the press release is really more of a sales pitch/marketing fluff piece than anything else. And from my experience, this type of fluff press releases are fairly common in the technology industry (US, PRC, etc).

    ReplyDelete
  19. I've been building systems that store data since the 80s, and I understand the issues involved. It is important to understand just how much cheaper data storage is getting. Yes, there are software issues, but the release makes the point: this isn't a relational database.

    There are big archiving requirements for surveillance these days; it is not unusual to have networks of 100s of cameras, and they all need to be recorded. Think monitoring on infrastructure, roads, trains, bank branches. The requirement isn't the photo taken at an ATM.

    Just a single train might have 30 cameras and an onboard system for storing 14 days worth of surveillance. Then consider a network of trains, stations, and you can see the need for a lot of storage. This stuff is practical now; it wasn't practical that long ago.

    Something like bank transaction data is much, much smaller.

    ReplyDelete
  20. John,

    Do the honorable thing and retract your statements

    ---Lim Kin San

    ReplyDelete
  21. If they are doing what they say they are doing in the accounts and buying the equipment for other people (read the last 20F which is what they say) then the margin will be lower.

    We have either (a) too high a margin or (b) too little capex or (c) a miracle.

    I do not need to retract my statements - its just that the nature of the miracle performed has changed....

    John

    ReplyDelete
  22. I am one of those people who thinks one Petabyte is not all that much. And again, the bean counters like spinning facts in their releases.

    ReplyDelete
  23. Who is Lim Kim San? And why so keen for John to retract? Are you speaking on behalf of Longtop?

    ReplyDelete
  24. Seriously Mr San.

    If they are buying equipment and on-selling it to the bank - something they say they are doing in the 20F - then the margin is unbelieveably high.

    I have a choice of miracles here. I am not however going to dismiss the possibility that the company is miraculous.

    J

    ReplyDelete
  25. Seriously Mr San.

    If they are buying equipment and on-selling it to the bank - something they say they are doing in the 20F - then the margin is unbelieveably high.

    I have a choice of miracles here. I am not however going to dismiss the possibility that the company is miraculous.

    J

    ReplyDelete
  26. Seriously Mr San.

    If they are buying equipment and on-selling it to the bank - something they say they are doing in the 20F - then the margin is unbelieveably high.

    I have a choice of miracles here. I am not however going to dismiss the possibility that the company is miraculous.

    J

    ReplyDelete
  27. I don't see what you're harping on here. They signed up a bank as a customer. They also claim to have PB-scalable RDBMS (which is bullshit, nobody has cracked that nut yet except for very flat databases). I don't see where they claim the bank has a PB worth of data (although an investment bank might, if they have enough tentacles). There's no claim or implication that they've bult a massive infrastructure.

    ReplyDelete
  28. I have to agree that the release does not say they built a petabyte cloud. They said they sold a solution that has the ability to manage a petabyte cloud.

    That's pretty easy to do, I just have to sell something that can use an EMC SAN effectively, or an AWS account LOL.

    FWIW, you can get to a petabyte with historical tick data on a few active stock exchanges pretty quick.

    There might be some other reasons the financials look suspicious, but this marketing puffery seems pretty typical and is not one of them.

    ReplyDelete
  29. John --

    I think you're off on this article.

    I'm an entrepreneur in the Hadoop/IaaS space, and from my perspective the press release is just standard fluff.

    First off, a Petabyte is not that much for a national size bank in the US (or China). To give you a sense of scale, if a transaction is a 100 bytes block, then a data warehouse containing 10B transactions would be a petabyte. A customer base of 100M would easily generate a couple petabytes of transaction data a year.

    In terms of the press release, if I didn't know better I'd just say that it sounds like they set up a Hadoop cluster for a bank to store their data. In this case, there are no licensing fees (hadoop is open source) and they would have booked most of their revenue as Professional Services/Consulting. It's entirely possible they just charged an installation fee/maintainence for installing their "solution" on top of bank purchased hardware. That's the standard arrangement for the industry. The bank in question wouldn't have been the first to drastically overpay for a storage/data warehousing solution.

    -A.S.
    P.S.
    I've been reading your blog for years and have learned a lot. Just think you're a bit off on the details on this one.

    ReplyDelete
  30. John,

    Do the honorable thing and retract your statments questioning Longtop's data archiving solutions.

    ---Lim Kim San

    ReplyDelete
  31. I do not retract the comments. The accounts are miraculous. They increased revenue by more than 40 percent on almost no incremental capital.

    There is a miracle going on. What miracle that is you can explain.

    Something is very - well - puzzling here.

    John

    ReplyDelete
  32. Lim Kin San

    Do the honorable thing and tell us you have no involvement with this company.....and you're not as shonky as the rest of Asian business.

    ReplyDelete