(Note: this is merely the ramblings of a startup founder and should not be treated investment advice of any kind whatsoever. Caveat lector.)
Apparently, we're in another tech bubble. I was wondering the other day how Color Labs were going to spend the $41 million they recently raised, and the only things I could think of were marketing and website hosting -- the former because the free press they got out of that ludicrous investment will die down, the latter because it sounds like they've built the app with a client-server model, so they'll need lots of storage and bandwidth to manage it.
As I write this, Color.com has the IP address 188.8.131.52, which ARIN tells us is in the Amazon EC2 subnet. So I guess we know where a chunk of that $41m is going.
So, one startup is using Amazon. At work, we're using it for our programmable cloud spreadsheet, Dirigible, too. But are all the cool kids using it?
The best way I could think of to check that out was to work out what Y Combinator-funded companies are using. They don't publish a list of their portfolio companies anywhere I can see, but there's an unofficial list here. So, taking the companies started since January 2010, 80 in total, I made this sheet to work out where they were all hosted.
- About 26% of the companies founded in January 2010 use AWS.
- About 63% of the companies founded in June 2010 use AWS.
- About 71% of the companies founded in January 2011 use AWS.
There's a pretty obvious trend there, but it's not clear what it means. Perhaps the startups kick off using AWS, then switch to other hosting providers once they've got traction. Perhaps many of the startups from January 2010 are now bust, and the hosting provider shown is the one for the holding page that now sits on their domain. Or perhaps it's something else entirely; I guess the best way to track would be to revisit the sheet periodically and look for changes.
However, one thing is very clear -- new YC startups right now are overwhelmingly choosing AWS. Back in the days of the last dot-com bubble, people often said that it was best to invest in suppliers to startups rather than the startups themselves -- better, in a gold rush, to sell shovels to gold miners than to start prospecting oneself. Back then, that would have meant buying Sun Microsystems stock, which, hard though it might be to believe these days, would have been a great investment -- inasmuch as any investment could be in a bubble. Certainly better than investing in a single startup, because even back then, most failed. (Obviously once you have money to spread your bets across a range of startups, things change. But I don't, and nor do most other people.)
So, is now a good time to buy Amazon stock? Well, if I could invest in AWS alone I would. But as far as I can make out from Amazon's last SEC 10-K filing, a maximum of 3% of their revenues came from AWS in 2010 (AWS comes under "other" sales, which totalled $953m, and total sales were $34bn) . It's hard to unpick what the associated costs where, as unfortunately they don't seem to split it out from their general technology spend. But their gross profit ($7bn) is seven times the best-case AWS revenues, so it doesn't sound like it's a major component.
So, ultimately, investing in Amazon is probably more a bet on their other businesses. They're pretty good at commoditising their complements, so may be a good pick. But they're not a bet on startup growth.
On the other hand, they're clearly blowing away the competition, at least as far as the sample set of companies are concerned. So investing in other cloud hosting providers like Rackspace sounds like a really bad way to bet on the bubble. And at least that's one useful (if tentative) conclusion.
I clearly need to post more stuff here so that it doesn't just turn into a blog announcing the LFPUG's meetings :-)
However, in the meantime, here are the details of the next one: it'll be on 11 March 2010, and is hosted this time by Man Investments Ltd at Sugar Quay, Lower Thames Street, London EC3R 6DU. As before, all are welcome, but for security reasons you need to register in advance; just drop an email to Didrik Pinte. (Update: old mailto link removed.)
Guest of honour this time around is Travis Oliphant, the creator of SciPy and the architect of NumPy. He'll be talking about NumPy memory maps and structured data-types, and Didrik will also give a talk about integrating C/C++ libraries using Cython. More suggestions for talks (or even better, offers to give talks!) are very welcome -- once again, just email Didrik, or post something in the LinkedIn group.
The next meeting of the London Financial Python Users Group will be on Feb 3, 2010 at 7pm, and is being kindly hosted by KBC Financial Products at their offices: 111 Old Broad Street, EC2N 1FP (just opposite Tower 42).
All are welcome, but for security reasons you need to register in advance; just drop an email to Didrik Pinte. (Update: old mailto link removed)
The topics planned for this meeting are:
- Improving NumPy performance with the Intel MKL - Didrik Pinte, Enthought
- Python to Excel bridges:
- "PyXLL, a user friendly Python-Excel bridge" - Tony Roberts
- Discussion on connecting Python and Excel (xlrd/xlwt, pyinex, win32com, pyxll, ...)
- Speeding up Python code using Cython - Didrik Pinte, Enthought
Didrik Pinte has put together a web page on the Python.org Wiki for the London Financial Python Users Group. Only a little content so far, but it will grow... if you're doing financial work in Python in London, do come along to the next meeting -- it will be 7pm next Monday (14 December) at MWB Regent Street, Liberty House 222 Regent Street, London W1B 5TR. You may have to put up with me talking for a while about a spreadsheet you already know everything about, but there will be interesting bits too ;-)
Last night, I went to the inaugural meeting of the London Financial Python Users Group. It was a small gathering (as you'd expect for a new group), just four of us, but a very interesting one. Didrik Pinte gave a presentation of an interesting new library called pandas, a useful layer on top of existing stats packages that provides some very neat data-alignment capabilities, and we also discussed what future meetings should involve (lightning talks are definitely on the cards).
As with all these things, the most interesting discussions were left for the pub afterwards. I was particularly interested in what Didrik had to say about Cython, a tool which lets you write CPython C extensions in Python (!). I'll have to play with it and see if it can easily integrate with Ironclad...
Anyway, the next meeting is very tentatively planned for 14 December; the best way to track the group is probably currently the LinkedIn group, though I will definitely also post the details when they firm up.