Technology publisher and Web 2.0 impresario Tim O’Reilly wrote a thought-provoking post today about the dynamics of the nascent cloud computing business. He makes some important and valid points, but his analysis is also flawed, and the flaws of his argument are as revealing as its strengths.
O’Reilly begins by taking issue with Hugh MacLeod’s contention that, thanks to “power laws,” “a single company may possibly emerge to dominate The Cloud, the way Google came to dominate Search, the way Microsoft came to dominate Software … We’re potentially talking about a multi-trillion dollar company. Possibly the largest company to have ever existed.”
O’Reilly argues that MacLeod mistakes the nature of power laws on the Net: “The problem with this analysis is that it doesn’t take into account what causes power laws in online activity. Understanding the dynamics of increasing returns on the web is the essence of what I called Web 2.0. Ultimately, on the network, applications win if they get better the more people use them. As I pointed out back in 2005, Google, Amazon, ebay, craigslist, wikipedia, and all other other Web 2.0 superstar applications have this in common.” O’Reilly goes on to argue that because many elements of cloud computing appear to lack this network effect – they don’t get better the more people use them – they won’t naturally evolve toward a monopoly or oligopoly. Here, he’s talking more about the infrastructure, or raw computing, services offered by, say, an Amazon Web Services and less about particular web apps.
Let’s stop here, and take a look at the big kahuna on the Net, Google, which O’Reilly lists as the first example of a business that has grown to dominance thanks to the network effect. Is the network effect really the main engine fueling Google’s dominance of the search market? I would argue that it certainly is not. And in fact, if you look back at that 2005 O’Reilly article, What Is Web 2.0?, you’ll find that O’Reilly makes a very different point about Google’s success. Here’s what he says, in a section of the article titled “Harnessing Collective Intelligence”:
Google’s breakthrough in search, which quickly made it the undisputed search market leader, was PageRank, a method of using the link structure of the web rather than just the characteristics of documents to provide better search results.
This has nothing to do with the network effect as O’Reilly defines it. What Google did was to successfully mine the “intelligence” that lies throughout the public web (not just within its own particular network or user group). The intelligence embedded in a link is equally valuable to Google whether the person who wrote the link is a Google user or not. In his new post, in other words, O’Reilly is confusing “harnessing collective intelligence” with “getting better the more people use them.” They are not the same thing. The fact that my neighbor uses Google’s search engine, rather than Yahoo’s or Microsoft’s, does not increase the value of Google’s search engine to me, at least not in the way that my neighbor’s use of the telephone network or of Facebook would increase the value of those services to me. The network effect underpins and explains the value of the telephone network and Facebook; it does not underpin or explain the value of Google. (Indeed, if everyone other than myself stopped using Google’s search engine tomorrow, that would not decrease Google’s value to me as a user.)
So why has Google’s search engine been able to steadily accumulate more and more market share at the expense of competitors? There are surely many reasons. Let me list several possible ones, all of which are likely more important than the network effect:
1. Google delivers (or in the past has delivered) superior search results as judged by users, thanks to superior algorithms, superior spidering techniques, or other technical advantages.
2. Google delivers (or in the past has delivered) results more quickly than its competitors (an important criterion for users), thanks to superior data processing systems.
3. Google has succeeded in establishing a strong brand advantage, in effect making its name synonymous with web search.
4. Google has, through partnerships, through the distribution of the Google toolbar, and through other means, made its search engine the default search engine in many contexts (and we know that users rarely change default settings).
5. Google has steadily expanded into new web properties and services that, directly or indirectly, funnel users to its search engine.
Now it’s true that, if you want to define market liquidity as a type of network effect, Google enjoys a strong network effect on the advertising side of its business (which is where it makes its money), but it would be a mistake to say that the advertising-side network effect has anything to do with Google’s dominance of the searches of web users.
The Google example, far from providing support to O’Reilly’s argument that the network effect is the main way to achieve dominance on the modern web – that it is the secret to the success of “all Web 2.0 superstar applications” – actually undercuts that argument. And there are other examples we might point to as well. Apple’s iTunes online store and software system has achieved dominance in digital music distribution not through the network effect (the company only recently got around to introducing a music recommendation engine that derives value from aggregating data on users’ choices) but rather through superior product and software design, superb marketing and branding, smart partnerships, and proprietary file standards that tend to lock in users. There are plenty of “social” online music services built on the network effect; none of them has dented Apple’s dominance. (I would also take issue with O’Reilly’s suggestion that Wikipedia’s success derives mainly from the network effect; Wikipedia doesn’t become any more valuable to me if my neighbor starts using it. Wikipedia’s success is probably better explained in terms of scale and scope advantages, and perhaps even its nonprofit status, than in terms of the network effect.)
“Ultimately, on the network, applications win if they get better the more people use them.” That’s a huge overstatement. Applications, or other kinds of online services, win for many reasons on the network. To be sure, one possible reason is the network effect. I’ve already mentioned Facebook’s success as an example. But there are plenty of smart network-effect services, including ones that O’Reilly singled out back in 2005, like Flickr and del.icio.us, that have not achieved widespread success. They definitely “get better the more people use them,” but they haven’t “won.” And there are plenty of other popular online applications – Turbotax Online, Apple’s MobileMe, MapQuest, Yahoo Mail, Basecamp, Google Reader, Mint, Zoho, etc. – that have achieved success not because of the network effect but because they’re useful, well-designed tools. (The original success of Salesforce.com, the most famous business web app, had nothing to do with the network effect, though Salesforce is now wisely trying to tap into the network effect, through, for instance, its force.com development platform, to extend its success.)
So what does this mean for the eventual shape of the cloud computing business? One thing it means is that, even on the pure infrastructure end of the industry, power-law distributions (wherein a small number of companies end up capturing most of the business) may well emerge for reasons having little or nothing to do with the network effect. Indeed, this new industry seems particularly well suited to a concentration of market power. Here are some of the reasons why:
1. Capital intensity. Building a large utility computing system requires lots of capital, which itself presents a big barrier to entry.
2. Scale advantages. As O’Reilly himself notes, big players reap important scale economies in equipment, labor, real estate, electricity, and other inputs.
3. Diversity factor. One of the big advantages that accrue to utilities is their ability to make demand flatter and more predictable (by serving a diverse group of customers with varying demand patterns), which in turn allows them to use their capital more efficiently. As your customer base expands, so does your diversity factor and hence your efficiency advantage and your ability to undercut your less-efficient competitors’ prices.
4. Expertise advantages. Brilliant computer scientists and engineers are scarce.
5. Brand and marketing advantages. They still matter – a lot – and they probably matter most of all when it comes to the purchasing decisions of large, conservative companies.
6. Proprietary systems that create some form of lock-in. Don’t assume that “open” systems are attractive to mainstream buyers simply because of their openness. In fact, proprietary systems often better fulfill buyer requirements, particularly in the early stages of a market’s development. As IT analyst James Governor writes in a comment on Macleod’s post, “customers always vote with their feet, and they tend vote for something somewhat proprietary – see Salesforce APEX and iPhone apps for example. Experience always comes before open. Even supposed open standards dorks these days are rushing headlong into the walled garden of gorgeousness we like to call Apple Computers.”
The network effect is indeed an important force shaping business online, and O’Reilly is right to remind us of that fact. (I should also mention that O’Reilly’s post includes other points that I’ve not discussed here.) But he’s wrong to suggest that the network effect is the only or the most powerful means of achieving superior market share or profitability online or that it will be the defining formative factor for cloud computing. Hugh MacLeod is probably right that we will in time see a striking concentration of market power in the cloud computing industry, and the network effect probably won’t have all that much to do with it.
O’Reilly seems determined to believe that his definition of Web 2.0 explains everything about the online business world today, arguing that “the cloud platform, like the software platform before it, has new rules for competitive advantage [and] chief among those advantages are those that we’ve identified as ‘Web 2.0,’ the design of systems that harness network effects to get better the more people use them.” That’s a half-truth parading as a truth. While the cloud may explain Web 2.0, Web 2.0 doesn’t explain the cloud.
UPDATE: Tom Slee chimes in: “if we are to really get to grips with industry concentration in new Internet-driven industries we need to acknowledge both sides of the story – that different industries are pushed by different forces (different parts of cloud computing will see different levels of concentration), and that there are multiple sources of increasing returns (it’s not all Web 2.0 network effects).” I agree with that, though I would add that one of the big question marks about the ultimate structure of the cloud computing industry – maybe the biggest – is whether the “different parts” of the cloud (infrastructure, development platform, apps) will remain separate or whether they will collapse into a single supply model. Will, in other words, the companies that run the data centers also end up supplying the apps? Already, Google is pursuing a model that spans all three layers, raising the question: Will the scale advantages in running the infrastructure also lead to advantages in supplying the apps? I hope not, for the sake of maintaining a robust web apps sector, but my guess is that they will.
James Governor isn’t making much sense.
I think he’s intimating that Microsoft has less data lock than Apple. That’s not true. Apple’s non-DRMd audio, for example, uses AAC – a standard Apple doesn’t own. Their document formats are no more or less proprietary than Microsoft’s.
Apple uses SQLite to store calendar, contact and task data, and FileMaker sells Bento to manipulate and extract that data.
Apple loves lock-in of course, but so does Microsoft. Governor is misguided and misleading.
On the other hand, Google, surprisingly, has published API for their data stores. They’ve made it possible to export and import blogs. Google’s Cloud has data freedom — signing up for Cloud services without data freedom is terminally stupid.
Speaking of which, I don’t see that you’ve pointed that out anywhere …
This is typical of folks who are more comfortable with the consumer side of technology and merely invoke Google, amazon, MS. Any enterprise cloud discussion which does not include IBM, HP, the major telcos, major sw vendors like SAP and Oracle foolhardy. Since all of them are gearing up to provide cloud services, combined with amazon, a bunch of start ups, at least in the near future there is little risk of any given vendor cornering the market. In the enterprise outsourcing market today there is plenty of fragmentation with the biggest – IBM with less than 10% market share. No reason to suspect the cloud services market will be as if not more fragmented
You are of course right in that Google’s advantage is not due to network effects.
I see Google’s dominance mainly due to first mover advantage (first with page rank) and switching costs (hard to get used to a different ui). Still, Google does enjoy network effects in that the larger the user base, the more an advertiser will pay for ads.
To me O’Reilly’s article was about “The big winners are those who best grasp the rules of the new platform.” which for any entrepreneur is deciding if this is a great business to be in.
Google’s search may not rely on network effects, but its advertising system certainly does. Google sees increasing returns to scale from its network in matching ads to content, setting prices (especially for obscure keywords), and having a wide network of publishers to syndicate ads.
Nick,
I think you fail to see the network effect driving Google. It’s really that they have done a better job than anyone else in harnessing the underlying network effects of the web itself. Despite your contention to the contrary, pagerank is in fact a mechanism for extracting a kind of implicit user contribution that no one else had realized was there. What’s more, Google is better at spidering, and better at using data from what people are searching on to improve their results.
As to the ad auction, network effects play a role beyond even the one you identify. It’s Google’s ability to mine clickstream data that made it possible for them to sell ads not to the highest bidder but to the best combination of price and expected click through rate.
This is obviously a very different kind of network effect than is demonstrated by facebook. That’s the most trivial and obvious kind, a network that gets better because more people are using it.
I suppose that I am using the term a bit differently than most people do when I suggest that Google, and Amazon, and Wikipedia gain advantage from network effects in user contribution. These systems all do get better the more people use them.
I agree that I probably am overstating the case when I say that this is the only source of business advantage. Of course it isn’t.
Apple, as you note, is a good counter-example. Yet they also demonstrate a strategic response to software commodification, one that I’ve also written about in the past. See http://www.oreillynet.com/pub/a/oreilly/tim/articles/future_2003.html, where I wrote:
Another source of value is in design. In his essay “The Birth of the Big, Beautiful Art Market” (collected in the book Air Guitar: Essays on Art and Democracy), Dave Hickey describes how, after WW II, Harley Earl of GM turned the marketing of automobiles “from being about what they do to what they mean.” His point was, as industries become commoditized, intangibles play a greater role in product differentiation. This is now happening in the computer market. Apple has been a pioneer in marketing computers for what they mean rather than what they do. Everything from the 1984 ad to “Think Different” speaks to the self-image of the user who chooses an Apple product. But the rest of the market is catching up to them.
In short, the kind of premium that proprietary software has enjoyed in the Microsoft era is likely to be significantly reduced, till it is equivalent to the proprietary value that, say, Mercedes has relative to Toyota or BMW or GM. Anything good will be copied quickly. There will be some engineering advantages, but new market momentum often comes from design innovations rather than technical superiority.
In any event, you failed to address my main point, namely that cloud computing is likely to be a low-margin business, with the high margin applications found elsewhere.
Real “lock-in” is less caused by technology than by business dependencies. The opportunity for cloud computing as practiced by Amazon and Google will be to what extent these companies are able to drive and participate in the revenues of the companies whose services are hosted on them. Amazon allows ISVs to build and sell subscription applications with Amazon providing the commerce and billing features. Google undoubtedly anticipates a similar mechanism with AppEngine but advertising driven. Companies that use either of these avenues for monetizing their applications will be locked in to a higher degree than any API dependency. The iTunes Application Store is the prototype for this type of business model. Without this interesting commerce relationship between developer and cloud platform, most cloud services are simply a form of automatically provisioned server clustering, which, don’t get me wrong, is very useful, but wont be a platform game changer like some believe.
As I understand Google’s spelling correction and suggest services, the quality of those services relies on the aggregate behavior of others using the site. And Google can probably make use of Feedburner stats, Google Reader subscriptions, links passed around within Gmail, Google’s bookmarks program, etc. to create higher quality search by analyzing the many signals they’re able to aggregate within their network.
This discussion makes me think about a way to make Hugh right, while taking into account the concerns rightly raised about the current state. Right now, cloud computing is dominated by infrastructure players, because the hardware is expensive. It doesn’t have to be. Several articles have been written about the cloud as a service, and abstracting the cloud implementation. If a company delivers this middleware layer between cloud hardware and the developer user along with a SETI@home distributed computing client they can capture network effects while delivering cheap cloud computing service.
Millions of computers, across the internet, serving up pages for whoever needs them, getting paid adsense size amounts for the server time, with this hypothetical company taking a slice of every deal.
google profits best from harvesting the network effect of the web itself, hosting and indexing copies of an ecosystem of webpages produced by commercial and noncommercial partipants. running a homepage, a blog or a social network profile as a consumer or small business is different from running a web presence as an organisation or company. the reliance on data becomes even more critical if you go into intranet applications, databases, workflow etc. consumers can flock from myspace to facebook, or give up one or the other web app. companies which invest into it-infrastructure still need more value propositions to be able to trust into the cloud. maybe they are still waiting for a network effect?
>> Wikipedia doesn’t become any more valuable to me if my neighbor starts using it.
Yes it does. To use your own reductio ad absurdum, if everyone but you stopped using Wikipedia tomorrow, it WOULD become less useful to you. Less users equals less potential editors equals a less accurate encyclopedia.
Otherwise I agree with you.
Brilliant post! Enjoyed it.
Not a very good post.
PageRank gets better with more internet users and usage. AdWords benefits from more usage as well.
Wikipedia clearly gets better with more usage.
And I don’t think Tim meant that you will necessarily win if you harness the law.
PageRank only gets better with more internet contributions. You can’t really call this a network effect IMHO. Of course most of us have become contributers with blogs, link aggregators …
But I think you can only call it a network effect if its bidirectional. If my neighbour links some pages, I profit from it because I’m a Google user. My neighbour doesn’t profit from me using Google, though.
Nick, lots of other people have commented on why you’re wrong about Google, but I’ll add a few more morning thoughts (I was pretty jet lagged when I wrote last night):
The kind of network effects you talk about are only the most obvious. The essence of Web 2.0 is figuring out new angles on how to harness (or harvest) network effects.
Phones have network effects, but a phone that used its call history database to build social network applications for you would be one that saw more deeply into the potential of the network.
Not all network effects need be endogamous (i.e. ones in which the user comes to live on the network.) Google can be thought of as a kind of exogamous network, in which the network effect comes from people who are spread throughout the broader network, which google then harnesses.
If there were no network effect that google could harness, everyone would use robots.txt to tell it to stop spidering their sites. As a number of people point out, Google uses user clickstream data in a number of their applications, but their harvesting of link data IS a network effect insight, contrary to what you claim.
They understood that there was more to the network than just the raw data, that everyone was contributing additional meaning by their pattern of linking.
Anyway, good food for thought.
Meanwhile, on your other point, that there are reasons why cloud computing can tend to monopoly, I just don’t buy it. Yes, there will be big players, just like there are big banks, big phone companies, big hosting providers, but they won’t get monopoly rents unless they figure out how to get lock-in, and I just don’t see that here.
Consider IBM vs MSFT: IBM is twice MSFT’s size, but has half the profits.
Yes, there will be big companies making lots of money in providing cloud computing services. But the point is that these will be ordinary business profits.
Finally, as to your point that sites like Flickr and del.icio.us didn’t become financial powerhouses – seems like a nit. They became dominant in their niche. Flickr blew past sites like ofoto (now kodakgallery.com) and shutterfly by harnessing network effects; just because photobucket blew by them by piggybacking on the bigger network effect implicit in Myspace doesn’t take away from their achievement. What’s more, even today, flickr has dominant share in programmable web access to photos, which will play out in interesting ways over time.
I’m sorry, but I have to disagree with most people in here: network effects was grossly described in early 80’s and many developments have refined it since.
Google’s business model is best expressed through the two-sided market model: they need both users and advertisers to make sense — and users drive ads money up, and ad money allow the quality of the service to be paid for. Credit card, match-makers, newspapers, job markets, conferences, etc. all work the same way: see Rochet et Tirole “Platform Competition in Two-Sided Markets” in Journal of the European Economic Association (2003) for technical details.
http://ideas.repec.org/a/tpr/jeurec/v1y2003i4p990-1029.html
Monopoly is assured if multi-homing (searching on several engines or advertising with several platforms) is expensive or rare; there are reasons for concerns.
To be more accurate, Google has more sides then two: webmasters are another important one, and the efforts towards OAuth let think that Web 2.0 platforms will structure yet another multi-sided platform.
Although it is hard to tell the future, the interactions to expect from Web 2.0 are certainly best described by Brousseau et Pénard “The Economics of Digital Business Models: A Framework for Analyzing the Economics of Platforms” Review of Network Economics (2007).
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1086370
Basically, a company has many ways to control others, but it is unlikely to have more then a monopoly for each solution: Apple for instance, survived because it kept the hardware design (and its latest innovation all come from that).
I would love to say a lot more on the subject, but I am late for a seminar. Don’t hesitate to ask me later today.
Vinne’s right. The value chain for hosting is already quite deep.
[Bob’s Friendly Hosting Company] – tech support -> [Chenai Technical Support]
|
servers
|
[Server Farm Inc.] – IP Transit -> [Level 3 Comms]
|
space+power
|
[Redbus Datacenter]
Etc., etc.
How virtualisation changes the economics of this chain would determine the nature of competition in “The Cloud”. It might be that the technology makes it better to build 1 massively CPU dense water cooled “computer”, rather than a conventional datacentre building.
I’d dispute Tim’s assertion that you can’t make money from commoditised markets. Commoditisation is about how users consume services, not about how you provide them. Innovation, profits and growth happen across the whole value chain.
The more I read between Tim and Nick’s articles and the corresponding comments, the more I realize that cloud adoption-both user and enterprise is still a sort of enigma; that there is no single factor that will tilt the scales towards the cloud’s favor at this point except maybe, the miserable economy and companies operating on survival mode. The realization of the cloud benefits will come later for them.
Best.
alain
I think iTunes/iPod succeeded *in spite* of proprietary file lock-in (presumably here you are referring to their DRM – the file format itself, AAC, is not proprietary). In any case, in my experience, the vast majority of most people’s music libraries are not filled with music bought on the iTMS, so there is no lock-in there. I don’t believe Apple wanted this kind of proprietary-ness, it was forced on them at the time by the music industry, and in the last year or so they have relaxed somewhat. The iTunes/iPod system succeeded so well because it provided a great end-to-end system that reduced the conceptual and actual workload on the user significantly. This was achieved with lots of “small” things – the seamless syncing, the one-price-for-all pricing scheme, the behind-the-scenes database that iTunes built, etc. This is essentially what your point in #6 argues, but I don’t think proprietary file format and resultant lock-in actually has much to do with it.