Wednesday, January 23, 2013

Google Fiber "Is Not a Hobby"

Patrick Pichette Google CFO, said on Google’s recent earnings call that Google Fiber is not a hobby. That could mean lots of things, so ISPs should not necessarily make assumptions about what that statement means.

“We really think that we should be making good business with this opportunity and we are going to continue to look at the possibility of expanding, but right now we just got to nail because we are in the early days,” said Pichette.

“Not a hobby” could mean that Google does not intend to lose money on the venture, and is not simply spending money on a “hero” initiative that is not intended to directly sustain itself over the long term.

Contrast that with Apple’s statements some years ago that Apple TV was, in fact, a hobby, implying that commercial impact was not expected.

But “not a hobby” would unsettle other large ISPs much more if it implied Google was seriously entertaining the notion of becoming an ISP in its own right, on a bigger scale.

Those sorts of fears have been expressed in the past, about Google “becoming a telco.” But Google has become a handset supplier, on a limited scale. Google Voice does earn communications revenue. Google Docs does compete with Microsoft’s “Office” suite.

Google does operate a large global backbone network. Likewise, there are, from time to time, discussions of whether Google (or other big application providers) want to become mobile service providers.

And even at the recent Pacific Telecommunications Council meeting, at least a few attendees I spoke with did express concern that Google might in fact be considering a wider and more significant entry into either the local access or backbone transport markets.

In other words, there remains considerable unease about what Google might decide to do, in the communications business.

The concern might be overblown. But there is no doubt about what Google would prefer, and that is higher speeds for most end users and more investment in access networks by the leading ISPs to enable that.

Google’s challenge to leading ISPs is clear enough.


In a highly-competitive market, the low-cost provider tends, over time, to win. That is true with respect to large tier one telcos competing with large tier one cable operators, for example. You might argue that cable gains in high-speed access and fixed network market share provide a clear example.

Some now would argue that ISPs--both fixed network and mobile ISPs--need to match Google’s own costs, on a gigabyte per cents or gigabyte per dollar basis. How well that can be done, and if it can be done, is the question.

But Google has affected service provider thinking before. Remember several years ago when executives started to routinely say they had to “innovate at Google speed?” Doubtless, most would say no telco really is able to innovate that fast. But it might be argued that service providers do now innovate faster than before.

So it might not be unreasonable to argue that if Google continues to demonstrate new cost models for very high speed access, that service providers will respond.

Shifting to costs equivalent to Google’s costs might be a daunting prospect, but less daunting than what could happen if legacy revenue streams erode faster than new revenue replacements can be created.

It is one thing to argue that telcos, for example, need to incrementally reduce current operating costs. But that argue also hinges on a crucial assumption, namely that current revenue continues to grow on a relatively stable basis, while revenue losses from legacy products do not accelerate in a destabilizing way.

Some might argue that the risk of unexpected revenue trend deterioration is greater than most now assume. In that case, one way or the other, service providers will have to make further adjustments. That is one reason why Google hints that it might expand the Google Fiber program.

No comments:

Costs of Creating Machine Learning Models is Up Sharply

With the caveat that we must be careful about making linear extrapolations into the future, training costs of state-of-the-art AI models hav...