[UFO Chicago] Is google too big to fail?

[Neil R. Ormos]

Brian Sobolak wrote:

> We often apply this moniker of TBTF to banks,
> under the theory that without access to money or
> credit the economy would collapse.  Is the same
> true for data?

One problem with the notion of TBTF--particularly
but not exclusively as applied to banks--is that
it sets up a false dichotomy between complete
failure and a particular type of
government-sponsored rescue intervention that
leaves the operation largely intact, without
addressing what specific services the entity
provides that are essential to the function of the
economy or society and whether those can be
preserved through some other form of relief, e.g.
by decoupling them from the larger organism.

> But are Google or Amazon now too big to fail?
> With so much data and services routed through
> their servers, is that true for "cloud
> providers"?  If Amazon goes bankrupt, do you get
> the data hosted by them on behalf of others
> (Netflix, Instagram, countless others)?  Is
> Netflix critical like banks or transport?

> They are private companies and one will someday
> fail.  Even more, we should prepare for them to
> fail.

> Obviously by my tone I think they are TBTF. But
> what am I not considering?

To address this question, I think you need to
decompose the possible failure modes and define
what you really mean by TBTF. As to the meaning of
TBTF, does "failure" of one of these entities shut
down the world's or this nation's economy?  Or
perhaps the result is something less drastic. Does
failure create widespread and long-term economic
disruption that cannot be repaired without
extensive government intervention? Or does failure
simply create a temporary disruption--a blip--that
can eventually be worked around by the normal
adaptation of businesses and markets.

I tend to believe the failure of either Google or
Amazon's computing service businesses (as distinct
from Amazon's store and Google's
direct-to-consumer application services) would be
in the blip category, at least at the macroscopic
level. A few Web-based service companies that rely
on Amazon or Google as computing service providers
would probably flame out because there would not
be enough spare capacity elsewhere to immediately
handle their workload, and enough consumers would
defect to rebalance the competitive landscape to
the incumbent's permanent detriment.

As to failure modes, consider what might happen to
cause a failure and how complete that failure
would be.

If it's simply the failure of the business model
of one of these company writ large, it's very
unlikely that someone will come in and press the
"Emergency Stop All" button.  For one thing,
pieces of each business are surely independently
viable.  For another, a significant fraction of
the value of each business is in intagibles and
would be completely destroyed in a n4ked
liquidation.  Also, the physical assets would be
hard to liquidate. Creditors are better off with
half a loaf than none, so the computing service
units are likely to continue in operation while
they're sold or spun out as independent
businesses, perhaps with adjusted pricing.

If the problem is something more
fundamental--e.g., an attack on physical
facilities--that renders one company completely
inoperative, while leaving competitors intact
(which I think is far-fetched), business customers
would eventually migrate to other services, albeit
incurring signficant inconvenience, conversion
cost, and greatly increased short-term service
costs.

So IMO, this boils down to the question, "What
services do these vendors provide that are
reasonably essential for commerce and for which
there are no real competitors--at all, or at a
scale that can accept large influxes of commercial
loads, or for which replacement requires the
vendor's cooperation?" For Google, I would suggest
those services are search, ad syndication, and to
a smaller extent, the fee-based Gmail/Apps service
(and only because some governments have adopted
it).

Information-based business customers of Amazon and
Google have to re-configure their processes every
couple of years or so (if they don't do it
continuously) in any case, so I expect mitigating
this risk is being done in the ordinary course by
existing personnel at tolerable cost.

After Amazon outages in the last couple of years
that disrupted some high-visibility web-based
services, I'd be surprised if their big users
hadn't already designed business-continuity
solutions that allow their workloads to be
redistributed to other vendors.

> Maybe the flameout won't be an accounting
> scandal but a data theft ala Target that causes
> a failure.  Then what?

How do you imagine that would cause a failure?

While not economy-breaking, I think it's worth
considering the risks to individual consumers of
reliance on services they get directly from Google
or Amazon.  Most consumers neither understand
these risks nor know how to mitigate them, and
since the consumer lacks a planned response, the
event is very disruptive.

Rather than the scenario where Google or Amazon
turn out the lights, I'm more worried about their
withdrawal of services on which people rely,
breakage of APIs that destroy existing investments
in devices and software, and the annulment of
long-standing assurances to consumers about how
these companies will use the consumer's data, all
at the company's absolute whim, and perhaps
accelerated by changes in business conditions or
management.

--Neil