Should I Stay or Should I Go


Tl;dr I’m now writing posts that follow up on posts I didn’t actually get around to writing

Last summer I outlined a playlist of posts based on Clash songs. But after borrowing some great titles, my motivation faltered.

Should I Stay or Should I Go was triggered by the cloud repatriation debate. Career Opportunities detailed how to compete with the hyperclouds. Straight to Hell examined CNCF’s “architecture”.

Now fast forward to summer 2022. The cloud repatriation debate has returned, albeit in different and perhaps unrecognizable form. It gives me an excuse to dust off this title.

Two, Four, Six, Eight – Repatriate!!!

“The Cost of Cloud, a Trillion Dollar Paradox” instigated the cloud repatriation debate. This manifesto, from the more fungible and mutable wing of venture capital firm A16Z, contended mature software companies have a financial imperative to move workloads off the cloud, and “repatriate” them to their own data centers.

The “paradox” of the title acknowledges the flip side to the hoary VC platitude that the cloud makes it easier than ever to start startups: if you rent your computing infrastructure, and the company grows, your bills also grow.

Those bigger cloud bills apparently came as a shock to some VCs. Their big takeaway is those bills are “excess costs” that negatively impact the valuations of bigger companies running in the cloud, particularly mature SaaS companies.

The authors believe there is an inflection point when software companies need to shift their priorities from growth and optionality to cost containment über alles. (That inflection point seems indistinguishable from the moment a software company is acquired by private equity.)

The underlying assumptions are highly stylized, but the claim was you can halve your cloud costs by running your own infrastructure. They asserted a sample of 50 software companies were forfeiting $100 billion in market capitalization by using the cloud.

A less charitable summary might be “we’re kind of bummed the cloud providers have better margins than our portfolio companies”.

The repatriation post triggered a vigorous debate. Many pointed out that repatriation is far more than a financial issue, and complex tradeoffs abound. Does every company have the chops to replicate what the trillion-dollar hyperclouds do? Dropbox made for a poor repatriation poster child, and was arguably a cautionary tale of allowing one’s franchise to atrophy while the company was focused on repatriating their plumbing. The debate eventually petered out (even as the cloud kept right on growing).

“This is our big short right now”

But the repatriation debate was quietly revived last week. Preeminent short seller Jim Chanos announced he’s shorting the data center Real Estate Investment Trust (REIT) stocks, which lease out data center space.

This trade is long cloud and short cloud repatriation. The data center REITs should be prime beneficiaries of cloud repatriation if it is happening.

Chanos, who famously shorted Enron and Wirecard (and more recently IBM for their financial shenanigans), believes the likes of Digital Realty Trust and Equinix are both overvalued generally and likely to lose ground to AWS, Azure and Google Cloud.

He argues the data center REITs will lose to the hyperclouds in two ways. Directly, the hyperclouds prefer to build their own data centers to renting space from the REITs (they do throw down some CAPEX). And indirectly, they compete for the same customers: will those customers prefer to run in the cloud or operate their own infrastructure in a co-location data center rented from the REITs?

Chanos says:

“The story is that although the cloud is growing, the cloud is their enemy, not their business. Value is accruing to the cloud companies, not the bricks-and-mortar legacy data centres.”

“Their three biggest customers are becoming their biggest competitors. And when your biggest competitors are three of the most vicious competitors in the world then you have a problem.”

So now we have a hard money bet against cloud repatriation to watch.

The Return on Repatriation

The nice, big, suspiciously round number of $100 billion in “suppressed” (they don’t quite say “stolen”) valuation now looks a little quaint, as it was estimated at peak 2021 startup valuations, aka the Altimeter Apex, the Coatue Crest, the Softbank Summit, the Tiger Top, and the Zenith of Zero Interest Rates (also aka the ZZ Top).

Since then, the financial North Star for startups has shifted from growth to profitability. That boosts the case for doing things that drop more dollars to the bottom line.

On the other hand, the valuation multiple for those profit dollars has declined with the overall market. (As you may have heard, software valuations have come down a tad since, including the profit-based metrics, not just the bubblicious revenue-based metrics).

If we look at the financial case study from the repatriation post, Datadog, their valuation as a multiple of (current year) gross profit dollars is down over 25% from last summer’s number. So the return on repatriation has declined, paradoxically, even as companies are being told to get profitable.

Career Opportunities

The most disappointing – but not surprising – thing about the repatriation manifesto is it resorts to the lens of valuation, seemingly the only perspective available in these times of endemic software financialization. Implementation is just assumed.

It is really just a plea for ever cheaper cloud pricing. The lenses of competition and disruption are much more interesting ways to explore that future.

Looking at competitive threats to the hyperclouds in the coming years, it is hard to imagine new entrants with similar scale and scope (even Google struggles to keep up with AWS and Azure). Meaningful competition for the hyperclouds will not come from hyperclouds, new or existing.

Challengers are much more likely to pursue specific segments and compete asymmetrically for specific workloads. With hundreds of billions of revenue in play, those niches can become quite large. I can imagine a half dozen or so different categories – most of which are beyond nascent today – emerging that could collectively pry material business away from the hyperclouds.

Like Chanos, I’m skeptical repatriation will propel the data center companies to be among the winners in the hypercloud shadows (which would be a great title for a post, even if it isn’t a Clash song). I have the best of intentions to write more on this topic (unless I am Overpowered by Funk).

2 responses

  1. Nice take, thank you for writing it. I am super curious about those half dozen competitors. Could you name some examples? I can only come up with server less edge computing related things and Cloudflare.

  2. I will try to write up this summer. Edge not on my list. Edge networks definitely.

Get Updates By Email

%d bloggers like this: