Costco's International Growth Opportunity

In 2020, RBC released a research paper estimating COST could grow its US store base by over 400 units. The population size, density, and median income of many regions of the US are still under-penetrated by Costco. Over the last years, COST has added 15 US stores annually. At this rate, COST has another ~30-years of potential unit growth in its home market.

But what about International growth?

This is the greater untapped opportunity for COST.

Over the last 16 years, international store units and sales per sqft has grown 10% and 2.65%, respectively. This has led to 560bps faster annual revenue growth than the US.

Source: Costco filings, IP
Source: Costco filings, IP

Last week, we published two interviews with two former Costco executives, who each spent over 30 years at COST, to discuss the international growth opportunity. One of the executives was responsible for scouting new markets, opening multiple APAC countries and more recently worked for Sam's in China.

We also published our analysis estimating COST's potential international store count growth for decades to come.

We share a few snippets of last week’s work below. The full interviews and analysis are only available to members at

COST International: 650+ Store Opportunity

RBC use five major factors to estimate new store openings:

  1. Population size
  2. Median income
  3. Population density
  4. Competition
  5. Store cannibalisation

GDP per capita per store is a key metric: population density combined with earning power is the starting point. Additionally, Costco will consider its existing store presence and competition from other club operators or grocers to drives store opening decision making.

Although we haven’t got specific store location data, we can make a rough estimate of potential international stores based on public information of two different countries: Canada and the UK.

Canada and the UK are on two different ends of Costco’s performance. Both countries have very different population densities, GDP, and store counts. Canada has 107 stores in a country with $1.8 trillion in GDP, $17bn GDP per store, and the UK has only 29 stores or £74bn GDP per store. GDP per capita per store is over double in the UK vs Canada.

Canada is still slowly growing. The UK hasn't added a unit in 5 years. Both are mature, saturated markets that we can use to forecast the potential mature store base for China, Korea and Taiwan, Australia, LATAM, and the Nordics.

Source: COST, IP
Source: COST, IP

We believe Costco could open up to 300 stores in China and a minimum of 650 international stores.

You can see the full international store breakdown by country in our full write-up.

We’ve excluded any growth from Southern Europe as COST's growth has been slow in Spain and France. It seems European governments are hesitant to let an iconic American brand shape its community and put local mom-and-pops out of business:

The biggest challenge in those countries is government rules restricting size limits on retail facilities…Still, the biggest challenge is government restrictions, fighting against behemoths like Costco, and having it affect the small mom-and-pop businesses you see in Italy and France. That's a big part of their economy, it's very protected, and it slows the development down in those countries. - Former COST Director, China

If COST has a 650+ store international opportunity, and open an average of 24 stores per year, this provides 40-years of unit growth, a ~2% CAGR.

Store unit growth plus increasing sales per square-feet from maturing stores will continue to power high-single digit revenue growth for decades to come.

Also, margins are higher abroad for structural reasons. In the US, ~30% of total labour cost is healthcare spend, which isn’t paid in other countries with nationalised healthcare systems.

[US employee healthcare cost] runs between 40% and 30% depending on certain areas, but yes, it is about a third of their cost, and it's a big hit. It's a huge hit. In Asia, we didn't have to deal with it that much. In each country, there's a cost for national healthcare. In some countries like Taiwan and Japan, the company pays a third of that cost, but that cost is fractional compared to the US. The US has a huge hit, and it's not a big hit against the company in Asia. In China, there isn't any hit; the Chinese government provides everything. - Former COST Director, China

We don’t need to explain why COST is a phenomenal business.

The question is more: can management maintain the culture and execute abroad?

The model is replicable, but every market is slightly different. And there are signs of Jelinek slightly tweaking Sinegal’s philosophy.

Although Costco isn’t cheap at 30x FY23 earnings, a membership fee royalty that grows high single-digit for 40-years is interesting.

The full interviews and write-up is available for members at

GCP, BigQuery, Databricks and Snowflake

A Former Google Cloud VP believes BigQuery is not enough alone for GCP to sign enterprise clients.

Google is not getting entry into large enterprises yet. It's ugly, and we were one of the first ones to come in, then they brought in TK, next phase. So Amit Singh and the team didn't quite get enterprise well when they tried to do it. I came in under Amit Singh's regime, but then Diane came, and it started to move more towards enterprise. It's still a distant number three in enterprise. People are going to pick AWS and Azure first, and what happens then is you're not going to pick BigQuery simply because GCP is last. You can argue that BigQuery has nothing to do with GCP but yes, it has everything to do with it, it's going to run on GCP, but you can get just BigQuery if you're looking at it from their perspective. But nobody's going to do that. No CIO is going to go out and buy it for any company that does a billion or over in revenue. Mid-market might, but mid-market has many other problems to deal with than to deal with multiple vendors. - Former GCP VP

This point was even more interesting:

Maybe there's a bit of bias creeping in here, but for one, I don't think there’s any product better than GCP. AWS is a very close second from a pure product perspective. But the minute you bring the packaging, deployment, ease of use, all of that stuff in, security – Google's great for security as well – there's no chance. Google stands no chance of winning at enterprise. AWS has got that figured out. - Former GCP VP

This reminds us of an old Marc Andreessen interview that Modest Proposal recently tweeted.

Distribution trumps product.

And both MSFT and AMZN have the scale and enterprise sales team to distribute for all layers of cloud computing.

The executive also believes Databricks and Snowflake are the most interesting software companies today. The runway is significant.

After all the changes we made at a large bank, I'm going to put a number out that we managed to get 5% of it off to the cloud. There's a whole lot of data and everything, and we had our own data lakes and lakes upon lakes and a bunch of different things, so consolidation has to happen even for people who've built all this stuff on-premise. I'm not sure what numbers Databricks and Snowflake are now projecting in the cloud market, but I would say if they're putting a B behind it, they're wrong. It probably needs more of a T behind it; it’s in the trillions. - Former GCP VP

AWS Dialogue

We hosted a group call with 5 professional investors on AWS. One investor framed both sides of the AWS thesis:

There are two ways to look at this market. One is that it is a commodity, renting storage and servers and some enterprises do look at these products like that; we want to commoditize this as much as possible, containerize everything, use zero higher-level services and everything goes to the cheapest provider. I’m becoming increasingly convinced that there is going to be technology out there that helps them actually do that, and it’s going to be technically feasible. I haven’t found great examples of that yet, but probably Netflix can do it.
The other way to look at this is that this is a dev tool, at the end of the day. The value proposition is not cost; the value proposition is agility. You want to use whatever empowers your devs to get the most product out the door as possible. I think that is the right way to look at the market, in my view, which is why I’m so optimistic on this space. This enables dev productivity so you want to enable them to use higher-level services and there is going to be no one that does that better than major public cloud providers.

Another investor also shared fears of industry specialists taking share from hyperscalers in the future:

What worries me about the hyperscalers, in general, is stuff like commoditization of their services. Look at AWS; they already have 200, 300 services. It is getting so complex and I worry about best of breed solutions for some industry verticals, where you have AWS or the hyperscalers more as a utility function, just offering compute and storage. Then you have someone really good on healthcare, where it can give you the whole suite of services, from compute, to machine learning, to databases...there are cases in other industries. Payments in the restaurant, for example, where you have a solution like Toast, who are offering that entire full-suite solution for the small restaurants. I just think it is conceptually hard to think that AWS or Azure would be great in every single verticaL

Over the next few weeks, we plan to walk through how an enterprise customer ports workloads via Kubernetes to the cloud and explore the risk of commoditisation.

Cogent's Sprint Fiber Deal

Looks like Cogent's owner-operator CEO has architected a way for the company to enter the wavelength and dark fiber business at negative cost...

Cogent buys for $1 and T-Mobile buys transit for $700 million; they don’t plan to use any of it. That part is solid. The hard part to unpack is, currently, the company is burning through $300 million a year, of cash. The question is, can Dave and his team transform that asset? This is like Dave’s playbook for acquisition. Can he strip it for its parts and reconfigure the network in such a way that you have some additional growth opportunities for wavelength and dark fiber and then can you get some cost efficiencies of the core Cogent business, which sounds as if he can? If you’re curious, the way he has laid it out is, something to the tune of 225 million of network costs between the two companies, just by pulling off net circuits from Sprint, onto the Cogent network and combining the networks internationally to reduce some of the international costs that Sprint has. Then getting off the Lumen long-haul IRU that Cogent has been on for a long time saves another 15 to 20 million of maintenance spending...The way he describes it, he’s getting access to the wavelength market and the dark fiber market, with negative cost.