No it's not pretty obvious. Reducing the rental cost of sold servers means directly reducing revenue, that sounds like madness to me.

The saner alternative would be to _increase_ the cost of the newer servers, but as they point out that could result in them being less competitive.

The real solution would be to devise some way where happy customers just retain their current servers, and new customers can get the latest hardware for the lowest price.

For some reason the old customers do enjoy upgrading their machines though, so this lands them in a tough spot. They can't have the best of both worlds, both the lowest prices and the freedom contracts. So they're going to take their time finding out what their strategy on this is going to be.

Ideally, the old servers would have (1) paid for themselves, and (2) generated a reasonable amount of profit, before they became so obsolete that the market demands drastically reduced prices. A high revenue stream from that server after that point should never have been part of OVH's business model in the first place.

A host that expects customers to keep paying the same monthly amount for a 2-year-old server is no different from a rotisserie that expects customers to pay the same price for last week's leftover chicken. If the rotisserie's business model depends on selling stale poultry at a higher price than what the market would bear, it is clearly unsustainable and deserves to have its revenue reduced.

Yes, it's obvious enough that the author specifically calls it out:

> So yes if it's a turnover issue, we could offer the old server at the same price as the new one as an incentive.

It might or might not be correct, but it's an extremely obvious idea.

> For some reason the old customers do enjoy upgrading their machines though

This is largely their own fault; they make doing this relatively cheap and easy, and are thus punished for it.

I guess the price is already quite low and at some point the cost of the power, space, etc would be higher than the rental price. Do you want a new new server or a server that is 2+ years old and it's reaching its end of life?

I met people using OVH 4 years ago and they told me support wasn't great (I can't confirm this at all!) and that it was usually faster to order a new server and migrate than to wait until a faulty disk was replaced.

If people get good doing that kind of thing, it makes sense it's a problem for OVH (+150k servers): instead of waiting until a server is over 3 years old to be retired, just order a new one every year and benefit of the newest hardware. It's the same price, isn't it?

>I met people using OVH 4 years ago and they told me support wasn't great (I can't confirm this at all!) and that it was usually faster to order a new server and migrate than to wait until a faulty disk was replaced.

This is still pretty much true, especially under their best effort support. I have their higher SLA (<2hour) on several large servers and it still takes them a day or more to get back to me and then a day or more to get the hardware replaced. And now they've introduced a new iteration of servers (which, IMO, was designed to fuck the customers over...they've figured out how to oversell the rack at scale) that was ~20% the cost of the 2012 models of the same servers.

The issue is that they need a certain amount of revenue per box.

So, they buy a new Core i3 box with 8GB of RAM and they want to get $40 over the next two years to cover the investment. If they have to reduce its price to $30 after 6 months and further as time goes on, they don't cover the investment.

Now, they could price the box at $60 and lower it over time, but then you lose signups since it's a higher price (even if the average price over two years would be the same). You can say: well, have the older servers whose price has already been reduced available for sale as well. That complicates the lineup and is hard to manage. Unlike the new servers, the old servers only become available when someone gives one up which makes the supply unpredictable. With a new server, they can always buy one. Making matters worse, if the old ones have a reduced price, people are more likely to stay with them.

A decent part of this is that a lot of people are probably "needlessly" upgrading. By that, I mean that a lot of people probably aren't upgrading because the new server will make their usage better, but because they have a sense that they should be getting the top that their dollar will get. They were running perfectly fine on the old box, but why not try the new?

Ultimately, this is why VPSs are a lot easier. You can lower prices as time goes on in a reasonable manner. Someone wants the 1.7GB/1GHz 2007 box? That's a fraction of a new piece of hardware we can deploy. On the contrary, if someone wants a two year old model at OVH, they can't deploy a new two year old model: it's only available if someone gives it up. So, it becomes hard to capture the cheap-end of the market unless those signups keep paying for that box.

> Now, they could price the box at $60 and lower it over time, but then you lose signups since it's a higher price

That feels like the crux.

A friend claims Warren Buffett said something along the lines of "it doesn't matter how many insurance policies you sell, if you sell them at the wrong price".

If your customers are losing you money by paying at your published prices, then acquiring more customers isn't a sensible thing to do. Lock the customers in, or price higher at the start. Why chase customers who aren't making you money?

Yes that was my thought, have a price function that falls over time you have held the server, perhaps starting a little higher than the current one. Its a kind of amortised up front fee in some ways.

Alternatively, start a PAAS business to use the servers people dont want to buy, although I guess the VPS business does this to some extent.

This is not what they want because they are trying to price their new hardware competitively by milking their older customers. If they would reduce the profit of their cash cows they can not invest in reducing the prices of the hardware that attracts new customers.

Another would be to tie the customers to their current servers through other means which is what they're trying.

For example, one of the things they're trying now is to link the number of IP addresses you can have attached to the cheaper servers with how long you have been running it. Which is pretty clever if you think about it. Who are the savvy customers switching to the new offer the day it comes out? Hosters who need mutliple IPs for their users.

Their profit margins are likely not huge as it is, and it costs them just as much (probably more, given changes in power usage in servers over the last few years) to run a Core 2-arch server as an Ivy Bridge one.

It all depends on the margins, maybe if they lower the old ones they can't afford these total lower costs