Why business models for HPC are so very important

You need a sound business model. Not a sound business plan, but a concept of where revenue comes in, and how you will profit from it, and what your costs are, before you should build and sell a product. In the case of state sponsored infrastructure, any model that looks like this:

  1. Build it
  2. ???
  3. Profit!

is a failure waiting to happen. Its not a business model. Actually grant models which for point 2 have “and we will go and find users in the business community for this resource” are also failures. You have to start out with “and we have signed up these users, for this much minimum time and revenue, to this resource, with no single user responsible for more than 25% of our revenue” if you want to have a fighting chance at being successful. Unfortunately, the New Mexico state supercomputer is like this. Build it and they will come, with only one customer, who went away. And left a nasty unpaid bill. A business requires users who will pay for a product or service. If you can’t find users willing AND ABLE to pay for the product or service, you have to alter the product or service in such a way as to find your market.

We did that years ago when we realized that undifferentiated clusters were undifferentiated clusters. Anyone can stack a bunch of web servers and call it a cluster. Many folks do that. Few focus upon the real design issues. Few design to the customers needs. So these days, most of the clusters we build are bespoke … specific to the task at hand. We stay away from the non-differentiated ones. There’s no margin in them, we can’t support running a business building just clusters. Or desktops. We’ll occasionally get requests for desktop quotes … not to the level we thought we’d see, but folks who think they need something genuinely powerful and customized. And then they inevitably compare them with random brand X, with a tiny subset of the spec they gave us, and a lower cost to go with that spec. Building to that spec would make no sense to us … again, insufficient margin on a $1000 unit to support a company. You need volumes of millions of units for this to make sense. This is also why we don’t sell parts. There’s no real value in selling parts. Can’t support a company on this without huge volumes. This is how Amazon is pretty much becoming the single goto market for stuff. They can lower pricing and make money with greater sales volume than their competitors. And they can live on the lower margins thanks to the greater volumes. Just wait until they no longer have meaningful competitors though. So we avoid business that we can’t live on. We try to make sound business decisions. We take carefully focused risk. And sometimes we reap rewards (JackRabbit, DeltaV, siCluster, siFlash). Sometimes not (pegasus). But there is no one backing us up. If we fail, we fail, and our business gets wound up. If we succeed, we grow and hire, and release more products to address markets that we see as being important.

Other rules of business … don’t have a single customer or small group of customers who you are solely dependent upon for your revenue. No more than 25% of your revenue should come from a single customer. Anything else is suicide. Success in business means you have a set of opportunities you can profitably exploit. Ignore the president’s comments (previous post) on someone helping you out … the guy obviously doesn’t know or understand what it takes for a business to succeed. Success requires perseverance, focus, goals, adaptability. It requires you find your customer. You identify them. And you get them to pay you for your service or product. And you take steps to prevent not being paid. We have this as a problem ourselves, with large multi billion dollar corporations and universities wanting us to take anywhere from large to gigantic risks associated with providing them a service, without being able to charge them a premium for that risk. We walk away from crappy business (think onerous terms and conditions), and those who don’t pay us tend to speak with our legal folks. And we require partial prepayment on purchases larger than a certain amount, and full prepayment from various parts of the globe. We’ve learned. Some geographies have people whom are, how shall I put this, reluctant to pay their bills until they need you. Then they are happy to. Since we’ve had so much pain from those areas, everyone there gets to prepay for what we offer. We’ve looked at getting some sort of payment insurance … basically hedging against customer payment. There are ways that this can be done today. But the risk is still on you to a degree. You can sell your accounts receivables (AR) to other companies, and let them work collections if in arrears, though this costs you quite a bit. I’ve looked into hedging, and derivative products to see if we could get something akin to a Credit Default Swap (CDS) against AR (e.g. effectively insuring the credit we grant is in fact safe for us to grant as we have an insurance policy against non-payment). I should also point out that this spawned a really cool idea I don’t see in market (not this, but something else), that would take lots of capital to get started, but would be one helluva nice business (in the multi-billions per year very quickly). Would be interested in working to get that launched. The US government offers Export-Import insurance, though this is targeted at the mega exporters/importers. This is to guarantee payment. It costs, but its not terrible, and you can add it to the bill for the customer. If they prepay, then the cost of insurance, cost of capital can be removed, and darn it, the stuff costs them less money. Go figure. Not sure why the GOP in the House of Representatives opposed the re-upping of the Ex-Im bank authorization. Sometimes our government does not do whats in our best interests. That was stupid of them. It sounds to me like the Encanto business model was broken … they depended heavily upon a single customer for most of their revenue. A non-profit company (bad idea for this) running an HPC center. Build it and they will come. Only one big one came.

Cerelink was working on film/video rendering and other bits. Leveraged NM’s production tax relief. A note at the bottom of their page is curious

Ummm … no. Not correct as a general statement. And not correct in this case either. Though its quite possible that the Encanto folks signed a sole reseller model with Cerelink, which made other business development efforts problematic. It is very rare that we would ever consider doing this ourselves. And we would make sure such agreements have exit clauses. But this would go to my statement about using a non-profit for running/operating the machine. Makes no sense. They won’t make business decisions the way a for-profit would make them. Well, there’s another issue. Large projects which consume large amounts of state financial resources are as often as not political targets. New Mexico is not the wealthiest state in the country. Tax revenues have fallen since the onset of the Great Depression v2.0. And it absorbs capital that politicians want to spend elsewhere. So it has a bullseye on its back. When it lost its customer, and had a $1M outstanding AR that it wouldn’t be able to collect on … yeah … that bullseye started sporting an arrow sticking out of it. Another issue is that Sandia National Labs happens to have a very nice facility, and good set of computing facilities in Albuquerque. Its pretty likely that they offer some access to these for businesses. Short version of this … they probably really didn’t have a good business model (single large customer), and it came back to haunt them. This is not saying this was a bad group, or anything else like that. They were unlucky, and likely, due to their instantiation, they could not unwind the problematic implementation without making substantial changes. Which they were probably enjoined from making (non-profit). I feel bad for them. Don’t get me wrong. But it would be a stretch to imagine them thriving. I’d love to be proven wrong. University level supercomputing centers are common. State/regional level centers are subject to politics as well as difficult business choices. Some have thrived, even in the face of painful changes (PSC). Others, not so much. Part of it is getting the right leadership to make the hard choices, the right business development team to sell the resources, and the right local/regional climate to encourage usage. I do hope a formula is found to enable them to continue, but it probably will be in pieces or as a private concern. The latter would be more likely to make better choices.