Advanced Computing in the Age of AI | Saturday, January 28, 2023

Rackspace Hires Bankers To Explore Its Options 

The potent combination of OpenStack cloud computing software, Open Compute Project minimalist hardware designed explicitly for hyperscale datacenters, and fanatical support may not be enough to allow for Rackspace Hosting to standalone and compete with its much larger rivals in the public cloud.

Rumors that Rackspace was exploring its options broke last week in the wake of the OpenStack Summit in Atlanta, which is still dominated by Rackers and ex-Rackers, and followed the hosting and cloud company’s reporting of its first quarter financial results. The news comes three months after Rackspace CEO Lanham Napier surprisingly decided to step down from the company, and perhaps it sheds some light on that decision.

According to a report from Bloomberg, banker Morgan Stanley has been helping Rackspace weigh its options. Rackspace has spent the past four years trying to transform its hosting business into a public cloud computing facility, and it has considerably less capital than rivals Amazon Web Services, Microsoft Azure, and Google Cloud Platform to build out infrastructure on a the global scale necessary to attract the largest customers and get the best economies of scale for its own business and therefore compete on price. Rackspace has north of 106,000 servers under management, the vast majority of them built by Supermicro. At this point, AWS, Microsoft, and Google all have at least ten times as much capacity in their server fleets and even IBM has woken up and spent a rumored $2.2 billion to buy SoftLayer and has pumped another $1.2 billion into it this year to get its server fleet up to several thousand machines running in 40 datacenters by the end of the year.

Rackspace simply cannot play in that league and has been the subject of takeover speculation ever since Google and Microsoft got serious about their public clouds a few years back. This is a bit of an exaggeration, but only by a little bit: Every time AWS cuts prices, its stock price goes up and Rackspace’s goes down and every time Rackspace cuts prices its own stock price goes down. And when it doesn’t cut prices to compete, Wall Street is similarly unhappy. Rackspace is also the largest publicly traded cloud and hosting provider, and unlike its rivals, its numbers are right out there. No one knows for sure what revenues and profits AWS, Microsoft, and Google are getting from their cloud and hosting businesses, and ditto for smaller players and telcos who also sell capacity by the month or by the hours. The pressure from Wall Street has Rackspace looking at its options, and maybe had the company a chance to do it all over again, it would have stayed private.

“In recent months, Rackspace has been approached by multiple parties who have expressed interest in a strategic relationship with Rackspace, ranging from partnership to acquisition,” the company said in a filing with the US Securities and Exchange Commission late last week, confirming the rumors. “Our board decided to hire Morgan Stanley to evaluate the inbound strategic proposals, and to explore any other alternatives which could advance Rackspace's long-term strategy. No decision has been made and there can be no assurance that the board’s review process will result in any partnership or transaction being entered into or consummated. The company has not set a timetable for completion of this process and does not intend to discuss or disclose further developments with respect to this process unless and until the board approves a specific partnership or transaction.”

Rackspace has been trying to position itself not as the low-cost provider of cloudy infrastructure, but rather as a provider that offers enterprise-grade technical support and other add on services including networking, database, and others services that are similar to those offered by the big three.

Last week, during the OpenStack Summit and ahead of the rumors about the hiring of Morgan Stanley, Rackspace CTO John Engates made the argument that all of this focus on pricing on the public cloud was proceeding from a false premise that pricing was the important factor.

“The vendors leading the race to the bottom on the price of raw infrastructure would love for you to believe all cloud services are identical,” wrote Engates. “The very definition of commodity is that the ‘importance of factors other than price are diminished.’ These providers believe the lowest infrastructure price will win the most market share and that customers will choose a provider based solely on that price. Their actions implicitly acknowledge that they’re selling an undifferentiated commodity and that its price is their best lever to win business. This is a dangerous and expensive way for developers and businesses to make critical cloud buying decisions. Success with the cloud requires much more than just renting access to cheap infrastructure. The real differences among clouds are the services wrapped around them and how those services solve your business’s IT needs.”

It is hard to argue that point, but then again, for very large companies with sophisticated IT skills, some of these services are perhaps superfluous. Anything that Rackspace can learn about hyperscale computing Fidelity Investments (which owns a 15 percent stake in the cloud and hosting company) or Goldman Sachs, just to name two prominent enterprise members of the Open Compute Project, can learn for themselves or from their peers. The appeal of cloud is indeed skills and services, but the appeal is really for smaller companies that cannot afford to foster those capabilities internally.

Own The Base, Rent The Peak

For large enterprises, the chant is much the same: Build buy your own private cloud to support your base workloads and rent the workload peaks from public cloud providers. This message was said time and again at the user sessions at the OpenStack Summit last week. Because of its compatibility with OpenStack, Rackspace should be able to garner some enterprise business. But Google has nine regions globally in its public cloud, Microsoft has sixteen, Amazon has ten, and Rackspace has only five (Dallas and Chicago in the United States, London in Europe, plus Hong Kong and Sydney in Asia/Pacific.) Having OpenStack is important, but having global scale is equally important.

And as Microsoft pointed out in discussing its Azure cloud last week, having hyperscale capacity is necessary. It is not enough to be near every major business center in the world, but you have to have hundreds of thousands of cores in each region so you can get the cheapest computing and storage to appeal to a large number of enterprises of varying sizes and workloads. The public cloud is a volume game every bit as much as the chip business has always been, and it is only by lowering the costs of datacenters to their absolute minimum can any player hope to survive unless they have some sort of captive customer base that will pay a premium. Clouds for running mainframe and other proprietary operating systems or even Unix workloads will no doubt be more costly, as physical systems running their platforms are today. But these are corner cases, not the volume of the market, which is dominated by X86 systems running Linux or Windows workloads.

Rackspace has played its hand as brilliantly as it could, under the circumstances, and has unlike many hosting companies, made the transition from hosting to utility-style cloud computing. It helped found the OpenStack community with NASA and shepherded it, with a certain amount of grousing, into an independent foundation that will see this cloud controller perhaps become a kind of Linux for cloud computing, wrapping around all operating systems and hypervisors to become a universal control plane for compute, storage, and networking in enterprise datacenters as well as in the facilities of the telecommunications giants, who are tired of paying premiums for specialized networking gear.

The business that Rackspace has been able to build certainly would have appeal to a number of companies that are desperate to get their public cloud businesses up to scale quickly.

Dell lost interest in operating public clouds when it saw the investment necessary and it content to sell general purpose and custom-built gear to cloud builders. Any one of the telecommunication companies that want to scale up to compete – Verizon/Terremark, AT&T, CenturyLink/Savvis, TimeWarner/NaviSite come immediately to mind – might be interested in acquiring Rackspace. Hewlett-Packard still has public cloud aspirations with its recent $1 billion Helion push and would have been wise to have bought Rackspace several years ago rather than waste $10.5 billion on Autonomy, a provider of analytics and data management tools that has caused it more grief than joy.

IBM may have bought recently SoftLayer but it has yet to get the kind of scale it needs to compete with the likes of AWS, Microsoft, and Google. Merging together SoftLayer and Rackspace would certainly get IBM closer to its goal of generating $7 billion in cloud revenues as it exits 2015, but the resulting business would not necessarily be much more profitable than the System x X86 server business that IBM is trying to exit. SoftLayer has about 25,000 customers using both cloudy and bare metal capacity, and Rackspace has several hundred thousand customers across its cloud and hosting businesses, with just under a third of its revenues coming from the cloud. Rackspace’s cloud business is only growing a little bit faster than its hosting business, and both have slowed in recent quarters.

The real issue is that Rackspace, like most companies on the stock market, is very expensive indeed given the relatively skinny profits that this hosting and cloud business throws off. In the trailing four quarters, Rackspace has brought in $1.59 billion in revenues and is growing in the middle double-digits at the moment. But it has spent $414.9 million on datacenter build outs, hardware, and software development over that same time, and has only brought $84.9 million to the bottom line over those four quarters.

That is a mere 5.3 percent of revenues, and that is more like the kind of margins a wholesale distributor of IT hardware and software sees than Rackspace would probably like. This makes sense if you think about it. Cloud and hosting are just another form of IT distribution, with lots of competition keeping revenues and profits down. It also makes sense that the scale of the customer base and the asset base matters in such a business, too.

Regardless of the competitive pressures, Rackspace would be a fairly expensive acquisition for anyone to make. The hubbub around OpenStack and the rumors that Rackspace might be on the block pushed its stock price up by 37.7 percent last week, giving the company a market capitalization of $5.1 billion. Last summer, before AWS. Google, and Microsoft kicked off a cloud price war, Rackspace was worth twice as much as this recent high. Even given that, it is trading at more than three times current revenues with skinny margins, and only IBM, HP, or someone who wants to build their base of cloud customers quickly and then scale up fast from there would be willing to pay what Wall Street would no doubt want for Rackspace. The real question – and one that has been on a lot of minds in the past year – is what happens to Rackspace if no one buys it? OpenStack has a large enough community to keep progressing, and Rackspace has plenty of customers. So it is not going to evaporate.

One Response to Rackspace Hires Bankers To Explore Its Options

  1. Phil Hughes says:

    “Open Compute” is designed for data centers using free cooling. This implies a low power density thus a very large building. It is dangerous illusion for companies like Rackspace as they do not have the financial muscle to build big enough data centers to profit from it. To compete, they need to add capacity much faster than every 2-3 years. ebay on the other hand espouses high density. With each generation they can ad more capacity in an existing space.Air cooling will take you to 30-40kW per rack and liquid to 200kW. What is more, with a good liquid cooling architecture, just about any building with access to power can be adapted to server as a data center.

Add a Comment