The chant for years and years from hyperconverged storage pioneer Nutanix has been “Ban the SAN.” But going forward, as the upstart is moving closer to its initial public offering, Nutanix wants to do much more. With two recent acquisitions, of PernixData and Calm.io, Nutanix is trying to transform itself into a proper, self-contained platform.
It will take either more acquisitions or lots more development to accomplish this goal. So Nutanix is by no means done. PernixData was equally ambitious in flash-accelerated and all-flash storage, and seems to have overextended itself as it invested in an effort to bring an all-flash array product to market and apparently ran out of cash. (The play-by-play of the PernixData efforts and rumors of a possible acquisition by Nutanix have been covered by our sister publication, The Register.)
With Nutanix co-founder and CEO Dheeraj Pandey and PernixData CEO and founder Poojan Kumar being colleagues at Oracle a decade ago, a discussion between the two seems inevitable, particularly once PernixData had run out of cash to do its all-flash development, as has been rumored. PernixData was acquisition, and Nutanix, which had delayed its IPO and which is increasing its losses even as it grows its revenues, needs a new – and better – story to tell Wall Street if it is to get its IPO on track. Nutanix itself admits that the hyperconverged storage market is crowded, with over 37 players, many of which have soaked up lots of venture capital to get off the ground to chase this nascent and fast-growing sector of the storage and server markets. There is no way that all of the hyperconverged and flash storage makers are going to make it, and Nutanix is trying to make sure it does.
There is always room for improvement in any storage platform, and with Nutanix partner Dell working to finish its whopping $67 billion acquisition of EMC and VMware, Nutanix will face some very stiff and staunch competition from that Austin, Texas titan. There is just no way around that, and it doesn’t matter how many nice things Michael Dell says about Nutanix. That deal closes, and the priority will be to get the VMware and EMC houses in better order and telling a consistent hyperconverged story. Dell will be pushing its platforms, vSphere and vSAN for heavily virtualized environment and ScaleIO for block storage. It would be good, we think, if Dell could come up with a unified block/file/object storage that is hyperconverged, but that remains to be seen if it is technically possible or economically desirable. (Red Hat will get there eventually with Ceph.) Dell could move up the stack from infrastructure to application development and management, finishing off its stack from CPUs all the way up to containers and code repositories.
To us, particularly with the PernixData and Calm.io acquisitions, Nutanix is able to snap these two companies up and start building a platform of its own with which to compete against VMware and whatever Microsoft and Red Hat cook up. While Hewlett-Packard Enterprise has its own hyperconverged storage by virtue of its StoreVirtual (LeftHand Networks) software, the intense competition from Dell might compel HPE to take a run at Nutanix before it goes public, and ditto for Cisco, which has a very loose relationship with Nutanix at this point, with the latter certifying a software-only variant of the Xtreme Computing Platform stack on Cisco UCS iron as of a few weeks ago. Nutanix sells appliances based on Supermicro iron, and has OEM agreements with Dell and Lenovo to distribute the Xtreme Computing Platform, which it now refers to as Acropolis, on their iron. It is hard to say what a Nutanix acquisition might cost, but it has raised $312.2 million in six rounds of venture funding between April 2011 when it came out of stealth, plus a $75 million loan from investor Goldman Sachs back in January of this year. When it took down $140 million in the summer two years ago, it was valued at $2 billion.
Just like cloud is just the way we are all going to do computing (whether public, private, or hybrid across clouds or hybrid across private and public), hyperconverged is the way we are all probably going to be doing storage in the long run. So the short-term expectation that hyperconverged storage could be a $3 billion to $4 billion market (depending on who you ask) with Nutanix possibly breaking through $1 billion in sales and presumably getting profitable within the next few years might be a radical undersizing of the market. What we can say for sure is that it is early days for a commercial-grade, Google-style infrastructure platform as Nutanix and several others claim to be, and that Nutanix has a ways to go before it can be profitable if the most recent numbers are any indication. In its latest S1 filing with the US Securities and Exchange Commission, for the first nine months of fiscal 2016 ended in April, Nutanix posted $305.1 million in sales, up 82 percent from the year-earlier period. But its losses widened to $126.1 million in that period, compared to $84 million in losses in the same three quarters of fiscal 2015. The gap is still getting wider. In April, the company had 3,100 customers, and we estimate that right now, based on past growth, Nutanix has maybe 3,500 customers and 50,000 nodes in the field in a market with millions of customers and north of 10 million shiny new servers sold every year.
Nutanix has not even scratched the surface.
In a conference call going over the acquisitions of PernixData and Calm.io, Pandey did not divulge the price that Nutanix paid for the two companies, but did say that the $75 million loan that it took from investor Goldman Sachs back in January was not used for the acquisitions and, importantly, has not been tapped at all. It is a cushion to give it maneuvering room as it presumably tries to get back to that IPO and raise lots of cash.
While the Acropolis platform has an integrated KVM hypervisor and obviates the need to pay VMware big bucks for the ESXi hypervisor and vSphere extensions to create a virtual SAN, Nutanix has a ways to go to turn this into a true platform – something that it wants to do, and Pandey makes no bones about that. With PernixData, and its FVP flash acceleration software and Architect management software, Nutanix will be able to substantially goose the performance of the underlying Nutanix distributed file system, and equally importantly, will be able to quickly absorb NVM-Express links to flash and other non-volatile storage such as Intel’s 3D XPoint into the Nutanix architecture.
This is obvious, and so long as Nutanix did not pay a lot for PernixData and its technologies, this will be good for the Acropolis stack. The Calm.io acquisition is less obvious, but points more to the real aspirations that Nutanix has – and the kinds of acquisitions it will be looking to make.
“Our roots were in the data plane, and in the last three years we have done a commendable job on the compute side with our own hypervisor,” explained Pandey. “And obviously in the past five years have done a lot to redefine to redefine Prism, which offers a single pane of glass to manage data centers as well. So the enterprise cloud operating system is not complete without an equal emphasis on one-click automation and orchestration. Developers increasingly to meld with IT ops, and want to think top down about applications rather than infrastructure. While we have successfully sold to IT ops and application administrators in the last four years, it is time for us to go higher up the value chain, even closer to users, that is, DevOps. Calm.io was a Sequoia funded company that was thinking deeply about applications and visual design of app workflows, even as Nutanix has been about top-down design and Prism has humanized infrastructure in the way we have thought about extremely mundane things about IT. That is exactly what we have found in the Calm.io team – democratizing orchestration without writing too much code. Their vision is highly aligned with ours and the hybrid nature of future IT, and their pane of glass helps design and orchestrate applications across AWS, Azure, and on premises webscale infrastructure. The lightbulb moment came when we saw how elegantly they had integrated Nutanix into their story.”
When the S1 was updated in the middle of August by Nutanix, it said it had acquired a technology company for 528,517 of its own shares and $1.2 million in cash. Presumably this is what it has paid for Calm.io, but it has not been precise. That works out to seven-tenths of a percent of shares that Nutanix has at the moment, which if the company is valued at around $2 billion still works out to something like $14 million in stock plus the $1.2 million in cash for Calm.io. PernixData had raised $62 million in cash and was looking for more to push its all-flash arrays and to ramp up its FVP flash management software, but apparently it could not get additional funding. Everyone is presuming that Nutanix paid at least this much to the owners of PernixData, but we won’t know for sure until the S1 is updated yet again.
We think other acquisitions will be in the works for Nutanix as it seeks to build up its enterprise cloud from hyperconverged storage to a full-on compute and storage platform, including virtual machine and container management and application orchestration.
The pressure is no doubt on Nutanix to grow its business and get closer to profitability at the same time. It is just shy of a $500 million annualized run rate and adding just under 500 new customers a quarter. Double that up again and Nutanix could become profitable. A lot depends on if Dell will continue to sell Nutanix software on its cloud PowerEdge-C variants and if Nutanix can convince HPE, the volume leader, to back its stack over VMware’s or Microsoft’s. Nutanix will be very fortunate if it can get Dell to continue to sell its stack and for HPE to start. If not, Nutanix may just have to focus on supporting an open version of its software that can run on any commodity X86 hardware and stop trying to directly get the major server makers to endorse it. It needs to make it an affordable software product that end users can pull into their datacenters, like Linux was a decade and a half ago, and less of a push from channel partners. It needs hundreds of thousands of customers, like VMware has. It can’t get there with partners who push competitive products, but it can get there with users who certify to specific hardware and can buy the software directly and at an affordable price.
The one thing that Nutanix needs to do is elaborate more clearly about how it intends to weave containers into its Acropolis platform. It is not enough to say that people can run containers on Nutanix. It is going to have to choose the method, and actually weave it in, not simply say that customers can run the CoreOS Tectonic container platform atop Acropolis or that there is a Docker plug-in for Acropolis. The secret to the success to Nutanix is that it makes complex clusters with virtualized compute and SAN-style storage look easier than they are. It needs to do the same for clusters, and then hide the complexity away from users. This is what VMware, Microsoft, and Red Hat are both doing with their respective container efforts.