Here we are Friday ahead of the start of Nvidia’s flagship GPU technology conference. And we’re no doubt excited in anticipation of what we suspect will be a deluge of computing and networking hardware, and puzzled over what this omniverse really means (like many of you).
But for some reason our thoughts keep returning to software and how Nvidia is building an increasingly complete and elaborate set of system software to run HPC, AI, data analysis, visualization and now world simulating applications.
Nvidia has been creating drivers for GPUs for decades and started building commercial HPC and AI frameworks and the underlying libraries and compilers a decade and a half ago. And Nvidia’s big acquisition here — big is a measure of importance, not spend — was The Portland Group in July 2013. And since that time, it’s built stacks of HPC and AI frameworks neatly packaged in Kubernetes containers the basis for applications in the fields of simulation, modeling and machine learning.
Nvidia acquired Mellanox Technology in March 2019 for its networking software — literally the bits, but also the know-how embodied in its workforce – as with its switch ASICs and network interface cards, and almost immediately it acquired Cumulus Linux in May 2020 for the same reason. Expertise is more important than any specific configuration of bits that we call software. With Mellanox and Cumulus Nvidia has a deep expertise in interconnects and network operating systems which complements their own very good SerDes design team and which among other things created the NVLink interface and the NVSwitch storage area network – well that’s what it’s all about after all Lashing GPUs together into a kind of shared memory SuperGPU.
More recently, Nvidia acquired Bright Computing for its eponymous cluster configuration and management tool, and more recently Nvidia has entered the storage industry with its rapid acquisitions of distributed block storage provider Excelero and distributed object storage maker SwiftStack. We still think that Nvidia needs high-performance filesystems if it wants to own a full stack, and we also think the Excelero and SwiftStack deals were more about getting people than a specific piece to get software-defined storage. As far as we know, Nvidia will develop its own file system with object or block underpinning and optimize it for the HPC, AI and data analysis workloads to track them in the data center.
What else is Nvidia missing from its software stack? Well, a low-level OS – which obviously has to be Linux – as the interface between its system software and the iron and a high-level OS – which probably means Kubernetes, but Docker and Hashi Stack are also options – plus containerize and pod- Microservices and denser applications. If not Kubernetes, Docker is probably a bad idea given how far it has fallen from grace as Kubernetes has risen as the new platform abstraction of choice, but Hashi Stack is a reasonable alternative to Kubernetes for building containerized applications at scale to manage, even if it did, would be an expensive acquisition given HashiCorp’s market cap of $5.6 billion. (That’s less than $13.3 billion from HashiCorp’s December 2021 IPO, so it’s just got a lot cheaper.)
When rumors arose after Pat Gelsinger switched the position of chief executive officer from VMware to Intel that Intel might buy VMware, we suggested that Nvidia might buy VMware if the arm deal went through. It did, but Nvidia doesn’t appear to be making the switch to VMware, which is good because with a market cap of $47.7 billion at press time, VMware is far too expensive. And Intel needs to save every tens of billions of dollars it can scrape together to build the foundries it should have been building over the past decade. Nvidia could have made Arm bigger and stronger, we think, but will itself be more nimble against rivals in CPU and GPU computing, as well as networking and now storage, without having to digest and defend the Arm acquisition.
Rich in cash — Nvidia had nearly $20 billion in cash in the bank when it ended its 2022 fiscal year in January — the company should add another block to its software stack by gaining control of Linux and Kubernetes distributions obtained.
IBM bought Red Hat for $34 billion back in October 2018, taking the company off the negotiating table. Canonical, owned by billionaire Mark Shuttleworth, has put together a very respectable Ubuntu Server distribution of Linux with all sorts of extras like bare metal and Kubernetes support, and would go well with the Nvidia stack.
But there are two executives who sell things to the data center who also see green. One of them, of course, is Jensen Huang, co-founder and chief executive officer of Nvidia, and the other is Melissa Di Donato, chief executive officer of SUSE, the longtime provider of enterprise Linux and now a fairly complete Kubernetes platform.
(The green for Nvidia symbolizes growth and change, and her eye looks to the future; for SUSE, the green comes from the chameleon adapting to its surroundings, and if you look closely, this chameleon has golden ratio proportions and is downright really friendly, although nothing escapes his eye.)
SUSE is perhaps a better fit than Canonical and is actually acquirable because it was bought by Novell in 2003, Attachmate in 2010, Micro Focus in 2014 and EQT Partners, a Swedish private equity firm, after a long line of acquisitions that bought SUSE in 2018 for $2.5 billion. In May 2021, EQT took SUSE public on the Frankfurt Stock Exchange in Germany, raising $1.35 billion and giving SUSE a then market capitalization of $6.1 billion.
We know SUSE in terms of its finances because it went public, and we’ve compiled key financial details in this table:
These numbers allow us to look at SUSE and make comparisons over the trailing twelve months, and when you do, SUSE had revenue of $596.8 million, up 15.1 percent, and income before income taxes , depreciation and amortization (EBITDA) of $203.7 million, up 3.6 percent. The core Linux business accounted for $517.4 million in trailing-12-month revenue, up 10 percent, and the emerging business, primarily Kubernetes with Rancher management extensions, Neuvector container security, and Harvester hyper-converged distributed storage for container platforms . (Harvester is based on the Longhorn distributed block storage system developed by Rancher Labs, which SUSE acquired in July 2020. (We covered Rancher extensively a few years ago.)
Nvidia wouldn’t buy SUSE because it needs the revenue stream. At least it doesn’t need that now. Instead, Nvidia would enter into such a deal at a reasonable premium to the company’s current $5.1 billion market cap because it wanted a team of 2,400 Linux experts, including engineers and salespeople, to lay the groundwork for a full Linux Kubernetes stack for all the higher-end software that Nvidia has already developed and even more that it will develop in the years to come.
We recognize the close partnership between Nvidia and VMware and that VMware server virtualization is the substrate for most systems in the enterprise world. And Nvidia packages its software so that it can be packaged to run on the Tanzu flavor of Kubernetes, which runs on the ESXi hypervisor and in conjunction with VMware’s NSX virtual network and vSAN virtual storage. We also know that Nvidia has been working with VMware on its “Project Monterrey” port if the ESXi hypervisor is to run on Nvidia’s BlueField family of DPUs. (Three times my brain wanted to type “moneytree” there. . . . Hmmm.)
So what. As has been the case for years, Nvidia will enter into partnerships with competitors. It makes servers and helps others build servers. It can control and optimize its own Linux operating system and Kubernetes distribution while simultaneously running on Red Hat Enterprise Linux and Ubuntu Server and their respective Kubernetes, or optimize it for the Linux distributions and Kubernetes container services running on the largest available are clouds.
One does not negate the other. But there are obvious benefits in terms of performance, ease of deployment, and price control that come from a full stack. Not everyone will want to pay the VMware premium – hefty, mind you, as Nutanix learned the hard way in the hyper-converged storage space, forcing it to create its own flavor of KVM. VMware ESXi still underpins many Nutanix storage, but for those who are price conscious, Nutanix’s Acropolis stack helps cushion the software price hit for HCI. This is no different.