Liftr Look Ahead: 2019 Q3 Earnings Calls - Cloud Providers Help NVIDIA Grow
2019 Q3 cloud providers and supply chain earnings calls, the wrap-up:
Quarterly reports for cloud providers and their supply chain usually occur in two clusters. This week, we’ll focus on the second cluster of 2019 Q3 reports, from cloud provider Tencent, middleware vendor VMware, chip vendor NVIDIA and system vendors Cisco, Dell EMC, HPE and Lenovo.
Tencent Cloud is part of Tencent’s Cloud & Smart Industries Group (CSIG). Tencent focused its earnings call on increased operational efficiencies to offer more cost competitive products and services. It has shifted procurement from branded equipment (OEM) to unbranded (ODM), which enables Tencent to order more customized servers and other gear at lower prices, even as it increased its spending on servers to expand Tencent Cloud.
Earlier in the quarter Tencent Cloud announced its first self-developed server, called Xinghai, based on a custom version of AMD’s latest EPYC second-generation processor. Tencent Cloud positioned Xinghai as more than 35% faster than its currently deployed solutions over a wide range of applications.
Although we covered GCP in our previous earnings analysis, Google recently acquired CloudSimple. This acquisition estranges Azure from its primary VMware integrated solutions provider. AWS is tightly integrated with VMware Cloud on AWS and does not rely on CloudSimple. Azure will have to follow AWS and integrate more closely with VMware—if VMware will let Azure do that—or Azure will need to evolve another solution that is not dependent on Google-owned CloudSimple software.
VMware threw Azure a bone during its earnings call by mentioning an expanding partnership between the two companies, but without any detail. It also leaned into a shift to security services via its acquisition of Carbon Black.
During the call, VMware continuously reiterated the success of its “growing hybrid cloud subscription and SaaS portfolio” and its continuing investment in that business to drive VMware’s top-line growth. However, investment in its new Tanzu Kubernetes-native platform speaks to future concerns. VMware designed Tanzu to bridge traditional virtual machine development and deployment with cloud-native container development, deployment and orchestration.
The top tier public clouds are getting much more serious about attracting mainstream enterprise customers. This competition will intensify in 2020. AWS may have more surprises in store in early December at its re:Invent conference, so we’ll wrap up the year in public cloud competition after that.
Like VMware, Cisco emphasized that it is growing a security and access management business. Cisco labeled its core enterprise networking business as experiencing a “challenging macroeconomic environment”. It is moving to automate customer connectivity across any cloud and announced a partnership with Microsoft to improve campus connectivity to Azure.
Dell Technologies CFO Thomas Sweet stated, “We have deliberately walked away from a number of large server deals [in China] where the pricing just most absolutely made no sense to us, principally with the hyperscalers.” This is consistent with Tencent’s statements that it moved from branded to unbranded server purchases.
In response, Dell is focusing on building out on prem private cloud and HPC deployments using hyperconverged infrastructure (HCI) with emphasis on storage, EMC’s traditional strength. Dell is also promoting its “on demand” pay-as-you-go private infrastructure financing plan.
HPE highlighted its new HPE Container Platform, a software platform leveraging its acquisitions of BlueData and MapR. HPE Container Platform is designed to enable app developers to develop and deploy containerized apps on bare metal or virtual machines at any public cloud and in private edge infrastructure.
HPE grew its HPC business over the past year, mostly based on its acquisition of SGI three years ago. HPE also grew its Composable Cloud and HCI businesses, and intends to keep expanding its HPC business, having recently acquired Cray.
Lenovo’s Data Center Group (DCG) highlight was continuing to narrow its losses after nine consecutive down quarters. Lenovo DCG overall revenue declined about 14% due to “softness in demand from some of the largest hyperscale customers”. However, DCG non-cloud revenue was up 13%, which highlights the magnitude of DCG’s earlier cloud sales, with non-cloud revenue in China up almost 50%. Lenovo is also focusing on HPC and flexible private cloud deployments.
Dell, HPE and Lenovo are all moving up-market into HPC and on prem AI training and inferencing deployments. They are all focused on software-defined infrastructure and HCI hardware. As public clouds become more viable options for mainstream and high-performance applications, pressure to move faster into the high-end of the market will continue. All of which reminds us of the minicomputer market in the 1980s as modern server architectures buried that market. We do not think future cloud growth will drive branded server sales in a meaningful way. As Tencent said, their buying habits have shifted to ODMs.
NVIDIA reported in Q3 that its overall enterprise revenue was down year-over-year, while cloud revenue grew year-over-year. The clouds give NVIDIA three areas of growth:
- Internal cloud services for AI training,
- Internal cloud services for AI inferencing, where conversational AI is a major driver for NVIDIA, and
- Public cloud services for AI, which is mostly start-ups at this stage of the industry.
NVIDIA said that for the first time it shipped more of its Tesla T4 chips than its V100 chips. Tesla V100 chips are primarily used for training, while T4 chips are primarily used for inference, primarily for conversational AI. NVIDIA’s inferencing business doubled year-over-year.
Our Liftr Cloud Components Tracker will continue to track NVIDIA GPU deployments at the largest public clouds, including changes to cloud customers’ mix of new NVIDIA GPU chip deployments.
NVIDIA also highlighted extending its partnership with VMware to include Red Hat OpenShift. NVIDIA’s goal is to create a telco-ready enterprise stack that encompasses both VM and container deployment, and it is a different route than VMware is taking internally with Tanzu. VMware is trying not to be perceived as only a VM deployment company. However, with so many open source container delivery and orchestration solutions emerging, it is hard to see how VMware can differentiate for very long in that market.
Liftr Cloud Components Tracker: http://bit.ly/2QceXlT
Liftr Cloud Regions Map: http://bit.ly/2LGB5PV
Follow Us On: https://www.facebook.com/LiftrInsights/