Strategic briefs on far-reaching shifts in business & tech
Sign up for free summaries of our briefs
Nov 22 2019
Shifting tides around the leaderboard in the global cloud race
Reading Time Estimate
26 min
What’s Happening
In late Oct 2019, the finance ministers of France and Germany announced a partnership to work on a European open data infrastructure called Gaia-X. The German-led project, which will launch in 2020, is intended to support a European alternative to the current set of global cloud giants based in the US and China. In the announcement, the French and German leaders cited the need for European “digital sovereignty.” Gaia-X will develop “safe and sovereign” data infrastructure that includes data warehouses, pooling and interoperability, and help foster the growth of European cloud providers. Members will work on joint standards and APIs to connect different cloud services and facilitate data-sharing and portability, under EU-style data sovereignty and protection rules. Gaia-X is envisioned as a “managed environment of thousands of nodes” in the cloud, fog and edge, with customers directing the characteristics of each node.
Project members include both private companies (e.g. Siemens, SAP, Deutsche Bank, Deutsche Telekom, Bosch, Festo, OVH, Dassault Systèmes) and public institutions such as Germany’s International Data Spaces Association. Details will be presented early next year to other EU member states, who will then be able to join. Non-European companies such as Microsoft who can certify compliance with Gaia-X’s data sovereignty and data availability standards would be able to participate as well.
The dominance of the top 4 cloud players (Amazon AWS, Microsoft Azure, Google Cloud, Alibaba Cloud) – which together represent 72% of the global market – is a key driver behind Gaia-X. With none of the 4 based in Europe, EU regulators and corporations are concerned about relying on cloud providers that may comply with demands from their own governments for data. The 2018 US CLOUD Act (Clarifying Lawful Overseas Use of Data), for instance, provides a framework for US law enforcement to obtain data related to criminal investigations from American companies – even if the servers containing that data are on foreign soil. While tech firms can challenge warrants if they violate foreign law (e.g. GDPR), compliance with the CLOUD Act may contradict Gaia-X’s data sovereignty standards and bar US firms from participating. European government agencies and some companies have been reluctant to put sensitive data in a US cloud – one factor in why only 20% of German companies have transitioned to the cloud.
  • Multi-cloud: Enterprises are turning to multiple public clouds for many reasons, including greater resilience and reliability and avoiding vendor lock-in, as well as gaining negotiating leverage. Organizations will also opt for an additional vendor if it can serve specific needs better – such as regulatory compliance with local governments (e.g. data sovereignty), data centers closer to certain geographies, specialized services, or better performance in a particular cloud service or for a particular use case. One Apr 2019 survey showed 97% of large organizations were using a multi-cloud approach for mission-critical applications.
  • Hybrid cloud: Organizations are also using hybrid clouds more often, combining public cloud with private cloud (on-premise or hosted) and/or legacy systems. Hybrid cloud architectures can offer a greater sense of control and security over sensitive data as well as greater customization and cost efficiencies. For instance, a private-cloud application might lean on public cloud to handle spikes in demand. There is a large and growing ecosystem of tools – including from the big public-cloud players – to help balance workloads, transfer data, manage applications, or otherwise help public and private infrastructure work together better.
  • Cloud-native & cloud-agnostic applications: Applications are being built for the cloud with a modular architecture using loosely coupled microservices (which serve components of a larger application), containers (which package microservices so they can run independently and in different contexts), container orchestration tools (such as Google’s Kubernetes), and APIs (for communication). These applications may tap services from multiple cloud vendors, in addition to being more portable across them. Modern DevOps workflows – e.g. continuous deployment, automation – are helping accelerate delivery of these applications, enabling teams to be more responsive to the market. 90%+ of apps built may be cloud-native by next year.
  • Industry clouds: Cloud solutions targeted to specific industries are gaining momentum, with the major vendors growing their slate of offerings. For instance, nearly all the big cloud players are marketing solutions for gaming – e.g. Amazon Game Tech – to serve the burgeoning realm of cloud gaming (see our Nov 5 2019 brief). There are also industry clouds for advertising & marketing, automotive, education, financial services, government, healthcare, life sciences, manufacturing, telecom, media & entertainment, nonprofit, oil & gas, power & utilities, retail, and travel & hospitality. The cloud players are going head-to-head at the level of verticals such as financial services and public sector. In Jul 2019, a TechRepublic survey found 64% of organizations had either implemented industry cloud offerings or planned to do so in the next 12 months. Organizations are expected to spend $45B in 2019 on industry cloud services across just 5 verticals.
  • Cloud AI & data management: The big cloud players have recognized since at least 2016 – when they began offering cloud-based machine learning services – that cloud and AI are two halves of the same coin. The cloud offers data pipeline management and high-performance computing capabilities for the training of AI models. Conversely, AI offerings – e.g. machine learning, computer vision, natural language, translation, chatbots, recommendations – have become table stakes (and additional revenue streams) in the highly competitive public-cloud market. One 2019 forecast by Tractica predicted that AI could comprise up to 50% of total public-cloud services revenue by 2025.
  • Connecting the cloud to the edge: The edge – i.e. the billions of sensors, smartphones and connected devices – is becoming a distributed network for data processing and computing in real-time. Cloud players have realized that enterprise customers need “cloud-to-edge” solutions – e.g. IoT platforms that extend cloud management systems to edge devices. For instance, durable storage of data and training of AI algorithms might take place in the cloud while learnings are sent to the edge for decision-making based on local context. Use cases can include fleet management, equipment tracking, and smart buildings. With the rollout of 5G wireless underway, cloud-to-edge is only becoming more important.
  • “Serverless” computing: Also known as “function-as-a-service,” serverless computing represents the next layer of abstraction beyond infrastructure-as-a-service (IaaS) and platform-as-a-service (PaaS). With serverless, developers can focus on a block of code and pay the cloud vendor for only the compute resources needed when a given event is triggered (e.g. recording and analyzing a video when motion is detected). Servers are still needed but all the infrastructure is abstracted away. For developers, this means on-demand scaling, faster deployment, enhanced performance, and lower costs without having to manage the underlying infrastructure. Most of the big cloud players now have serverless solutions – e.g. AWS Lambda, Google Cloud Functions. The most significant disadvantages of first-generation serverless have been limitations on languages/libraries and vendor lock-in. This is changing, however, with Google’s Cloud Run introduced earlier this year – an environment for serverless-style containers that combines the ease and speed of serverless with the portability of containers.
  • AI-optimized chips: Hardware has been a major front in the cloud wars for the past couple years, with the big cloud players (not to mention chip players, other tech firms, and startups) working on their own AI-optimized chips. Most are focused on the second “inference” stage of AI execution (rather than the upfront AI training stage, which is dominated by Nvidia’s GPUs). Google’s Tensor Processing Units (TPUs), launched in 2016, is already on its 3rd generation. Google has also lately been exploring chips that enhance security and could help prevent hacks on low-level firmware and hardware. Microsoft, focused for years on the more flexible reprogrammable FPGA chips, appears to be diversifying its hardware strategy with a recent partnership to bring Graphcore’s AI chips to Azure customers as well as participation in the Open Cloud Project for cloud hardware design. Amazon is also expected to launch its Inferentia chip this year. China has been a particular hotspot with Alibaba and Huawei launching new chips in 2019, along with a new $29B state-backed fund to catch up to the US in chip manufacturing and design.
The “Big 4” public cloud giants
Amazon AWS
  • By any metric, AWS is the global cloud leader. It originally launched as the first modern public cloud in 2006 and had a 4-year head start before Microsoft entered the market with Azure. AWS’ share of the public-cloud market remains around 40% in Q3 2019. It has maintained share even as the market grew 37% (vs. Q3 2018) and AWS rivals have inched up at the expense of smaller players. While AWS’ growth over the past 3 quarters has been slowing, it still has twice as much market share as next-largest rival Microsoft Azure and managed to grow 35% in the most recent quarter (to $9B, for a $36B run rate). AWS is also far more profitable than Amazon’s retail business with a 25% operating margin (though this has also been declining).
  • AWS’ breadth is massive. It’s available in 22 regions globally with plans to expand to 4 more soon. With its head start, AWS can claim “significantly more services, and more features within those services, than any other cloud provider” – with 165 “fully featured services,” including 40+ not offered by any other major cloud provider. Services range from more traditional compute and storage to AI, IoT, robotics, security, and blockchain. In 2018 alone, AWS launched 1,957 new features and services. One recent offering was AWS Data Exchange, which lets customers discover, subscribe to, and use data products. AWS also has the largest set of industry cloud solutions among the top 4 public-cloud players. (Some critics, however, have noted that Amazon’s vast services portfolio can also create complexity and confusion.)
  • In the past, Amazon has not been as welcoming as the other players to multi-cloud – which has opened AWS up to criticism from competitors. It was also one of the last players to announce a hybrid cloud offering with AWS Outposts in Nov 2018. AWS Outposts, which extends AWS services to private cloud, is projected to launch late 2019 in alliance with VMware.
  • AWS has had challenges with security in recent months, with some members of Congress blaming it for Capital One’s misconfigured firewall and massive data breach. It was also hit by a major DDoS (distributed denial of service) attack a month ago, bringing down parts of AWS for up to 8 hours. It is also facing pressure from retailers and other companies it competes with in non-cloud domains, who are actively steering clear of AWS and, in some cases, engaging in strategic alliances against Amazon. Some, like Walmart, are encouraging their vendors to avoid Amazon’s cloud as well.
  • Recently, in an effort to stave off competitor inroads, AWS introduced a new discount model called Savings Plans which cuts customers’ bills in exchange for 1- to 3-year commitments at a minimum spend. (Complex pricing resulting in unexpected bills has been an ongoing complaint about AWS.) It is also investing in building out its enterprise sales and marketing organization to compete with Microsoft’s large enterprise salesforce.
Microsoft Azure
  • Azure came to the market as the second major IaaS player in 2010, 4 years after AWS’ launch. Earlier this month, AWS CEO Andy Jassey asserted he believes Azure is still 2 years behind AWS in maturity and functionality. It has about half the market share of AWS at just under 20%. Azure, however, is the only serious contender to AWS right now, with both a clear claim to the #2 position and a faster growth rate than AWS – 63% in the most recent quarter (constant currency) vs. AWS’ 35%. Azure is seeing “material growth” in large $10M+ contracts, with strong and improving gross margins as well. As with AWS, however, Azure’s growth has fallen over the past 3 quarters, from 76% in Q2 2019 to 63% in Q1 2020 (in constant currency).
  • Microsoft's AI offerings are another key driver for Azure, which has what Microsoft claims is the “most comprehensive portfolio of AI tools, infrastructure and services.” Based on side-by-side comparisons, it’s certainly one of the broader portfolios. Microsoft has been actively developing cloud AI capabilities through acquisitions (e.g. the Jun 2018 acquisition of cloud ML startup Bonsai) and a bevy of product releases (most recently Azure Synapse, which helps ingest, prepare, manage and serve data for business intelligence and AI apps). Azure AI now has 20,000+ customers, with 85%+ of Fortune 100 companies using it in the past year.
  • Microsoft is also building out its suite of Azure offerings to meet demand for edge computing, investing $5B in IoT programs from 2018-2022. Last month, Microsoft unveiled several solutions that extend its Azure IoT platform – e.g industry templates for IoT apps, weather data for location-based IoT, time-series analysis for IoT data. Microsoft currently has about 23% of the IoT cloud market vs. AWS’ 34%.
Google Cloud
  • Google Cloud came into being as a platform-as-a-service called App Engine, which was launched in preview in 2008. However, it didn’t move into general availability until late 2011 and was only consolidated as a business unit in 2015 – which means Google was late to the game in taking public cloud seriously as a business. Since then, it has poured tens of billions into what is now known as Google Cloud, with $13B earmarked for cloud investment in 2019 alone. Google Cloud is now in 20 regions with 5 more regions underway, as well as connected by a network of 13 subsea cables. It announced plans back in Jul 2019 to triple its salesforce over the next few years, and has since continued to hire aggressively for both sales and technical roles.
  • Google currently appears to be in 3rd place with a market share of less than 10%. Though it doesn’t share exact figures for Google Cloud, in Jul 2019 it said run rate was $8B+ (up from $4B+ in Feb 2018). While Google Cloud has grown more strongly than the overall market, it has gained ground slowly in market share against its larger rivals.
  • Google Cloud’s key differentiators are its AI, analytics, data management, cloud-native tools, and security. Industry watchers are interpreting Google’s approach to the market of late as, “let us be your no. 2” cloud provider. Many companies such as Spotify, GitLab and Twitter are using Google Cloud for specialized workloads – particularly those involving data and Kubernetes (container orchestration). One survey at an AWS conference found that 24% of respondents were using both AWS and Google Cloud, and 18% were using all 3 of the leading clouds (though the most popular combo was AWS-Azure). Google recently consolidated its popular AI offerings under a centralized AI Cloud platform, bringing together tools across the lifecycle from data ingestion and preparation to training and deployment. Google is also building out its portfolio of industry-specific cloud-based AI solutions.
  • Hybrid and multi-cloud are currently key focus areas for Google Cloud. In Apr 2019, it announced Google Anthos, a hybrid/multi-cloud platform that lets customers build applications that run on their own servers, on the edge, in Google Cloud, and competitor clouds like Azure and AWS. Migrate for Anthos will help enterprises directly convert existing virtual-machine workloads into containers. Data migration has been a major theme, along with data processing, analytics and API management for hybrid and multi-cloud. Over the past year, Google has acquired migration-related startups Velostrata, Alooma, CloudSimple, and Elastifile, in addition to engaging in migration partnerships with SAP, VMware and Microsoft. On the data and analytics front, Google released Dataproc on Kubernetes to help manage data processing across cloud and on-premise, while $2.6B business analytics acquisition Looker is updating its offerings to support all 3 of the big cloud providers. Finally, Apigee (another Google acquisition) earlier this year announced API management for hybrid clouds.
  • Google Cloud’s reputation has been, relative to its rivals, less enterprise-oriented and more aligned with digital natives and startups. However, its customers include 9 of the 10 biggest media companies, 7 of the 10 largest retailers, and 6 of the top 10 enterprise companies (though not all may be using Google Cloud as their primary cloud). New leader Thomas Kurian is repositioning Google Cloud, emphasizing its differentiators and role in a multi-cloud strategy rather than seeking to make it enterprises’ one and only cloud.
Alibaba Cloud
  • After a major expansion in 2015, Alibaba Cloud has grown dramatically over the past few years and is now competing with Google for the #3 spot. Much of the Alibaba’s presence and growth stems from its domestic market, where it is the largest provider in China with 43% of the market (followed by Tencent with 17%, then Amazon and Baidu). It also is the leading cloud provider across Asia-Pacific with about 20% of the market. In the most recent quarter, Alibaba’s cloud revenue grew 64% year-over-year to reach $1.3B for a $5.2B run rate (though it’s not yet profitable). Globally, that means Alibaba Cloud’s market share is still only in the mid-single digits. Similar to the other cloud players, Alibaba has seen declining growth since its peak of 104% in the quarter ending Dec 2017.
  • Like Amazon, Alibaba Cloud originally developed much of its cloud technologies internally to serve its own massive ecommerce platform and adjacent businesses. It has invested in building out new features and services for the Alibaba Cloud platform, reporting about 1,300 new products and features in the second half of 2018 and 300+ in the most recent quarter (as it has focused more on value-add rather than quantity). Alibaba has extended its reach internationally and now operates in 20 regions globally – 8 in mainland China, 7 in Asia-Pacific (outside China), 3 in Europe and the Middle East, and 2 in the US, with one coming in Brazil.
What It Means
If it weren’t for the growing push for digital sovereignty among nation-states and political unions, it would be easier to predict what will happen in the public-cloud market – at least in the near term. A host of players – such as IBM, Oracle, Tencent and Baidu – have been drawn to a cloud market that is still in relatively early days with 30-40% annual growth and less than 20% of enterprise workloads on a public cloud. That said, the leaderboard has been surprisingly stable. The key shifts are:
  • Microsoft is picking up ground as a rising threat to Amazon in the enterprise
  • Google is steering towards specialized services in a multi-cloud environment
  • Alibaba is taking over Asia Pacific and competing for the less mature markets globally
  • Greater concentration in the top 4 mean the players lower down on the list are losing share
While the top of the leaderboard is unlikely to make any dramatic shifts in the near term, there is still room for a reordering, especially if one or more of the players lose focus and take their foot off the investment gas pedal. However, the scale and breadth of needs in the cloud space means there probably won’t be just one successful player.
Given where we are in the game, all of the big cloud players know what the cloud trends and table stakes are – e.g. hybrid and multi-cloud, cloud AI, industry clouds, etc. Maintaining a continuous stream of product updates and releases is just part of the buy-in to be a big cloud player. As soon as one player brings forth a significant innovation, the others will follow suit. For cloud customers, however, who may be managing multiple clouds, legacy systems, workload migrations, pressure to integrate more AI into applications or their business, concerns about cybersecurity, constrained staff, and new ways of working, the end state is likely to be a sense of overwhelming heterogeneity and complexity. Unfortunately for the cloud players, this may mean they are investing an enormous amount and are still not differentiated.
The outcome of greater complexity is likely to be one of two paths – a desire for better orchestration and solutions to manage cloud complexity or a desire for a more simple architecture and environment. There is room in the market for both. The larger enterprises will be looking for tools and consulting services on the roadmap to a new operating model. Others will bet on one primary cloud provider for the majority of their services, with other vendors in the portfolio to meet certain needs and for diversification.
Differentiation is the key challenge for the respective cloud players. For AWS, its key differentiators are its market leadership and vast breadth of services. For Microsoft Azure, the differentiators are its strength in the enterprise and hybrid cloud. For Google, it’s advanced AI and cloud-native tools. For Alibaba, right now it’s largely its scale in the geographic region. Most of these differentiators are subject to competitive pressure – e.g. other players building out their portfolio of services including hybrid cloud, AWS growing its presence in Asia Pacific.
Microsoft’s strength in the enterprise is among the more resilient differentiators in this group. If we unpack what it means to have strength in the enterprise – e.g. understanding of enterprise needs and IT across the organization, solutions that address enterprise needs like hybrid cloud and compliance, installed enterprise customer base already using its software, experienced high-touch enterprise salesforce, large active developer community, extensive partner ecosystem – it’s not easy to replicate. Microsoft has been able to parlay these advantages – as well as a relatively neutral position in ecommerce and other wars – into strategic alliances and big wins like the JEDI contract. However, even strength in the enterprise may not be a differentiator forever – for instance, it could be matched with a big acquisition and successful integration.
In the long run, it’s still anyone’s game. Amazon could successfully fend off all its rivals by shoring up its vulnerabilities and growing its enterprise capability. Microsoft could continue its steady climb, eroding AWS’ market share until it’s knocked off the top pedestal. Google and Alibaba could use their respective niches as footholds to expand their footprint. The basis of competition, however, will eventually shift from products and features to the customer experience. The cloud players will have to consider how they will shape a differentiated experience – perhaps one that is less complex, frustrating or overwhelming and more seamless and open.
Digital sovereignty and nationalism is presenting a major wrinkle to all of this. European “digital sovereignty” is the key rationale behind Gaia-X, and part of a broader shift towards “digital nationalism,” which has been gaining momentum on a global scale. While it’s not clear whether Gaia-X will be successful (it’s not the first time a European country has sought to build a competing cloud player), the underlying dynamics driving the project are not going away.
As US and Chinese cloud players go global, there will inevitably be pushback from political bodies who recognize the risks of handing over sensitive or proprietary data to corporations subject to another country’s laws. With the rise of national internet shutdowns by governments, the issue is not just about use; it’s also about access to critical data. Looking ahead, for the larger and more developed markets, it’s unlikely that governments will stand by long and allow a foreign player to dominate in the cloud. Certainly in China, Alibaba is likely to remain the leading cloud provider for some time; it is also in China’s interests to see Alibaba Cloud expand across Asia-Pacific and other areas where it seeks influence.
Digital sovereignty is also about access to assets that can help build national capability and competitiveness – e.g. data to train AI models and build associated capability in the workforce. Europe is already lagging behind the US and China in key technologies like AI. Government officials are realizing this, which is one of the reasons why data pooling and sharing is a major aspect of Gaia-X. One of Gaia-X’s goals is to “enable companies and business models to scale out of Europe competitively worldwide.” A similar dynamic is happening in the cybersecurity realm, where European business associations – recognizing the dearth of local capability – are encouraging governments to fund or otherwise support local cybersecurity players.
The advent of laws like the CLOUD Act are challenging data localization (or data protectionism), laws in 45+ countries instituted to prevent their citizens’ data from freely flowing beyond their nation’s borders. Regulations such as the EU’s General Data Protection Regulation (GDPR), with its stringent data privacy and security standards, are creating proxy localization laws by deterring market participation by smaller businesses and organizations afraid of heavy penalties. Data sovereignty and localization laws can impose unintended costs on the citizenry, by limiting access to services and information, reducing competition and thereby raising prices. European cloud services, for instance, are currently more expensive than US cloud services.
Digital sovereignty presents challenges to cloud players like Amazon and Microsoft whose business models rely on scale. They have condemned the project and say it could limit organizations’ choice in providers as well as their ability to scale globally, without necessarily increasing security. The threat of fragmented dataflows globally should be taken seriously by multinational enterprises as well as the cloud providers that look to serve them. Disconnected digital environments pose the following risks: Siloed and inefficient data collection, difficulties in performing analytics or applying AI, possibility of under-serving small populations, slower insights and communication, challenges in working with global ecosystems of partners and vendors, blind spots in supply chains and cybersecurity efforts, and overall increased complexity from managing across disjointed environments.
For the time being, it’s not clear to what extent Gaia-X will have an impact on the big global cloud players. First, we don’t know whether it will be successful. There’s also still the possibility that the cloud giants will be able to join Gaia-X, if their domestic regulation (like the CLOUD Act) doesn’t bar them from doing so. Furthermore, the effort so far doesn’t seem to include any measures preventing foreign cloud players from participating in the market outside of Gaia-X. In the near term, the US and Chinese cloud players may have to adapt their processes or take a hit in European market share, but it’s unlikely their operations in Europe will be halted. In the longer term, European companies may have to choose between global players and more local players (perhaps propped up with the new infrastructure). While corporations may have heartburn over sensitive data and the CLOUD Act, in many cases they will continue to opt for the market leaders because of the cost, scale and breadth of their offerings.
Disclosure: Contributors have investment interests in Microsoft. Amazon and Google are vendors of 6Pages.