Apart from increasingly difficult customers, what other issues exist?

Author: Li Xiaodong

Editor: Sang Mingqiang

The root of the issue stems from the actual market performance of domestic public cloud in recent years: starting almost from 2022, China's public cloud market has been in a state of decelerating growth. The official narrative is that the market is entering a period of stable development.

In contrast, although there are differences in data statistics, overall, the global cloud computing market still maintains rapid growth, and public cloud giants such as AWS and Azure remain resilient. That is to say, the market's demand for cloud, especially public cloud, is still certain.

Under these circumstances, the domestic public cloud market has been discussing two main topics in recent years: one is price reduction, and the other is going overseas.

After Alibaba Cloud fired the first shot of price reduction, other cloud vendors also followed suit. The reasons are numerous: intensifying market competition, vendors vying for market share, which is also the case abroad; the expanding scale of public cloud service providers, resulting in gradually decreasing operating costs, thereby further expanding their market share and achieving economies of scale; and technological progress, leading to cost reduction and universal benefits.

The trend of price reduction means lower IT costs, higher flexibility, and more choices for enterprises and developers. Of course, it also indicates the homogenization of vendors' basic product functions, including cloud hosts, object storage, various databases, and drag-and-drop website building templates, with no significant competitive edge.

Going overseas is another way out sought by players. For example, after Alibaba Cloud reduced its prices, this strategy became a means for it to compete for overseas markets. It's worth noting that there is a premise: Alibaba Cloud's price reduction effect in China was once not optimistic, and they have publicly acknowledged this. Therefore, while turning their gaze overseas, they are also continuously expanding the scope of price reductions for products and customers.

As Liu Weiguang, President of Alibaba Cloud's Public Cloud Business Unit, said, in fact, many Chinese companies do not have cloud computing, do not use public cloud, or do not enjoy the true value brought by public cloud, or use cloud as server resources. In other words, domestic customers' willingness to pay for public cloud is not ideal, so they want to encourage enterprises to migrate to the cloud through price reductions.

However, in the past two years, while the growth of public cloud has slowed down, there is another trend that makes cloud vendors nervous: some enterprises are starting to migrate off the cloud. What are the considerations behind this?

01

Escaping from Public Cloud

There is no clear record or consensus on the exact identity of the first company to implement "CloudExit." However, Elon Musk played a significant role in making this issue a hot topic.

In 2022, Musk acquired Twitter and renamed it X, after which he implemented a series of major reforms aimed at reducing the company's operating costs, with "cutting cloud services and additional servers" as a key move.

For example, by shutting down the Sacramento data center and reconfiguring servers, X can save over $100 million annually. Additionally, optimizing cloud service usage and running more workloads locally has reduced X's monthly cloud costs by 60%.

This data shatters many people's perceptions. Generally speaking, the purpose of migrating to the cloud is to save costs. Traditional IT systems require purchasing hardware equipment, building data centers, maintaining networks, etc., which require significant capital investment.

Cloud computing adopts a pay-as-you-go model, where enterprises do not need to purchase and maintain their own servers and hardware equipment, but only pay for the resources they actually use. Moreover, cloud service providers usually take responsibility for system maintenance and upgrades, saving enterprises from incurring additional maintenance costs. The high flexibility and scalability of cloud computing also allow enterprises to quickly adjust resources based on business needs.

Musk's actions have made the industry re-examine the cost-effectiveness of cloud computing. The long-term operating costs of cloud computing may not always be as low as expected. However, the industry has noticed that X has not abandoned cloud computing but continued to operate its business using private cloud, especially in scenarios with massive data volumes requiring frequent processing and analysis, where local or private cloud solutions may be more cost-effective.

It is worth mentioning that it's not just Musk who is escaping from public cloud. Almost at the same time, David, the co-founder of 37Signals, also officially announced his departure from AWS and Google Cloud, which he had been using for years. The most important reason for not going back to the cloud is also cost:

37signals pays over $3.2 million annually to two public clouds, including more than $1 million for Hey, an enterprise email service with 300,000 users. They concluded that for steadily growing medium-sized enterprises, renting computers is not cost-effective in most cases.

It is understood that David estimates that migrating off the cloud can save at least $150 per year. They expected it would take several years, but the last application was successfully migrated back to local hardware in just six months.

"The two main benefits of migrating to the cloud are: first, when the application has simple functionality and low traffic, it can be fully managed by cloud services, which is friendly for startups; second, cloud computing can handle highly irregular load changes, such as sudden surges or large fluctuations in usage, with elastic solutions."

However, 37signals has been in the cloud for many years and is a steadily developing company. Its complexity has not been significantly simplified due to cloud migration, and the number of operations and maintenance team members has instead increased year by year. "The benefits of cloud migration are no longer the dividend of 37signals," said David. Compared to when Hey was first launched, its user base exploded to 300,000 in less than a month, and now Hey's growth has stabilized.

In short, paying a huge price to handle rare cases of sudden surges seems disproportionate to him. If companies rely too heavily on proprietary managed services or serverless solutions, they may find it difficult to escape once the bills start soaring.

But when it comes to the fundamental motivation for migrating off the cloud, David believes it allows companies to rethink what kind of network a software vendor should use to run their business. In other words, companies need to reassess what software operating environment suits their current state.

While the public cloud environment provides convenience and flexibility, it also poses some limitations and challenges, such as issues with performance, security, and cost. By migrating off the cloud, companies can have better control over their operating environment and choose a network architecture and infrastructure that better suits their business needs.

02

Don't Stay in the Cloud Just Because of Hype

It is not difficult to observe that whether to migrate to or off the cloud depends on the enterprise's development status: most foreign companies believe that cloud computing is useful for enterprises in the early stages of their lifecycle, with insignificant expenses or those that may not be able to continue operating within two years. Of course, it also depends on the cost comparison under different solutions.

A study conducted by Citrix, a well-known American technology company, found that 42% of companies already have plans to migrate off the cloud, and 94% of the 350 IT leaders surveyed have participated in cloud migration projects in the past three years. This means that cloud migration is not just a concept or trend but has been widely implemented in practical operations. IT leaders are actively responding to and driving this change, possibly because they see the potential benefits of cloud migration for their business.

Looking domestically, such demand also exists. For at least the past decade or more, migrating to the cloud has been politically correct for enterprises and has even become an important standard for measuring a company's digitalization level. However, many people tend to overlook the significant costs that accumulate over time under the cloud leasing model.

Assuming a medium-sized enterprise runs its core business on public cloud, as the business develops and expands, the usage of cloud resources also increases. Since growth is usually nonlinear, costs will also increase nonlinearly. In the initial stage, the enterprise may only need to pay about $100,000 per month for cloud services, including costs for virtual machines, storage, databases, network bandwidth, and other resources.

However, as the business grows, the enterprise may need to add more virtual machines to handle workloads, possibly hundreds or thousands of them, more storage space to save data, and higher network bandwidth to support more user access. The increased usage of these resources will lead to an increase in cloud service fees.

It is worth noting that public cloud usually adopts a subscription model, where enterprises must continuously pay for the resources they use. Even during inactive periods, as long as resources are allocated and reserved, fees need to be paid. On the other hand, fees for sending data from the cloud service provider's network to the public network are usually calculated based on the amount of data transmitted, and some cloud service providers may charge anywhere from a few cents to several dollars per GB of outbound traffic.

This is also a direct reason why many enterprises choose to migrate off the cloud in the later stages.

Additionally, as data volumes grow, the costs of services such as data backup, recovery, and archiving will also increase. At the same time, to ensure business security and compliance, enterprises may need to purchase additional security services such as data encryption, intrusion detection, identity management, etc., which will also increase the total cost of cloud services.

In the later stages, if the enterprise's business continues to grow rapidly, the cost of public cloud services will increase rapidly. For example, assuming the number of virtual machines increases from 100 to 500, with a monthly cost of $200 per virtual machine; storage needs increase from 1TB to 10TB, with a monthly cost of $1,000 per TB; network bandwidth increases from 1Gbps to 10Gbps, with a monthly network cost of $50,000. In addition, there are extra security services such as data encryption at $10,000 per month and intrusion detection at $20,000 per month.

Applying the same scenario to China, while cost considerations are also important, in fact, large and medium-sized domestic enterprises, especially state-owned enterprises and government entities, which are major customers of cloud vendors, are more concerned about information security and compliance issues.

03

Has the Golden Age of Public Cloud Passed?

From a market share perspective, public cloud has indeed led for several years: according to data from China Academy of Information and Communications Technology, in 2019, China's overall cloud computing market reached 133.4 billion yuan, with a growth rate of 38.6%. Among them, the public cloud market size reached 68.9 billion yuan, an increase of 57.6% compared to 2018, surpassing private cloud for the first time in market size.

Recall that five years ago, with the increasing maturity of technologies such as AI, 5G, and AR, a large number of emerging IT enterprises emerged in China, and traditional industries also entered the fast lane of digital transformation under policy promotion. Public cloud vendors represented by Alibaba Cloud became China's largest cloud computing vendor, ranking first in the Chinese market for many years, driving the rapid growth of the entire public cloud market.

Today, however, many small and medium-sized technology innovation companies have not achieved sustainable development. Among them, even smaller companies cannot afford the long-term costs of public cloud. There are also more cost-effective alternatives to meet their needs outside of public cloud.

This is why some believe that the audience for public cloud is only suitable for companies that are very early-stage or will no longer exist in two years, as well as companies that do not care about sunk costs, e-commerce companies with highly volatile traffic, and companies facing overseas compliance issues.

For large state-owned enterprises and government clients, public cloud is treated as a capital expenditure in financial statements, while private cloud is a fixed asset. Although the amortization cost of private cloud may be higher, it is still reflected as an existing asset on the balance sheet, which better aligns with the needs of state-owned enterprises and government agencies for the preservation and appreciation of state assets.

The leadership of these enterprises will be concerned about any potential risks, such as if the cloud provider goes bankrupt or stops production, then the cloud software will no longer exist, and the documents and data created with these software will be locked. Therefore, they are very reluctant to be centrally managed and locked by external suppliers.

However, on the one hand, cloud outages and other incidents occasionally occur with cloud providers such as Alibaba Cloud and Tencent Cloud, which are endorsed by major internet companies; on the other hand, public cloud data is not deployed locally, making it difficult to clarify responsibilities in case of accidents. Locally deployed clouds are under their own control, which better meets regulatory and security requirements.

As a result, in recent years, while the public cloud market has shrunk, Alibaba Cloud's market share has also been compressed, from over 30% to around 20%. The market share it lost has been taken by companies with more specific significance, such as Huawei Cloud, Tianyi Cloud, and Mobile Cloud. In fields such as scientific research, higher education, government affairs, and finance, the erosion of public cloud by hybrid cloud and dedicated cloud with customization advantages is also evident.

In addition, many internet and technology companies that were originally cloud vendor customers now also choose not to pay for other public clouds. For example, ByteDance, which previously spent hundreds of millions of yuan annually renting public cloud, not only spends hundreds of millions of yuan building data centers every year but has also incubated a cloud computing company called Volcano Engine.

When migrating off the cloud before, David emphasized that cloud service providers always like to create a bunch of new concepts, such as "on-demand computing," which sounds cool and seems a century ahead of "renting computers," but there is no essential difference between the two. It can only be said that their marketing skills are highly sophisticated.

Similarly, AI is currently trending like crazy, with various industries undergoing AI transformation and adding AI concepts. However, it is not clear what the specific transformations will look like, including cloud computing vendors.

Nevertheless, there are also turning points.

AI assistants can replace traditional project management tools, and while the change is only in the form of human-computer interaction, for companies managed by David, such as basecamp and Hey, they may also need to complete AI-oriented transformations to maintain their current stability. Does this mean that in response to rapid changes in business, companies need to reconsider returning to cloud operations and whether the physical servers and self-built data centers purchased at present will become liabilities in the future?

Amid the AI wave in China, Alibaba Cloud was the first to propose a "public cloud first" strategy. In their view, public cloud is a must for AI big models, which often require thousands of cards to train and require enormous computing power to support inference applications. Therefore, it inherently requires a distributed computing architecture.

Coupled with the catch-up trend of price reductions, whether it is more cost-effective to migrate to or off the cloud may require enterprises to re-evaluate when making decisions.

This article is originally created by Xinmou. For reprint authorization and business cooperation, please contact us via WeChat.