Why Snowflake has become a bubble situation


The share price of Snowflake, a provider of data warehousing SaaS in the cloud, has risen much higher than expected.

How did it happen in the first place? How did this happen in the first place, and will it be possible to sustain the momentum in the future?

As a 30-year IT industry veteran, my guess is that AWS will catch up with Snowflake in the next few years. We’ll unpack why this is the case in this article.

What is Snowflake?

Snowflake is a DWH service. Traditionally, it was necessary to promote a large project to form a group of people who are familiar with DB, but now it can be easily achieved.

Of course, it is possible to build a very large scale DWH that was not possible in the past, and it is easy to change the configuration and maintain after starting operations. It won’t completely kill the existing DWH in that it requires a cloud environment, but a significant portion of it can be replaced.

It’s cheaper in terms of cost, so existing DWH vendors will have a lot of trouble. The following article summarizes the details.

That’s all. Have a nice day!

… if it is explained in this way, I don’t think you really know how much to believe in this article. So, I have decided to issue a “summary article” to supplement it.

Database embedded in Snowflake

The architecture of the DB is important when it comes to data processing. If you compare it to storage, it’s like block storage and file storage. The format of the data is determined by the transactional type, RDB, Time Series DB, etc. easy to handle.

In the case of Snowflake, it is said that it is classified as transactional type DB. And they are not only devising the transactional part, but also separating the storage from the compute.

It’s easier to understand from the above diagram. It is the same with other software, but DB is a program that processes data. In Snowflake, AWS EC2 and AWS S3 are separated and the data and compute are separated as much as possible.

Of course, when you have an RDB like this, you end up with two places with the same data.” Relational”.

That’s why we use “Transactioal” data format and processing method to increase data independence and processing performance at the same time. To put it crudely, the design philosophy is to optimize not only the part that processes the data, but also the whole process, from the part that takes in the data to the part that completes the processing, to get the performance up.

Therefore, if you have a transactional DB to do what a traditional RDB does, you can’t expect to get a spectacular effect. Also, you need an architecture that is designed for the cloud. That’s why it is valuable for very large scale DB applications that require scalability, and it is not a product that will destroy all existing DWH.

By the way, as you can see from the fact that RDB and OLTP are often used together, “transactions” are important for IT. Attempts to achieve strong performance with an integrated architecture are proceeding as a NEDO project in Japan as well.

Although the software world seems to be complete and mature, there is still a lot of room for it to transform in response to hardware.

Snowflake Advantage

Now, if Snowflake’s product architecture is so good, then the idea comes to mind that we should make something similar. It is the second-best business that some manufacturers are good at.

This is actually quite annoying. As you can see from NEDO’s project, to develop a new DB with a new architecture, you need a high level of technology and experience. People who have both are valuable, and it is difficult to develop in a short time.

And when Snowflake came out, many vendors such as Tableau and Treasure Data developed a collaboration function. That’s how a big ecosystem was formed.

Of course, although it’s called the cloud, most of it runs on AWS, and Amazon has RedShift.

However, the strength of AWS is not in its technology, but in its development system that flexibly incorporates the voices of developers. The same is true for quantum computers, for example, but it’s not good at developing products that put the most advanced crunchy technology in place.

And from the point of view of the members who form the Snowflake ecosystem, they are not inclined to respond positively to “the same old thing”.

In a tightly built world, a few percent is a big difference. However, the cloud is still a world with the potential to change, as represented by Kubernetes and others. If that’s the case, we’re going to put our resources, such as people and R&D, into securing sales in other areas rather than focusing on a few percent of it.

So for the time being, Snowflake’s position looks secure. That’s why growth potential is expected, and why prominent investors are coming to get their hands on it. And investors who like to go with the flow! Investors who like to go with the flow are rushing in, and it has become today’s situation.

Snowflake’s Fatal Weakness

Now for the foreseeable future, I explained that Snowflake will be safe for the foreseeable future. But in the long term, Snowflake has a big problem.

It’s AWS. No matter how much the technology isn’t great, it’s possible to develop a similar product if you take the time to absorb Snowflake’s talent and expertise and develop a similar product.

And the strength of AWS is the ecosystem. If similar services (products) exist, more and more vendors will gradually adopt the AWS side of DWH. The strength of Cisco, Intel, AWS, etc. is that they can practice “three years on a stone”.

And Snowflake is not a solutions vendor, but a startup that competes on technology. So the current ecosystem is not enough, we need to grow into an entity that can be treated by normal SI vendors, etc. AWS will catch up with and overtake us.

For example, SAP is the lone survivor, but Oracle has a powerful enemy in Microsoft SQL Server. If you don’t get your hands on the entire pie as soon as possible, you won’t even be in the same position as Oracle.

This is Snowflake’s fatal weakness. While the best partner for growth exists, it’s AWS. That’s why, for Snowflake, while forming an alliance with Micsoft Azure and others, it is necessary to push forward with a multi-platform strategy.

Wrap Up

As mentioned above, Snowflake has a lot of momentum in its current state. However, the cloud itself is changing, and in a few years, AWS may be catching up with it.

However, the risk of a sudden collapse, as some analysts worry, is low. If it does, the risk of antitrust violations exists with AWS today.

It looks like AWS is going to take a strategy of zipping away at Snowflake’s customers and partners.

Sei Yotsuba