Teradata ( $TDC ) was once king of the data warehouses. Can it re-gain mindshare and resume it's reign?
A thread
A thread
Let me start this thread by saying nothing has changed about my opinion of $ASTS. In my opinion, ASTS is still the most exciting company that I am aware of.
If the technology works and the company executes, they have a license to print money.
2/n
If the technology works and the company executes, they have a license to print money.
2/n
This thread is the result of analysis that I have performed and conversations I have had with fellow SpaceMob friend @thekookreport
My eyes were re-opened to Teradata when I saw Kook tweeting about Teradata.
3/n
My eyes were re-opened to Teradata when I saw Kook tweeting about Teradata.
3/n
What led to me looking closer was a tweet last week where Kook mentioned that Teradata had entered into a partnership with Amazon/AWS.
4/n
https://twitter.com/thekookreport/status/1456354748769071129
4/n
Teradata isn't some hot, new startup.
They have been in business for almost 40 years.
And this isn't a company that I just became aware of. I have been familiar with Teradata for decades and have built a few data warehouses with their technology.
5/n
They have been in business for almost 40 years.
And this isn't a company that I just became aware of. I have been familiar with Teradata for decades and have built a few data warehouses with their technology.
5/n
So when Kook started tweeting about them, I was mildly curious at first - Kook has a good track record.
When I dug a little deeper, I started liking what I found.
This won't displace $ASTS, but I believe that $TDC deserves a position in my portfolio. I do NOT view this as
6/n
When I dug a little deeper, I started liking what I found.
This won't displace $ASTS, but I believe that $TDC deserves a position in my portfolio. I do NOT view this as
6/n
a trade. I view this as a long term investment because I believe that this company is misunderstood in the marketplace, and that the price is dislocated from the fundamental value of the company and its technology.
There are some hurdles that need to be cleared.
7/n
There are some hurdles that need to be cleared.
7/n
But I think that the risk/reward makes this a worthwhile stock for the patient investor.
The purpose of this thread is to explain why I believe it is a worthwhile investment.
First, I realize that most of my readers are not familiar with data warehouses or Teradata.
8/n
The purpose of this thread is to explain why I believe it is a worthwhile investment.
First, I realize that most of my readers are not familiar with data warehouses or Teradata.
8/n
Therefore I am going to start with a brief explanation of what a data warehouse is, and why they are important.
Next, I am going to talk a little about Teradata the company itself - both history and current events.
Following will be some information about the industry and
9/n
Next, I am going to talk a little about Teradata the company itself - both history and current events.
Following will be some information about the industry and
9/n
trends.
Through it all, I hope to give people an understanding of why I believe that this is a good investment.
So what is q data warehouse?
The term "data warehouse" refers to a type of database platform that is designed to store data in a manner that makes reporting and
10/n
Through it all, I hope to give people an understanding of why I believe that this is a good investment.
So what is q data warehouse?
The term "data warehouse" refers to a type of database platform that is designed to store data in a manner that makes reporting and
10/n
analytics easier for end users. Frequently data warehouses will contain data from multiple different systems. (Either systems that contain the same "type" of data - such as financial data, or operational data. (large companies frequently have multiple systems covering the
11/n
11/n
same functional need)
Integrating data into a data warehouse helps give data a common format and semantic meaning while allowing users to deal with very large volumes of data, and they make life easier for end users at the same time.
Historically, the concept of the
12/n
Integrating data into a data warehouse helps give data a common format and semantic meaning while allowing users to deal with very large volumes of data, and they make life easier for end users at the same time.
Historically, the concept of the
12/n
enterprise data warehouse is not new.
The concept originated in the 70s.
Teradata released one of the first database engines designed to work with data warehouse loads in 1983.
Data volumes grew as hardware became more powerful. As data volumes grew, data warehouses
13/n
The concept originated in the 70s.
Teradata released one of the first database engines designed to work with data warehouse loads in 1983.
Data volumes grew as hardware became more powerful. As data volumes grew, data warehouses
13/n
became more valuable to companies.
By the 90s, warehouses in excess of 1tb started coming into existence. (I know that 1tb doesn't sound like much today. But even with modern hardware, if you are dealing with a database that measures in the 10s of terabytes, you will run
14/n
By the 90s, warehouses in excess of 1tb started coming into existence. (I know that 1tb doesn't sound like much today. But even with modern hardware, if you are dealing with a database that measures in the 10s of terabytes, you will run
14/n
into performance problems if you are using a general purpose database like Oracle RAC or SQL Server Enterprise Edition)
The way that Teradata was able to deal with multi-terabyte loads with 90s hardware is through an approach than spawned an industry and has been
15/n
The way that Teradata was able to deal with multi-terabyte loads with 90s hardware is through an approach than spawned an industry and has been
15/n
"reinvented" is "newer" technologies like Hadoop.
Teradata pioneered Massively Parallel Processing (MPP) data warehouse appliances.
An MPP data warehouse is built on top of what is called a "shared nothing" architecture.
What that means in (semi) English is that a MPP
16/n
Teradata pioneered Massively Parallel Processing (MPP) data warehouse appliances.
An MPP data warehouse is built on top of what is called a "shared nothing" architecture.
What that means in (semi) English is that a MPP
16/n
warehouse contains a lot of CPUs - too many CPUs to fit within the confines of a single machine even in an era of multi-core CPUs. (That's the MPP part)
These machines are networked to form a cluster. But the hardware of each machine is independent of each other. (thats
17/n
These machines are networked to form a cluster. But the hardware of each machine is independent of each other. (thats
17/n
the "shared nothing" part)
The approach to making queries against a lot of data fast is to take a divide-and-conquer approach.
Data for each table is spread throughout the cluster. (Distribution of data is a complex subject that I am going to skip)
18/n
The approach to making queries against a lot of data fast is to take a divide-and-conquer approach.
Data for each table is spread throughout the cluster. (Distribution of data is a complex subject that I am going to skip)
18/n
When a user (or application) enters a query to the cluster, a supervisor node accepts the query, and passes it off to all nodes in the cluster.
Each node runs the query against the portion of data stored on that node, and returns data to the supervisor node over the network
19/n
Each node runs the query against the portion of data stored on that node, and returns data to the supervisor node over the network
19/n
The supervisor node accepts data from the worker nodes, and does any final processing that is needed to combine the result sets before returning those results to the user (or application).
Through it all, large amounts of data become manageable.
20/n
Through it all, large amounts of data become manageable.
20/n
For the technical people reading this, it probably sounds a lot like using Map/Reduce against a Hadoop cluster.
There is a reason for that
Despite all the hype surrounding Hadoop a decade ago, this architecture is not new
Conceptually, the differences lie in the execution
21/n
There is a reason for that
Despite all the hype surrounding Hadoop a decade ago, this architecture is not new
Conceptually, the differences lie in the execution
21/n
The NO SQL technologies have a similar hardware platform (though typically they run on less expensive hardware than data warehouse appliances use), and developers write map/reduce jobs in procedural (or object oriented) languages like Java instead of declarative languages
22/n
22/n
like SQL.
In fact, the fact that open source alternatives to data warehouses became available at the beginning of last decade, combined with the use of cheaper hardware led to a shift in the enterprise data warehouse approach.
Relational warehouses began to fall out of
23/n
In fact, the fact that open source alternatives to data warehouses became available at the beginning of last decade, combined with the use of cheaper hardware led to a shift in the enterprise data warehouse approach.
Relational warehouses began to fall out of
23/n
fell out of favor, and open source platforms began to take over that segment of the market.
To make a long history somewhat shorter, people discovered that technologies like Hadoop were difficult to manage, that talent was hard to find, and that SQL-like languages bolted on
24/n
To make a long history somewhat shorter, people discovered that technologies like Hadoop were difficult to manage, that talent was hard to find, and that SQL-like languages bolted on
24/n
top of NO SQL platforms didn't operate as smoothly as the SQL language built into database platforms.
So newer Business Intelligence systems started returning to their relational roots, and MPP data warehouses became the preferred solution (again) for large data volumes.
25/n
So newer Business Intelligence systems started returning to their relational roots, and MPP data warehouses became the preferred solution (again) for large data volumes.
25/n
Only another change had come to the industry.
When the NO SQL platforms first started entering the Business Intelligence niche, most computing work was done on-premises.
By the time that people realized that MPP data warehouses were still a pretty good solution, a lot of
26/n
When the NO SQL platforms first started entering the Business Intelligence niche, most computing work was done on-premises.
By the time that people realized that MPP data warehouses were still a pretty good solution, a lot of
26/n
that workload had shifted to the cloud. (This was done partially because of the complexity of managing the NO SQL platforms, but also the ability for IT Managers to get budget as part of Operational Expenditures instead of Capital Expenditures more easily)
27/n
27/n
Before the flirtation with NO SQL platforms and the shift to the cloud, Teradata was considered to be the king of MPP data warehouse appliances.
Though their hardware was expensive, when you looked at the largest data warehouse implementations in the world, most were done
28/n
Though their hardware was expensive, when you looked at the largest data warehouse implementations in the world, most were done
28/n
on Teradata.
When talking to an Oracle person I worked with on an Exadata implementation in the early part of last decade, he told me that Exadta was largely a response to getting beaten up repeatedly in the market for large databases by Teradata.
(As mentioned earlier,
29/n
When talking to an Oracle person I worked with on an Exadata implementation in the early part of last decade, he told me that Exadta was largely a response to getting beaten up repeatedly in the market for large databases by Teradata.
(As mentioned earlier,
29/n
MPP data warehouse platforms use a "shared nothing" architecture. Clustered databases like Oracle RAC, DB2 PureScale, or clustered implementations of SQL Server use a shared disk architecture, and that shared disk creates a performance bottleneck for queries that return
30/n
30/n
result sets)
In evaluations of data warehouse platforms on various projects that I have worked over the years, though Teradata was often a good choice, it was often skipped in favor of other vendors because of price.
Anyway, as a result of changing technical trends
31/n
In evaluations of data warehouse platforms on various projects that I have worked over the years, though Teradata was often a good choice, it was often skipped in favor of other vendors because of price.
Anyway, as a result of changing technical trends
31/n
the stock price of Teradata has been largely stagnant during one of the greatest bull markets of our time.
Here is a monthly chart:
It's up about 67% since the IPO 14 years ago, and though there have been ups and downs, the price is about where it was in 2013.
32/n
Here is a monthly chart:
It's up about 67% since the IPO 14 years ago, and though there have been ups and downs, the price is about where it was in 2013.
32/n
So why do I think that it is worth an investment now?
First, the technical underpinnings of the database software and engineered hardware platform are still there. Those underpinnings allowed Teradata to handle the largest data volumes before technical trends started to
33/n
First, the technical underpinnings of the database software and engineered hardware platform are still there. Those underpinnings allowed Teradata to handle the largest data volumes before technical trends started to
33/n
change, and they can handle the largest workloads today.
Data volumes are growing, and analytics is becoming a higher priority throughout the world.
Second, though the technical underpinnings behind Teradata are largely the same, there are important differences in their
34/n
Data volumes are growing, and analytics is becoming a higher priority throughout the world.
Second, though the technical underpinnings behind Teradata are largely the same, there are important differences in their
34/n
product offering that make it a viable choice with the trends in technology that exist today.
If you look at AWS, Azure, or GCP, you will find that you can choose something called Teradata Vantage. (If there is interest on the subject, I will write a separate thread on
35/n
If you look at AWS, Azure, or GCP, you will find that you can choose something called Teradata Vantage. (If there is interest on the subject, I will write a separate thread on
35/n
Vantage) But the short description of Teradata Vantage is that it consists of Teradata software and hardware, just located in cloud data centers.
In other words, unlike an MPP warehouse that runs against commodity hardware, this is a complete engineered system and will
36/n
In other words, unlike an MPP warehouse that runs against commodity hardware, this is a complete engineered system and will
36/n
like you would expect Teradata to scale.
But while the approach to the problem of large amounts of data has remained the same, there is something radically different about Teradata's cloud platforms.
This is something that I never expected to see from Teradata.
37/n
But while the approach to the problem of large amounts of data has remained the same, there is something radically different about Teradata's cloud platforms.
This is something that I never expected to see from Teradata.
37/n
Teradata is now cost competitive with other offerings.
Not only has the entire capex/opex argument shifted to opex with the cloud, but the systems hosted by the cloud vendors allow pay-for-what-you-use pricing plans, elastic scalability up and down as your needs change, and
38/n
Not only has the entire capex/opex argument shifted to opex with the cloud, but the systems hosted by the cloud vendors allow pay-for-what-you-use pricing plans, elastic scalability up and down as your needs change, and
38/n
prices that are cost competitive with platforms that are far less performant.
THIS CHANGES EVERYTHING.
As I researched the current offerings from Teradata, many thoughts struck me. One of the thoughts was that the company needs to overcome it's reputation as a good, but
39/n
THIS CHANGES EVERYTHING.
As I researched the current offerings from Teradata, many thoughts struck me. One of the thoughts was that the company needs to overcome it's reputation as a good, but
39/n
expensive on-premises offering.
In analyst coverage from Morgan Stanley where Teradata was upgraded from "equal weight" to "overweight", they highlighted the same concern. Here is a snippet from the "risks" section:
40/n
In analyst coverage from Morgan Stanley where Teradata was upgraded from "equal weight" to "overweight", they highlighted the same concern. Here is a snippet from the "risks" section:
40/n
So a lot of the potential for Teradata to move from being undervalued (in my opinion) to a proper valuation is going to be dependent on sales and marketing to be able to change the perspective of the company in the eyes of people who design and build systems.
41/n
41/n
But the technology is there. The cost structure (capex vs opex) is there. The technology is ther. And even the price is there for people to choose this technology.
And when you look at at valuation of the company compared to SNOW, Teradata is very attractive.
42/n
And when you look at at valuation of the company compared to SNOW, Teradata is very attractive.
42/n
SNOW TDC
Market Cap 111.5B 5.5B
Revenue '20 264.8M 1.84B
Revenue '21* 691.58m 1.473B
43/n
Market Cap 111.5B 5.5B
Revenue '20 264.8M 1.84B
Revenue '21* 691.58m 1.473B
43/n
SNOW TDC
Current Assets 4.27B 1.04B
Total Liabilities 985.27M 1.79B
Note: For Fiscal '21 Revenue, I included the quarter ended 12/31/20 for TDC and the quarter ended Jan 21 as well as the next 2
44/n
Current Assets 4.27B 1.04B
Total Liabilities 985.27M 1.79B
Note: For Fiscal '21 Revenue, I included the quarter ended 12/31/20 for TDC and the quarter ended Jan 21 as well as the next 2
44/n
quarters for each company.
Though Snowflake is growing faster, Teradata has almost over twice the revenue of Snowflake.
Given that Teradata is valued at less than 5% of Snowflake's market cap, I think the argument as a value investment is there.
45/n
Though Snowflake is growing faster, Teradata has almost over twice the revenue of Snowflake.
Given that Teradata is valued at less than 5% of Snowflake's market cap, I think the argument as a value investment is there.
45/n
But even though there is a compelling argument from both the technical side and the valuation side, the company needs a catalyst to unleash that value.
That catalyst *MIGHT* be in place right now.
If you look at the history of Teradata from the spinoff from NCR to date, in
46/n
That catalyst *MIGHT* be in place right now.
If you look at the history of Teradata from the spinoff from NCR to date, in
46/n
2007 to date, initially the company was led by Michael F Koehler. Koehler was the CEO and President of Teradata from 2007 through 2018. He took over that position after working for NCR since 1975.
Koehler was replaced as President and CEO by Oliver Ratzesberger in 2018.
47/n
Koehler was replaced as President and CEO by Oliver Ratzesberger in 2018.
47/n
Ratzesberger was replace as President and CEO by Steve McMillan in May 2020.
So Teradata was led by old-school NCR management, and during that time, not keeping up with changing technical trends led to the loss of mindshare.
The management change after old school NCR
48/n
So Teradata was led by old-school NCR management, and during that time, not keeping up with changing technical trends led to the loss of mindshare.
The management change after old school NCR
48/n
left doesn't appear to have worked out.
So Teradata made another management change.
Steve McMillan came from the executive ranks at F5 Networks, and was in charge of Global Services.
Prior to F5, Steve worked at Oracle.
He has inherited a company with VERY comnpelling
49/n
So Teradata made another management change.
Steve McMillan came from the executive ranks at F5 Networks, and was in charge of Global Services.
Prior to F5, Steve worked at Oracle.
He has inherited a company with VERY comnpelling
49/n
technology that sits at the intersection of growing data volumes and a growing awareness of the need to gain value from data.
Not only that, technically the product is positioned to work in the cloud, and is priced competitively.
The challenge that McMillan faces is to
50/n
Not only that, technically the product is positioned to work in the cloud, and is priced competitively.
The challenge that McMillan faces is to
50/n
the customer perspective of the product that his company is offering.
But it isn't like he has to overcome a bad product reputation.
Teradata has an excellent reputation.
They just need to overcome the perception of being too expensive, and of being an on-premises only
51/n
But it isn't like he has to overcome a bad product reputation.
Teradata has an excellent reputation.
They just need to overcome the perception of being too expensive, and of being an on-premises only
51/n
solution.
Also, he is in the fortuitous position of having a product that is not only cloud native, but that can also be run on-premises.
So customers do have the option of doing things that will be beneficiql financiqally - like running development and test environments
52/n
Also, he is in the fortuitous position of having a product that is not only cloud native, but that can also be run on-premises.
So customers do have the option of doing things that will be beneficiql financiqally - like running development and test environments
52/n
on premises, while running production instances in the cloud.
Given the attractive valuation of the company and their technology, personally I think that it is worth opening a position in TDC.
53/n
Given the attractive valuation of the company and their technology, personally I think that it is worth opening a position in TDC.
53/n
@threadreaderapp unroll
• • •
Missing some Tweet in this thread? You can try to
force a refresh
