Welcome to the Protocol Power Index, a ranking of the most powerful companies by tech industry subsector, as well as the companies best positioned to challenge them. This time: databases .

The modern enterprise must be able to fluidly collect, analyze and access its internal data. Cloud-native database companies play a key role in this process. Their products — data lakes, data warehouses and data lakehouses, to name a few — exist in a rapidly evolving slot within the tech stack, wedged between the public cloud and machine-learning services.

Those offerings may all sound similar, but they currently refer to important distinctions in intended use cases and end users. Still, the companies offering them are increasingly overlapping with one another, creating an arms race that's resulted in booming valuations and an ever-growing suite of products quickly becoming core to the industry.

But which companies have the lead right now? And which ones are challenging that dominance? We've ranked the market for you.

Ranking	Company	Score
1	Snowflake	71.32
2	Databricks	69.40
3	Rubrik	64.13
4	Teradata	55.0
5	Dremio	47.15
6	Cloudera	47.0
7	Cockroach Labs	45.24
8	Exasol	26.51
9	Panoply	26.46
10	Yellowbrick	26.02

Explore the Data

The Protocol Power Index is designed to view power through a holistic lens that reflects how modern tech companies amass and exercise their strength. To do so, the Power Index takes into account 30 metrics across five categories — Economics, Leadership, Innovation, People and Politics & Policy — and synthesizes them into a single Power Score. Read our full methodology statement here .

What happens next?

Three forces stand to shape companies operating in the database space over the coming years: a blurring of database technologies, a demand for the technology despite a shortage of talent to use it and the ever-looming threat of public cloud vendors.

Momentum Ranking	Company	Score
1	Dremio ↑	96.47
2	Snowflake	82.35
3	Cockroach Labs ↑	75.29
4	Databricks	70.0
5	Exasol ↑	55.29
6	Rubrik	52.94
7	Teradata	38.82
8	Panoply	32.94
9	Cloudera	22.35
10	Yellowbrick	16.0

Product types are set to blur. It may be tempting to write off database terminology as marketing jargon — is a data warehouse really that much different from a data lake? And what's a "data lakehouse" anyway? — but it refers to important distinctions in intended use cases and end users.

• Let's start with the difference between data warehouses and data lakes: Relative to data lakes, data warehouses tend to contain more structured data and be more readily queryable. Business analysts often use data warehouses to build dashboards using historical data; data scientists might tap into data lakes to build predictive machine-learning models that require months of development time.
• These definitions make meaningful distinctions today, but that's changing. Databricks and Dremio have started popularizing the "data lakehouse" concept. "The lakehouse is a technological breakthrough that enables you to do both in one place," Databricks CEO Ali Ghodsi said at a Protocol event in June 2021. "So all your data in one place on a data lake and have [business intelligence] to ask questions about the past and AI to ask questions about the future. So it sort of revolutionizes what organizations are doing with data and AI."
• Even if companies don't adopt the "data lakehouse" nomenclature, data lakes will continue to look more like data warehouses, and vice versa. For instance, Snowflake recently added Java and Python support in an attempt to appeal to data scientists, who tend to favor less-structured databases over its SQL-friendly data warehouses . On the flip side, Dremio recently added SQL functionality to its data lake platform, allowing end users to start an SQL query without first moving less structured data into a data warehouse. The convergence of product categories will result in more intense competition; whereas there's now room for several medium-sized fish in medium-sized ponds, soon those ponds will be converging.

Software will be shaped by availability of expertise. Many enterprises still lack the internal resources required to make the most of new database technology. Open-source database standards could help remedy the situation, allowing data scientists to tap into available code repositories and more easily migrate between jobs.

• Data analysis skills are still highly sought-after by U.S. tech employers. This comes even after the prevalence of data scientists in the tech employee pool more than tripled between May 2015 and May 2021. "It's not a technology challenge. It's a talent challenge," Whirlpool CIO Dani Brown told Protocol earlier this year.
• Hiring data scientists is only a first step. Before they can start generating insights, data scientists must get acclimated to a new database system: Even a veteran data scientist would need to become familiar with a new operating environment if they switch from a job using Snowflake to one using Dremio, which can take several weeks.
• Some big players in the database industry — most notably Databricks — are pushing for open architectures to help address this learning curve problem. Vendors including Tableau, Tencent, Alibaba and Informatica have started using and contributing to the open-source Delta Lake architecture supported by Databricks. The standardized architecture could also help bolster online code repositories, like those on GitHub, that can reduce the workload associated with complex database projects.
• However, an industry migration to open architectures is no sure thing. Snowflake product head Christian Kleinerman told Protocol this year that open-file formats "limit innovation in the long run." The company explained its stance in a March 2021 blog post: "It's not about rejecting open; it is about delivering better value for our customers. We balance this with making it very easy to get data in and out in standard formats."
• Snowflake has the ability to maintain this stance because it is already so large. Smaller companies in the database space will be more likely to adopt open architectures, since they reduce the learning curve associated with their products. Open architectures might also allow for greater interoperability and collaboration within an organization, because employees wouldn't need the same software licenses to interact with a dataset.

And the threat of dominant public cloud vendors looms large. It dominates the stack both immediately above and below vendors like Databricks and Snowflake: Amazon, Google and Microsoft command dominant positions in both the public cloud and AI/machine-learning markets.

• Ultimately, these public cloud vendors could be in a position to squeeze the independent database vendors from both directions. The firms could use bundling or preferential ecosystem access to perform this squeeze.
• The AI layer of database platforms in particular has become an important source of differentiation. This could play to the strengths of Amazon, Google and Microsoft, which all have their own database platforms and all offer robust AI systems . Amazon offers S3, Athena and SageMaker; Google offers BigQuery; and Microsoft offers Access and Azure ML.
• For now, though, database vendors benefit from their independence from public cloud vendors. Their platforms exist on a software stack layer above the cloud, which gives enterprises greater leeway to adopt them regardless of their cloud vendor setup.

Methodology

To rank the competitors, we've developed a formula that encapsulates 30 criteria. Those criteria span five groupings that factor into power: Economics, Leadership, People, Innovation & Politics and Policy. We then developed two systems for weighting the criteria — one for measuring power and the other for measuring momentum — such that companies can be scored on a 0–100 scale. Read our full methodology here .