The same year Amazon launches Redshift and Snowflake is founded, Google makes its own data warehouse, BigQuery, generally available. Three of the most important moves in cloud data are all clustering around 2012, which is not a coincidence; this is the moment the industry collectively realizes that the cloud is about to swallow enterprise analytics. But BigQuery deserves its own entry, because it takes a genuinely different approach from both Redshift and the traditional warehouses, and that approach points toward where the whole industry may eventually go.
Google’s Unusual Starting Point
Google has a peculiar advantage in building a data warehouse: it has spent over a decade solving the problem of analyzing almost unimaginably large datasets for its own search engine and advertising business. BigQuery grows out of an internal Google technology called Dremel, a system Google engineers built to run queries across enormous datasets at very high speed. In productizing that internal capability as BigQuery, Google is effectively renting out the same kind of data-crunching machinery it uses to run its own empire.
That heritage shapes BigQuery’s defining characteristic: it is fully serverless from the start, in a way neither Redshift nor the old warehouses are. Serverless is a confusing term, so here is what it actually means in practice.
What “Serverless” Really Means
With Redshift, you still have to choose and manage a cluster: you pick how many servers you want, you pay for them whether you are using them or not, and you are responsible for sizing them correctly. If you guess too small, your queries are slow; too big, and you waste money on idle machines. BigQuery removes that decision entirely. You do not provision any servers. You simply send it a query, and Google’s infrastructure automatically marshals however much computing power the query needs, runs it, and charges you based on how much data the query processed. When you are not running anything, you pay nothing for compute.
The everyday analogy is the difference between owning a car and using a taxi. Redshift is like leasing a car: you commit to a specific vehicle, you pay for it whether it sits in the garage or not, and you have to choose the right size up front. BigQuery is like hailing a taxi: you do not own or manage anything, you just state your destination, a right-sized ride appears, and you pay only for the trip you took. For workloads that come in unpredictable bursts, which describes a lot of corporate analytics, the taxi model is liberating. You never pay for an idle warehouse, and you never have to guess how big to build it.
BigQuery charges primarily by the amount of data each query scans. Ask a question that sifts through a terabyte of data, you pay for a terabyte scanned. This has an elegant honesty to it, you pay in proportion to the work you ask for, though it also means that careless queries against huge tables can run up surprising bills, a lesson many data teams are learning the hard way.
Why the Serverless Model Matters So Much
The serverless approach points at a possible eventual destination for the entire industry. Snowflake’s separation of storage and compute and BigQuery’s serverless model are two routes to the same goal: freeing companies from the burden of managing infrastructure so they can focus purely on their data and their questions. It is easy to imagine even Redshift and the on-premises holdouts moving toward serverless and separated-compute designs in the years ahead, because the convenience is simply hard to resist. BigQuery gets there first, and most completely.
It also gives Google Cloud a genuine flagship. In a market where Amazon has an enormous head start, BigQuery is one of the products that gives enterprises a real reason to consider Google Cloud, and it is shaping up as one of Google’s strongest data offerings.
What It Means for the Market
BigQuery hardens the three-way cloud dynamic taking shape in the data industry. Amazon has Redshift, Google has BigQuery, and Microsoft is building its own warehouse offerings on Azure. The three cloud giants each want to own the data layer of their respective clouds, because whoever stores and analyzes a company’s data has enormous influence over everything else that company does in the cloud. Data is sticky; once your analytical data lives in BigQuery, you are far more likely to use the rest of Google Cloud around it.
This three-cloud reality is exactly why the independent players, Snowflake and Databricks, make such a point of being multi-cloud, running across all three rather than belonging to any one. Their pitch is neutrality: use our platform and you are not locked into Amazon’s, Google’s, or Microsoft’s ecosystem. BigQuery, by being so deeply and excellently tied to Google Cloud, helps define the very lock-in that the independents position themselves against. The benefit to customers, meanwhile, is another leap in accessibility, world-class, Google-grade data-crunching available to anyone, with zero infrastructure to manage and a bill that scales with use. The taxi has arrived, and the whole industry may be about to learn to stop buying cars.