Big models tear up the database industry: Databricks and Snowflake let developers choose sides

Text: Li Hezi Editor: VickyXiao

The wind of large models has inevitably blown to the big data industry, and the smell of gunpowder is not small.

This unhappiness is approaching. Someone noticed that Databricks and Snowflake, the two "old enemies" of the big data industry, both chose the most important annual summit of each to be held at the same time this year-June 26-29.

This is something that hasn't happened in so many years.

Not only that, but they also "unanimously" moved the theme of this year's conference closer to AI. Databricks directly named the conference "Data+AI Summit", and one click on the official website is a huge "Generation AI":

Snowflake also added a great attributive to its conference, "the world's largest data, App and AI theme conference":

This shows what? It shows that the two companies are blatantly forcing the participants to make a choice: if you go to their meeting, it may be too late for us to come again.

You must know that these two companies, one in San Francisco and the other in Las Vegas, take 9 hours by car and 1.5 hours by plane. Participants want to have both fish and bear's paw, I am afraid that they will have to toss enough.

The next big thing is coming, not only the time coincides, both companies have invited important guests to give keynote speeches - Databricks invited Microsoft CEO Satya Nadella (Satya Nadella), Snowflake is the CEO of Nvidia Huang Renxun.

Everyone is a boss.

Some netizens reminded everyone that since the summits of the two companies were held at the same time, "if you haven't chosen a side yet, it's time for you to choose."

Of course, there are also many people who are ready to run on both sides. Striim product manager John Kutay plans to speak at both conferences, but will also "evaluate the distance and time between Las Vegas and San Francisco."

A netizen who hadn’t made up his mind asked him what his speech on Databricks would be, “I want to go, but I have to go to SF when the time comes, alas.”

In this regard, one netizen even had an outrageous dream, "I just dreamed that everyone was infected with the new crown at the Snowflake conference, because Snowflake didn't want us to participate in the Databricks summit..."

Both Snowflake and Databricks are currently leaders in big data analytics companies. The former was founded in 2021 and went public in September 2020. The stock price soared 111.6% on the first day of listing to close at $253.93, making it the largest software IPO in the history of the United States.

In addition, Snowflake’s shareholder list also includes well-known investors such as Salesforce and Buffett.

Founded in 2013, Databricks is currently a super unicorn in the primary market. In 2021, it has received two consecutive rounds of large-scale financing at the level of US$1 billion, with a valuation as high as US$38 billion (2021 data).

Some Chinese practitioners are more accustomed to calling it a "brick factory".

Not only are the two companies often compared by the outside world, but they also always compete with each other openly and secretly.

The most famous call is in 2021. At that time, seeing Snowflake achieve a market value of 100 billion by relying on cloud data warehouse, Databricks couldn't sit still, and published an article saying that its data lake technology set a new record in TPC-DS benchmark test.

The point is, Databricks also highlights third-party studies that show up to 2.5 times the actual performance of Snowflake.

Ten days later, Snowflake responded, releasing its own test results, while claiming that the performance comparison conclusions published by Databricks lacked completeness, and that the research itself was flawed.

The Snowflake founder also emphasized that such benchmarks are meaningless, publishing database benchmark results in this era is "turning normal technical communication into a marketing gimmick lacking integrity."

Unwilling Databricks responded again, and the founder posted a blog, this time with a more serious accusation: Snowflake actually changed the input data of TPC-DS for the test results.

Since then, the mutual choking between the two has not stopped.

In a conversation with investor Matt Turck last year, Ali Ghodsi, co-founder and CEO of Databricks, did not shy away from talking about competition with Snowflake.

He first commercially boasted that Snowflake has "probably the best data warehouse on the market", and "Databricks and Snowflake will coexist in maybe 70% of customers".

To add here, Snowflake mainly uses data warehouse technology, while Databricks uses data lake technology, which is also the main difference between the two technical ideas.

Then Ali Ghodsi promoted his own data lake, "Public cloud computing providers have an incentive to push more people to store data in their data lakes... I think the data lake paradigm will win."

The pursuit of product performance is not over yet, and now Databricks and Snowflake are secretly poking at the big model.

Databricks released an open source large language model called Dolly (supposedly in homage to Dolly, the first cloned sheep) in March of this year, saying that "for $30, a server, and three hours, we can teach Dolly starts to interact at a human level".

This is obviously aimed at products with a higher threshold like ChatGPT, which means that AI is no longer something that only large technology companies can afford. Without much financing, anyone can develop an AI that is truly human-like.

Databricks then released Dolly 2.0, an open source iterative version of the large language model, in April.

Snowflake is also constantly hyping up large models. It published an article in April stating that it is building a data-centric platform for generative AI and large language models, and explained in detail the basis for doing so and the impact it will bring .

The following May, Snowflake announced the acquisition of startup Neeva to add AI-based generative search to its data cloud platform.

Although the two companies have not yet stepped into each other's territory in terms of AI large models, the overlapping timing of this annual summit has already explained their attitudes.

Interestingly, when someone searches for "Snowflake conference" on Google, the first result is the conference link of Databricks, followed by Snowflake's.

It is equivalent to saying that Databricks bought the keywords of competitors to sell their own ads.

We also found that when Google searched for "Databricks+Snowflake" at the same time, the first two were advertisements (this result was displayed dynamically) - Databricks still advertised its achievements in the first position of the bidding ranking, but the second It is a technical service provider, and it is promoting its service to help customers migrate to Snowflake. This launch is even more interesting than the two direct bidding situations that someone discovered earlier. Databricks can still be interpreted as (in terms of commercial offensive) more aggressive. The stronger and more combative side, but in the competition between the two, some technical service providers are also looking for business opportunities, and seem to think that there is a greater need to transfer to Snowflake...

In any case, a big battle is about to start again. Now there is no shortage of spoilers in the field of large models. When it comes to Databricks and Snowflake, it may become a question of who will kill who first.