The statements says mainly "That’s why Snowflake is acquiring Neeva, a search company founded to make search even more intelligent at scale. Neeva created a unique and transformative search experience that leverages generative AI and other innovations to allow users to query and discover data in new ways."
Was Neeva providing useful tooling for this kind of search? I am unfamiliar with Neeva, never used it before. Is this a really useful thing or an acquisition for the investors for a startup to make sure their fund meets a return target?
As a Googler…way I see it, too big too quick too much google and wrong time. Launch a year later and you’re legit have some rocket fuel via LLMs to build something new.
I never used it till it shut down, and it looks like about 70% of google to me. That’s great, but you gotta avoid the xoogler trap of rebuilding the 20% of google you want to fix as a startup.
I agree - Neeva was just Google-that-you-pay-for, which might've worked as a niche... but then LLMs came out and afaik they didn't do anything to integrate with them.
From your statement, it seems reasonable to assume that the people and their experience are what Snowflake is actually acquiring. (Another comment here suspects a soft landing for investors as both companies are Sequoia backed.)
We're just free associating, only like 60 people know, I presume. And life just doesn't work like that, like, some of those 60 people will fervently believe one interpretation, another will believe in another. We'll know more in 5 years
I think I follow, but then does that mean their product was an inferior substitute-able service for Google? What was the innovation that made them stand out more? I understand that information retrieval space has entered a new realm of hype-vs-reality so one cannot discount as you say the LLM rocket fuel they missed. But why did they miss it? It is happening right now and even open-source models may be sufficient for a search startup (or use OpenAI or Bard as a default for now until specializing themselves).
I really wanted to like using Neeva because I want to pay for a good search engine, but the results were always bad and I had to go to Google 95% of the time anyways.
Xoogler understanding of privacy, trying to woo non-xoogler privacy fans
Google has a "collect a lot but don't sell it" approach to data, and Neeva had similar vibes (only giving "basic" search unless you opt in to having your data collected but its ok its private collection)
Big mismatch with its main audience, and that mismatch was raised early, but ignored.
The stock market reacted quite positively to this acquisition (Basically confirmed a few days ago). Buying a soon-bankrupt startup is very cheap, and getting a strong tech stack and already formed team with strategic synergies with the main business is going to be valuable.
This is contrasted with panic acquisitions like say Adobe & Figma.
And so, if this is the first time you have heard about Snowflake acquiring Neeva you now know the level of asymmetry of information that exists in the 'public' market.
Use this as a lesson to exercise caution when personally investing in individual stocks in the public markets - even if you are buying tech stocks and feel as though you follow tech trends closely because you work in the industry.
Wall Street already knew and priced it in before you did.
Which is totally fine, unless you’re day trading, which you probably shouldn’t be if your goal is to maximize returns.
Day to day and month to month ups and downs don’t matter if your investment time horizon is long.
The dips are actually great because they create opportunities to buy more of the stock cheaper, assuming you’re not planning to sell it all in a few days.
Traders may have heard this from The Information: "Snowflake in Talks to Buy Search Startup Neeva in AI Push" from 5/17, then "Snowflake Nears Acquisition of Search Startup Neeva" from 5/19.
Okay, but if we assume last week's small rise is attributable to the confirmation of the Neeva acquisition, that seems more like a "market reacted tepidly" to me.
Snowflake seems like it's been losing relevance ever since Clickhouse became more popular. Seems like they're struggling to maintain performance vs the other competitors out there. Not sure how this acquisition will help here. I don't think decoupling storage and compute was a good bet in the long run.
Thats interesting, i don't see these as occupying the same space. Clickhouse is in the space of realtime analytics and Snowflake is a data warehouse. Although you could use Clickhouse for similar things it will fail at doing large distributed joins and similarly Snowflake will have trouble meeting a subsecond SLO.
also FWIW Clickhouse's cloud offering also decouples storage and compute using an object store, but they found a good middleground where they keep local caches of hot data.
But CH is capable of the same “data warehousing” features that snowflake is. Which leaves snowflake as a slower, less capable, less open, and more expensive alternative.
Which brings me to the next point: I’m convinced the delineation between “data warehouse” and “olap” is largely a marketing move designed to segment the market along made up boundaries.
We use both MS SQL and Snowflake heavily. There are clearly instances where having row based storage is appropriate, and also instances where columnar storage outperforms. All based on your workload and not just marketing.
MSSQL is an OLTP based db (going to preclude discussion of its fancy column index stuff it’s capable of). OLTP db’s definitely, definitely have a different role.
I’m talking about the false difference between the likes of ClickHouse and Snowflake, where they’re both column oriented already. I’m asserting that the fundamental differences between “classic” column db’s and “data warehouses” is far less fundamental than the marketing would have us believe. Some of the db’s in this space have slightly different architectures and trade offs, and some deliberately operate at different scales, but they are built for, and operate in, basically the same purpose.
Snowflake and ClickHouse are very different in their focus.
Snowflake is focused on enterprise customers. It has a lot of features focused on that, like very granular security and governance and data marketplace. There's also some non-enterprise features that ClickHouse lacks, like the ability to execute Python in database (so you can bring ML in).
But the biggest difference is that Snowflake is storage segregated architecture. Scaling Snowflake is done by running "alter warehouse resize" or something. You can also dedicate specific compute slices to specific users and scale them up and down as needed. And this is all managed for you.
If you want to run ClickHouse at scale, you have to run your own k8s, figure out how to manage persistent storage, figure out how to replicate your data, manage cluster replicated tables, etc. Once you outgrow single instance, things get exponentially more difficult - both for the admins and for the users.
Also, while ClickHouse can do joins and is getting better and better optimizer as we speak, and is probably faster than Snowflake for the same money on "single big table analytics" kind of workload, I would expect it to perform much worse in traditional analytics queries, like you would find in TPC-DS.
Yeah, that would be my answer as well. I actually forgot to mention that - Snowflake and the like store data away from compute so no matter how you misconfigure clusters (though Snowflake isn't really that configurable) the data is safe. Messing up database that stores data locally means the data is gone - and that makes all operations like resizes and upgrades much more scary.
But of course the local storage is much faster. Tradeoffs.
I know ClickHouse Cloud uses S3 as well, but I don't know much about it, so I don't want to comment on it.
> If you want to run ClickHouse at scale, you have to run your own k8s, figure out how to manage persistent storage, figure out how to replicate your data, manage cluster replicated tables, etc. Once you outgrow single instance, things get exponentially more difficult - both for the admins and for the users.
This greatly overstates the difficulty of running ClickHouse as well as the current state of the market.
1. ClickHouse has a good Kubernetes operator written by Altinity that manages most of the basic Kubernetes operations. It's used to operate many thousands of ClickHouse clusters worldwide both self-managed environments as well as multiple SaaS offerings of ClickHouse. (Disclaimer: it's written by my company.)
2. If you don't want the trouble of running ClickHouse there are now multiple cloud vendors in every geographic region offering ClickHouse-as-a-Service. Among other things competition keeps prices reasonable and ensures plenty of choice for users.
There are real differences between Snowflake and ClickHouse but ease of operation is no longer one of them. For example one major difference between Snowflake and ClickHouse from a user perspective is the following: You can develop great Snowflake applications just with a knowledge of SQL whereas for ClickHouse you really have to know how it works inside.
I think it's more borne of the lack of scaling capabilities in the traditional sql databases, and I guess a lack of capability in summarising data.
In reality, you can probably scale something like vitess pretty far, and then by adding your own summary tables on top, you're probably good for most usecases.
I'm not an expert on this level of the stack though, so I'm probably missing a whole bunch of context.
don't disagree with what you said, but your Google Trends argument has a big asterisk against it - right in the page it says "This comparison contains both Search terms and Topics, which are measured differently. LEARN MORE"
As a data point, if you examine something more granular and trend/topic tied, like Snowpark (which is close to Clickhouse alone) or "Snowflake Table" I would propose the overall point being made kind of stands.
The original term is ambiguous (I wish Snowflake had different branding) but more specific terms to Snowflake still rank high and are maybe less wonky of a comparison.
Sounds like that's a you issue, not one for Oracle.
Don't mean to sound dismissive but that what your post reads like, jut because I've never encountered a brown rat does not mean it's not the most populous animal species on earth
It's extremely possible to have never run into an Oracle DB in an entire career (in the depts you worked in), and moreover it's quite possible to use one database in finance and another in engineering or operations (Postgres or cloud). It merely means you haven't worked at the type and size of organization that tends to license Oracle, or more specifically only in some depts. And sometimes the org didn't voluntarily pick Oracle for technical reasons, it was mandated by the end-user, or for compliance, or application stack, or Oracle's sales team beat out technically superior/more cost-efficient competitors.
None of that is denying Oracle exists.
And that isn't even an 'issue', just an observation. I imagine this used to be similar with encountering IBM DB2 or SAP or Amdahl or melamine deskphones and partitions, but I assume you wouldn't say those are issues.
Pretty sure the most populous animal species on earth would be some type of insect, probably an ant or a locust. According to wikipedia there are estimated to be over 1.4 billion insects for each human on Earth. Rats are numerous but not nearly that numerous.
What I’ve seen: most big companies have one or a few Oracle databases and hundreds or thousands of “all other DBs”, including licenses for MS SQL Server.
Corporate and education is infested with Oracle due to an army of salesdroids and large technical platform decisions being made by upper mgmt instead of infra staff.
I've also observed that Oracle stack people generally don't have experience with other platforms, so push it in whatever org they're working for.
These are pretty different products with different use cases IME. I haven't used clickhouse in production but we use Snowflake extensively and I'm a big fan of the product and the business model. The ecosystem also seems to be in sync with the needs of people building on top of Snowflake as well.
They're both data warehouses that do a great job operating on massive datasets and neither should be your primary source of truth. I guess my question to you is: Why aren't they competitive products?
What's more expensive: the data engineering staff you need to have on hand to optimize data loading and queries all to make sure your Snowflake/Databricks bill doesn't balloon out of control, or the staff to maintain your data on either cloud or self-hosted Clickhouse for equal or better query performance?
In a world of limitless VC money, one might choose the more familiar and battle-tested Snowflake dynamics every time... but the world is shifting quite rapidly, and the degree to which investment in a Clickhouse stack is much less likely to "trap" you in rapidly expanding spend on a more closed ecosystem is becoming notable.
Being in enterprise software I have heard AI/Generative AI so many times since chatgpt became mainstream. Every fortune company I have spoken to wants to get on this train.
That's not in any way to dismiss the real usefulness of the technology, it's just that in the short term it will likely be blindly applied to everything. Some of which will stick and become useful.
It seems the AI hype is quite justified, considering the massive progress that was made over the last ten years.
Edit: The Gartner diagram seems especially not suited for AI. Advanced AI is, in a clear sense, the final technology. The technology with the potential of creating new technology, including better versions of itself. Anything that is technologically achievable at all (consistent with the laws of physics) must be solvable with sufficiently advanced AI. Gartner, at best, applies to some forms of narrow AI.
Access to huge upfront principal/capital in order to build a talented team and compete against a Trillion dollar corp: gated on your connections.
A cushy exit when your ambitious plans fail: gated on your connections.
So the market is technically open but plebs do not have credentials to play at all. Or to quote someone, "The law, in its majestic equality, forbids rich and poor alike to sleep under bridges, to beg in the streets, and to steal their bread."
Not a rap on Neeva team, btw. They did nothing wrong and kudos for tackling a big problem even if they failed at it.
I was also thinking in these terms. The reality is there is virtually no possibility of starting a startup, getting VC money and making a buck without connections in your customers or investors. This is especially true with VCs.
Kind of disappoints the romantic story but it is what it is
Isn’t “connections” just another word for “networking”? If I’m an investor I want a warm introduction. There are enough scammers and fraudsters out there that a basic filter on “who can vouch for you” is important. Furthermore I like entrepreneurs that had a history in industry, working for other successful people and orgs, before they strike out on their own. That’s my personal preference. And it’s through that history that they make the connections and have people vouch for them.
It’s not some injustice that private firms raise money and then direct it to who they want. That’s just a mechanism for reducing risk in a game where everything is extraordinarily risky.
> virtually no possibility of starting a startup ... without connections in your customers or investors.
I feel this frustration is strongest with enterprise software which is usually the most capital intensive to build and requires a lot of trust to sell.
The number of people who get to make decisions about, for example, data warehousing services is already pretty small and the subset of those willing to be somebody's first customer is even smaller.
However, there are plenty of other startup opportunities outside of that space. Consumers, media people and developers are generally easier to chat with than CIOs and still need software to do what they want to do.
I'm curious what the implications would be for Snowflake if they legit didn't want the asset that a board-member was trying to pressure them into buying.
What if they didn't want Neeva but Sequoia knows Snowflake can absorb the loss for them? What's the procedure then?
Could anyone with more info potentially speculate on what a theoretical acquisition of this kind would look like for the founders?
I have to assume founders are basically wiped out entirely by preferred shares / levered shares given to investors? Staff with equity obviously wiped out as well?
Seems like a situation where early investors enter the FIFO queue to recoup as much of their original investment as possible?
Even if founders are wiped out, they are usually still paid as consultants (or employees) at the acquiring company to help facilitate the transition. Compensation for these arrangements are often (well) above market in order to incentivize the founder to stay long enough to facilitate the transition.
But yea, in general if a company is bought for less than they raised in VC, the returns usually aren’t huge.
I am both unhappy and excited. Hopefully Snowflake does something interesting that is consumer facing so I can utilize it.
But god I wish there was more innovation and competition in the premium search engine space; I was so happy with Neeva and excited for its future
Why were you happy with Neeva? I used it for several years and never saw much benefit. The sports results were quite bad. Recipes were decent. But in general felt myself unconsciously going to google for any complex queries.
Neeva was positioning itself as private search, that was high quality and was going to deliver “answers”. They looked oh so cool and useful until ChatGPT erupted on the scene and got 10billion in funding. Then Bing began to offer many of the features of Neeva minus the privacy. Now Snowflake has a team of smart kids with AI skills, and it makes total sense to accelerate their product roadmaps this way. Tough for the Neeva crew, they got caught in an innovation tidalwave.
> Now Snowflake has a team of smart kids with AI skills.
Or they have a bunch of people that wanted to work at a startup, now with golden handcuffs who have to keep going through the motions until their stock grants vest.
Similar USP now, but I thought Kagi built up to it the right way by building just a few features on search engine APIs and repeatedly asking people what they would actually pay for. They also appeared not to overstaff in anticipation of future revenue.
112 comments
[ 3.2 ms ] story [ 205 ms ] threadWas Neeva providing useful tooling for this kind of search? I am unfamiliar with Neeva, never used it before. Is this a really useful thing or an acquisition for the investors for a startup to make sure their fund meets a return target?
Is their current technology that good?
I never used it till it shut down, and it looks like about 70% of google to me. That’s great, but you gotta avoid the xoogler trap of rebuilding the 20% of google you want to fix as a startup.
It was the wrong product at the wrong time.
Google has a "collect a lot but don't sell it" approach to data, and Neeva had similar vibes (only giving "basic" search unless you opt in to having your data collected but its ok its private collection)
Big mismatch with its main audience, and that mismatch was raised early, but ignored.
This is contrasted with panic acquisitions like say Adobe & Figma.
How is Snowflake stock dropping 12% "reacting quite positively"?
Use this as a lesson to exercise caution when personally investing in individual stocks in the public markets - even if you are buying tech stocks and feel as though you follow tech trends closely because you work in the industry.
Wall Street already knew and priced it in before you did.
Day to day and month to month ups and downs don’t matter if your investment time horizon is long.
The dips are actually great because they create opportunities to buy more of the stock cheaper, assuming you’re not planning to sell it all in a few days.
also FWIW Clickhouse's cloud offering also decouples storage and compute using an object store, but they found a good middleground where they keep local caches of hot data.
but ch supports large distributed joins?..
Which brings me to the next point: I’m convinced the delineation between “data warehouse” and “olap” is largely a marketing move designed to segment the market along made up boundaries.
I’m talking about the false difference between the likes of ClickHouse and Snowflake, where they’re both column oriented already. I’m asserting that the fundamental differences between “classic” column db’s and “data warehouses” is far less fundamental than the marketing would have us believe. Some of the db’s in this space have slightly different architectures and trade offs, and some deliberately operate at different scales, but they are built for, and operate in, basically the same purpose.
Snowflake is focused on enterprise customers. It has a lot of features focused on that, like very granular security and governance and data marketplace. There's also some non-enterprise features that ClickHouse lacks, like the ability to execute Python in database (so you can bring ML in).
But the biggest difference is that Snowflake is storage segregated architecture. Scaling Snowflake is done by running "alter warehouse resize" or something. You can also dedicate specific compute slices to specific users and scale them up and down as needed. And this is all managed for you.
If you want to run ClickHouse at scale, you have to run your own k8s, figure out how to manage persistent storage, figure out how to replicate your data, manage cluster replicated tables, etc. Once you outgrow single instance, things get exponentially more difficult - both for the admins and for the users.
Also, while ClickHouse can do joins and is getting better and better optimizer as we speak, and is probably faster than Snowflake for the same money on "single big table analytics" kind of workload, I would expect it to perform much worse in traditional analytics queries, like you would find in TPC-DS.
I think they have option of standalone cluster, where all of this kinda easy to configure.
But of course the local storage is much faster. Tradeoffs.
I know ClickHouse Cloud uses S3 as well, but I don't know much about it, so I don't want to comment on it.
Signing for cloud infra also adds lots of complexity and risks.
This greatly overstates the difficulty of running ClickHouse as well as the current state of the market.
1. ClickHouse has a good Kubernetes operator written by Altinity that manages most of the basic Kubernetes operations. It's used to operate many thousands of ClickHouse clusters worldwide both self-managed environments as well as multiple SaaS offerings of ClickHouse. (Disclaimer: it's written by my company.)
2. If you don't want the trouble of running ClickHouse there are now multiple cloud vendors in every geographic region offering ClickHouse-as-a-Service. Among other things competition keeps prices reasonable and ensures plenty of choice for users.
There are real differences between Snowflake and ClickHouse but ease of operation is no longer one of them. For example one major difference between Snowflake and ClickHouse from a user perspective is the following: You can develop great Snowflake applications just with a knowledge of SQL whereas for ClickHouse you really have to know how it works inside.
See the presentation for examples: https://presentations.clickhouse.com/meetup74/ai/
In reality, you can probably scale something like vitess pretty far, and then by adding your own summary tables on top, you're probably good for most usecases.
I'm not an expert on this level of the stack though, so I'm probably missing a whole bunch of context.
In terms of measuring popularity, I love
https://db-engines.com/en/ranking
Google Trends is interesting too
https://trends.google.com/trends/explore?date=2021-04-24%202...
Disclosure: I work for Snowflake
The original term is ambiguous (I wish Snowflake had different branding) but more specific terms to Snowflake still rank high and are maybe less wonky of a comparison.
Don't mean to sound dismissive but that what your post reads like, jut because I've never encountered a brown rat does not mean it's not the most populous animal species on earth
None of that is denying Oracle exists.
And that isn't even an 'issue', just an observation. I imagine this used to be similar with encountering IBM DB2 or SAP or Amdahl or melamine deskphones and partitions, but I assume you wouldn't say those are issues.
See this chart from Gartner DBMS Market Share stack ranks - Oracle was #1 for a long time:
- https://www.linkedin.com/posts/aronthal_dbms-gartnerda-cloud...
Snowflake is now #9 on this chart.
(high res: https://media.licdn.com/dms/image/D4D2CAQGZqgH3ta2R0A/commen...)
(Formerly)
I've also observed that Oracle stack people generally don't have experience with other platforms, so push it in whatever org they're working for.
https://app.snowflake.com/marketplace/
In a world of limitless VC money, one might choose the more familiar and battle-tested Snowflake dynamics every time... but the world is shifting quite rapidly, and the degree to which investment in a Clickhouse stack is much less likely to "trap" you in rapidly expanding spend on a more closed ecosystem is becoming notable.
https://en.wikipedia.org/wiki/Gartner_hype_cycle
That's not in any way to dismiss the real usefulness of the technology, it's just that in the short term it will likely be blindly applied to everything. Some of which will stick and become useful.
Edit: The Gartner diagram seems especially not suited for AI. Advanced AI is, in a clear sense, the final technology. The technology with the potential of creating new technology, including better versions of itself. Anything that is technologically achievable at all (consistent with the laws of physics) must be solvable with sufficiently advanced AI. Gartner, at best, applies to some forms of narrow AI.
A cushy exit when your ambitious plans fail: gated on your connections.
So the market is technically open but plebs do not have credentials to play at all. Or to quote someone, "The law, in its majestic equality, forbids rich and poor alike to sleep under bridges, to beg in the streets, and to steal their bread."
Not a rap on Neeva team, btw. They did nothing wrong and kudos for tackling a big problem even if they failed at it.
Kind of disappoints the romantic story but it is what it is
It’s not some injustice that private firms raise money and then direct it to who they want. That’s just a mechanism for reducing risk in a game where everything is extraordinarily risky.
How about let's start there.
I see no statically compelling reason that should be the assumed position for most people
I feel this frustration is strongest with enterprise software which is usually the most capital intensive to build and requires a lot of trust to sell.
The number of people who get to make decisions about, for example, data warehousing services is already pretty small and the subset of those willing to be somebody's first customer is even smaller.
However, there are plenty of other startup opportunities outside of that space. Consumers, media people and developers are generally easier to chat with than CIOs and still need software to do what they want to do.
Besides, the "free market" is not a goal on its own. It's a method that is frequently the most beneficial one, but not always.
What if they didn't want Neeva but Sequoia knows Snowflake can absorb the loss for them? What's the procedure then?
I have to assume founders are basically wiped out entirely by preferred shares / levered shares given to investors? Staff with equity obviously wiped out as well?
Seems like a situation where early investors enter the FIFO queue to recoup as much of their original investment as possible?
But yea, in general if a company is bought for less than they raised in VC, the returns usually aren’t huge.
Most likely got acquired for less than the valuation the founders desired.
That's probably why the acquisition price is undisclosed.
Or they have a bunch of people that wanted to work at a startup, now with golden handcuffs who have to keep going through the motions until their stock grants vest.
Maybe they mismanaged their money and/or time or failed to deliver on the tech maybe.
https://news.ycombinator.com/item?id=36013945