AnalystANGLE with Sanjeev Mohan, SanjMo | Google Cloud Next ’24

[John Furrier]

Welcome to theCUBE’s live coverage of Google Next

Welcome back to theCUBE’s live coverage here at Google Next. We’re at day three of our wall-to-wall three days coverage. I’m John Furrier, host of theCUBE, joined with Savannah Peterson, Rebecca Knight, Rob Strecce.

Breaking down the action at Google Next

Breaking down all the action. As usual, CUBE, our flagship program, we go out to the events, extract the signal from the noise. Day three, we got the analyst angles segments and a lot of great innovators coming on.

Looking back at the first two days of Google Next

As we look back on the first two days of the show, pretty much all the big announcements are out there. We got Sanjeev Mohan here, the principal owner and CEO. Sanjeev Mohan, great analyst, deep in data, deep in cloud, deep in data, former legendary Gartner analyst and now a successful firm. And member of theCUBE Collective. Great to always have you on theCUBE. Good friend and love your work. Your research is phenomenal. You always got your finger on the pulse and getting more data. So I got to ask you, well, first of all, welcome to theCUBE.

[Sanjeev Mohan]

Thank you so much. Always a pleasure to be here.

[John Furrier]

Google Next is a data show and a cloud show

So this is data show and hiding in plain sight with a cloud show.

[00:01:03]

I mean, everything about AI is about data. This has been a great move for Google Next. Now remember, eight months ago, they had the other Google Next. Hasn’t even been a year. So we know Google’s been working really hard, but I was kind of like, not skeptical, but curious of how it would play out following on so fast because one, eight months is not a short time for pulling content together, building product, but the market was changing so fast.

Sanjeev Mohan’s take on Google Next

So what’s your take of the show? What do you think of what’s happened? How would you look at this? Give us an update on how you see how things played out. What’s the coolest thing?

Google has a full stack from hardware to applications

Give us your take.

[Sanjeev Mohan]

So John, last year when we were at this event, I wrote a blog and I created this seven layer OSI stack and I basically mentioned how Google has a full stack and is one of the very few companies that does from all the way hardware to applications at scale.

[00:02:00]

This year, that story came together. We actually saw, I strongly feel that Google Cloud has thrown the gauntlet and I know I’m hearing that a lot of other competitors are having sleepless nights because what Google has done is they brought, so the entire stack is AI infused and I don’t want to over index on AI, Gen AI, everything is about AI, but because now they own their own multi-model foundation model, so it is so deeply integrated. They can handle text, they can handle audio, they can handle video, all in real time and we saw a lot of demonstration of that and now they have a whole build process for agents.

Google Cloud has a unified stack from TPUs to agents

So right from their own TPUs, their own new chip that they announced, all the way up to agents, they have a unified stack.

[00:03:02]

[John Furrier]

Google Cloud’s BigQuery with vector embeds is a game changer

I think that’s a great point and it’s very nuanced and I’m glad you brought that up because what’s not obvious in some of the coverage and the other analysts out there and the news is that they’re covering all the sexy announcements. Oh, the new processor, which by the way is big news. Custom silicon, you got to have that and we had them on theCUBE. But the BigQuery with vector embeds, with the multi-modal and the fact that it connects to other data sources and other clouds is a huge game changer because Google’s basically saying, hey, it’s a multi-cloud world, it’s a super-cloud world, we’ll connect, but we think it’ll end up in BigQuery anyway because we’ll have the better product, which is unique. So now you got a product that takes away all that complexity. So having BigQuery, having all that in one place is new for Google and it’s a game changer if they can pull it off. Now again, a lot of this stuff’s in public preview, so we have to still see the meat come on the bone, so to speak, but directionally, powerful.

[Sanjeev Mohan]

BigQuery Omni is a cross-cloud materialized view

So let me ask you a question. You’ve had a number of speakers come here.

[00:04:01]

Has anyone mentioned BigQuery Omni? No. No, I was expecting that answer. They did not. Because it’s not AI and no one wants to talk about it. You know what they announced in BigQuery Omni? That did not even make the headlines, that you can have, BigQuery Omni lets you run BigQuery not just in Google Cloud, but in other cloud providers. They announced a cross-cloud materialized view, so you can query BigQuery on AWS much faster and cheaper. So this cross-cloud materialized view is a big deal, but in all this noise about AI, no one even talked about it.

[John Furrier]

Well, we did talk about the BigQuery fine-tuning with Vertex, that was cool.

[Sanjeev Mohan]

BigQuery is developing a unified architecture for streaming and batch

Yes, and that is actually, okay, so let’s talk about that. So this is how BigQuery is developing. First of all, I’m a huge believer in having a unified architecture for streaming and batch.

[00:05:06]

Because we are moving to streaming, we want to be able to run intelligence on the real-time data via stale data. So now, the way BigQuery works is, let’s say a new document is loaded, that document gets into their cloud storage, BigQuery gets notified, immediately it starts vector embedding it, I can run my vector search. So a lot of companies have it, but if an audio file gets uploaded, I can transcribe it in real-time, I can store the transcription in BigQuery, embed it, and I can do vector search.

[John Furrier]

Vector search is a huge deal for BigQuery

The vector search is a huge deal for multiple reasons. One, it makes BigQuery a much stronger solution because it’s got unstructured data, and you get the retrieval, it makes search available, which is going to game-change. And combined with taking in other documents and reasoning them, it’s going to be a huge game-changer to ingest, forms, procurement, every vertical will be impacted.

[00:06:05]

So there’s going to be some goodness there. But also, there’s an industry perspective, Sanjeev, and remember we said on theCUBE, I think it might have been last year at MongoDB Local in New York, that this whole vector thing is really powerful, but it’s not a company, it’s a feature. And I think Google’s, again, another company that’s got vector, it’s a feature of something bigger with BigQuery, and so Mongo has vector search, so vector search is not a company, it’s a feature. And so here you go, well, again, another checkbox, BigQuery with vector, I can run it there.

[Sanjeev Mohan]

Google is doing more than just vector search

Yes, but there’s a difference in what Google is doing. It is not just vector search. If it was just vector search, it’d be fine. So the story I was telling about documents coming, audio comes in, now a video comes in. So you can start tagging that video and embedding it. Let’s say it’s an insurance company that wants to look at not just claims, but they also want to look at the video.

[00:07:01]

But in the video, there’s a license plate. That license plate should not be visible to the general public, so you can obfuscate that.

[John Furrier]

Google is using computer vision for multi-modal data

Yeah, you need computer vision. Yeah, you need to have multi-modal.

[Sanjeev Mohan]

Correct, so there’s vision API. So it’s no longer just about search. You can take that data that you brought in, embedded it, and you can fine-tune a model. You can train a new model.

[John Furrier]

No, no, this is a good point. Let’s expand on this. I think this is a great point you’re bringing up. So what you’re saying is, and this is really, I want to just put it on the table. It’s not just search. Bring it in, do some retrieval, augmentation generation, which is unstructured data. It’s multi-modal, meaning you’re creating the ability to address data so that at runtime, when you’re generating answers or reasoning, you can do it. So it’s not just text. You have to look at all the things. So for example, I’ll give you another example. Multi-modal, we were talking about some of the biometric stuff, I mean biology stuff around healthcare and health sciences, and DNA is stored.

[00:08:06]

That’s a mode. Modal, too. DNA is multi-modal. It’s not text. So multi-modal is a really important concept because a license plate in a form is an image. A video has audio. So this is just all about making it automated.

[Sanjeev Mohan]

Google’s multi-modal approach can be used for DNA analysis

So see, let’s talk about DNA. This is all, I’m just literally thinking on my feet here. You know, CRISPR has been such an amazing technology, lets you do gene editing, and now you can simulate, you can try to find cure for cancer and Alzheimer and all that. What if all that information is hiding in plain sight in 20 million documents from New England Journal of Medicine, CDC, WHO, with generative AI, I can now find these relations, these semantic search that are similarity, like distance close to each other.

[00:09:05]

So this is where I see Gen AI having the biggest impact. Not summarization, great. Not call center automation, IVR. Easy stuff. Yeah, but in finding what is in my 20 million, 40 years of unstructured data, and then finding new cures.

[John Furrier]

Google’s Gemini 1.5 has a context window of one million tokens

Yeah, and so that brings up the whole context window. Gemini 1.5’s got one million tokens. Correct. Remember Jensen at NVIDIA, and again, good point about Jensen not being here. There’s no Jensen here.

[Sanjeev Mohan]

Right, so. Yeah, I was telling you that before. We didn’t talk about that. See, I think this is a, if you look back, every conference you and I went to, there was obligatory presence by Jensen, by CEOs of other companies. Who came to this conference? Nobody. We had videos of Uber CEO and customers, but they didn’t come on the stage, and this is the power of Google.

[00:10:06]

Google is giving customers the whole package, simplified and unified

What Google is saying is that we are giving you the whole package, simplified, unified, but we give you optionality. You can go to Model Garden. You can get an open source model. You can get Apache Iceberg, so you can keep your data on desktop, but we don’t need NVIDIA to be successful. We don’t even need OpenAI to be successful.

[John Furrier]

So at your point, again, so your summary of the show is Google put together the package. Yes. All in the stack. Correct. Bottom to top. Correct. With data infused throughout. Yes. And where AI will be introduced and scaled, and then you get the Kubernetes container piece to orchestrate it together. Correct, correct. And that’s now a full workable package for an enterprise.

[Sanjeev Mohan]

Yes, build and runtime. Runtime through Kubernetes, you know. Google Run is getting a lot of traction. Yep, yeah. A lot of these new pieces we are looking at, like BigQuery Canvas, for instance, for like natural language.

[00:11:03]

It’s all running.

[John Furrier]

Okay, so as a research analyst, you do a lot of events. Let’s just zoom out.

Assessing the steak of Google Next

Yeah. The folks watching who didn’t attend the one, maybe they kind of checked out the Twitter stream, they see all the sizzle. Where’s the steak on this show? Where’s the beef on this show? How would you assess the show? Give us a quick rundown of what you think, what the big points were, and what it means for customers looking at their cloud business transformation with AI.

[Sanjeev Mohan]

Google Cloud Next is the biggest ever

So, this is the biggest Google Cloud Next I’ve ever been to. Actually, it’s twice the size of Moscone Center by just moving here. The number of analysts who were in my two days of analyst summit, 120 from 38 different companies. So, I see this new level of confidence in Google. They think their moment in the sun has arrived, and they think they made the right bets. All these years, you’ve had people come on the show and say, well, I don’t know about Google Cloud.

[00:12:01]

Google Cloud is now enterprise friendly

They’re not really enterprise friendly.

[John Furrier]

They go to markets weak. They don’t have an ecosystem. They’re not bringing their tech to the table. I mean, I’ve been very critical of Google in a very positive way, and I remember saying years ago, if they brought that tech to the table and cleaned up their motions with customers and built an ecosystem, they’d be great. Now, this is pre-gen AI. Now, guess what? They get a lucky strike. Gen AI comes, and they got a user interface with Workspace. They have big iron backend scale, just add GPUs and TPUs, and then they’ve been working on Kubernetes for 10 years.

[Sanjeev Mohan]

Google has a world-class infrastructure

Yep, so Kubernetes is a big thing. All the infrastructure that they’ve laid out, under submarine cables and data centers, that’s world class, because YouTube, search, ads, they all run on that at planet scale, so no question about the hardware. Security, this is a week we are hearing about what happened to Azure.

[00:13:01]

Google has reported zero security incidents

Do you know how many security incidents Google has reported? Zero.

[John Furrier]

They had one little incident in public sector, I don’t know what Google’s, but it was kind of weird on the side, but it was a nation-state attack, so it was definitely different, but I would agree with you, and Dave and I called this on theCUBE years ago, and we said Apple and Google have the best security on the planet, because they have a consumer business. Google had security advancements at the biometric level, with Android hardware level, and look at Apple, same thing.

Google has end-to-end themes coming up

Now, the big thing that happened at the World Congress this year, as we get into the supercomputing AI infrastructure, is it’s devices into the cloud, so IoT or handhelds, handheld devices, so your end-to-end themes is coming up too, so to me, I want to ask you this, because the theme that’s jumping out at theCUBE is a continuation of the same theme of end-to-end workloads can be an advantage to scope those out now, lock them in, and scale them up. That has come big, so from device in, so you need the full stack to manage that in, for AI, so to your point about full stack, I’m a company, I don’t have to redo my entire IT, I can take one workflow that has an app at the end point that users use, and that whole workload can be optimized.

[00:14:16]

That’s a huge theme here, workload optimization, end-to-end.

[Sanjeev Mohan]

Google is using AI for workload optimization

They’re using AI both inside and outside, inside being workload optimization. FinOps has been a very big topic, I don’t know if you’ve had anyone talk about FinOps?

[John Furrier]

No, not yet, not here, not today.

[Sanjeev Mohan]

So, there’s an entire FinOps division, they have launched a bunch of new features for FinOps, a new single pane of glass, so there’s a lot of movement going on. So John, I really feel, personally, that the keynote did more on agents than, it did so much on agents that all this other amazing stuff got hidden.

[John Furrier]

Google had too many announcements at Next

Yeah, well they had too much announcement, like always, they were trying to check the box.

[00:15:02]

Like AWS event, there’s so many announcements. I mean, I heard from the whispers in the hallway, there was over 600 announcements that had to be pared down to 200, they just couldn’t get the volume out, so there’s a lot more.

[Sanjeev Mohan]

In eight months, 600 new innovations in eight months.

[John Furrier]

Google pulled together a lot of content in eight months

So I give Google really high marks on this event, one big event, they pulled the content together, again, all this work they’re doing in eight months, and overlay that on the industry change that’s been significant. So it’s been a moving train on the rapid change in the industry combined with getting this word out. So they built on ModelGuard and they got ModelBuilder and then AgentBuilder, so they got now 130 miles in Vertex, Gemini 1.5, I’ve been playing with it, it’s pretty good. It’s got cross-modality analysis and reasoning. That, to me, is a huge deal, and I think that’s going to be the secret gem that pops out of the show.

[Sanjeev Mohan]

Gemini 1.5 is a huge deal

So we saw Gemini at work yesterday at the developer keynote. During the developer keynote, towards the end, they took all the video and they sent it through, so that was 627,000 tokens.

[00:16:10]

So out of a million, they used 627, but it just literally took a couple of minutes, and they were asking questions. I mean, it takes a lot of guts.

[John Furrier]

Advice for customers looking at Google Next as a viable cloud

Okay, final question to wrap up here. First of all, thank you for your time, I know you’re busy. What’s your advice to customers that you recommend you pull the data in from the show, you’re going to probably do a big report, you’re going to synthesize it, reason all the data. What’s your early directional position for the posture for customers when they look at Google Next as a viable cloud? How are you thinking about positioning that?

[Sanjeev Mohan]

Google Cloud is now a viable option for enterprises

So I think all these years that we’ve been questioning viability of Google Cloud are now behind us. I think this is the first year. I think we are now going to see an acceleration, so I would say we are already hearing how 90% of the startup unicorns already use Google Cloud. I think Google Cloud, even for enterprise, is now a very viable option.

[00:17:03]

[John Furrier]

Google’s ecosystem is performing well

I would totally agree with you. I would add one more thing to that analysis, at least from my perspective, is the doubters can, now that’s all gone, Google’s viable. Ecosystem has been really performing well. They’re standing tall with their booths, they’re paying up for their sponsorship, they’re having parties. If Google can maintain this year and make the ecosystem successful and not overdrive that piece, then it’s the last piece of the puzzle. They got to get the ecosystem 100% on board, no cognitive dissonance, because right now I’m sensing like, did I buy the right car? I love this car, it’s almost too good. Everyone has those kinds of doubts, so Google should have to reinforce, no, we are here for you. Go to market, support them, drive business through them, that will be key.

The ecosystem is the last piece of the puzzle for Google

And again, public sector, a whole nother great position that they got, so love the public sector, and I think the ecosystem’s the last piece of the puzzle.

[00:18:04]

BigQuery is the unsung hero of Google Next

And of course, I’m a big fan of what BigQuery is, I think that’s the unsung hero, because that’s the engine. The cross-modality reasoning will be the big piece of the puzzle that will make everything work, and of course, the glue up top is the orchestration with Kubernetes containers and serverless.

[Sanjeev Mohan]

Dataplex is the metadata layer for Google Cloud

And by the way, one really critical piece, when I was on KubeCon three weeks ago in Paris, on theCUBE, we were talking about the metadata layer. Everything that goes into Google also goes into Dataplex, which is, and it gets tagged and classified, and you can apply security on this, role-based access control and attribute-based access control. That metadata layer is that secret sauce to how this whole thing works.

Sanjeev Mohan’s contact information

Great to have you on theCUBE. Where can people get a hold of you? So the best way is through LinkedIn, so please feel free, follow me on LinkedIn.

[00:19:03]

Also my medium blogs and my YouTube podcast.

[John Furrier]

Sanjeev Mohan is a CUBE contributor and industry analyst

All right, Sajeeb Mohan, CUBE contributor, industry analyst. This is the space, your wheelhouse, data and cloud. This is theCUBE bringing you the analyst angles.

Day three of theCUBE’s Google Next coverage

Day three of our coverage. We’ll be right back with more after this short break.