Everything Announced at Google Cloud Next in Under 13 Minutes

0:00 Welcome everyone to Google Cloud Next.

0:02 Just one year ago, we stood on this same stage and promised a new future for AI.

0:10 Today, that future is running in production

0:13 at a scale that the world has never seen.

0:16 Gemini Enterprise is now the end-to-end system for the agentic era.

0:21 Intelligence plus automation must deliver value.

0:25 To make this work, you need context.

0:29 and action.

0:30 Intelligence comes from your data.

0:33 Automation is driven by agents.

0:35 To solve this equation at scale, you need a complete integrated system.

0:41 While the agent platform is where your technical teams build and govern agents,

0:46 the Gemini enterprise application is the primary

0:50 environment where your business actually operates.

0:53 Imagine I work for a global furniture retailer.

0:56 Here's my personalized homepage where I can interact with internal

1:00 business context and external sources in a single pane of glass.

1:05 The agent gallery hosts my approved selection of agents,

1:09 including built-in ones by Google and my company's

1:12 agents like this one for price and margin optimization,

1:15 which autonomously orchestrates across agents, tools, and sources.

1:19 Now, let's see Gemini Enterprise in action.

1:23 To bring some less popular product lines back to life,

1:27 we'll ask our agent to analyze current interior design trends,

1:31 identify dead stock in our warehouse, and orchestrate a relaunch campaign.

1:37 With that one prompt,

1:38 multiple agents complete a series of actions in minutes instead of hours.

1:43 We're going to pull up a trick from your 2017 Burton US Open.

1:46 Oh, wow.

1:47 I'm sure you remember this one.

1:48 It's a throwback.

1:49 So, let's uh let's slow this down, right?

1:51 Yeah.

1:51 And to do this, we need to analyze

1:53 this frame by frame and see what Google Cloud sees.

1:56 And we're going to do this essentially with every single stop in this.

2:00 So, first, let's start with your pose.

2:03 Yeah.

2:03 So, this is what's super cool.

2:05 We built a model in collaboration

2:07 with Google DeepMind that can track you spatially

2:09 and it creates a three-dimensional essentially pose

2:12 of you from a flat two-dimensional video.

2:14 So, awesome.

2:15 Let's talk about the next thing.

2:17 Right.

2:17 These are stats powered by Gemini.

2:19 Here we're tracking your flight dynamics,

2:22 rotational velocity, and even your tuck compression.

2:24 Yeah, this is amazing.

2:25 We didn't have this before.

2:27 I can see how much time I've spent in the air.

2:30 And now take old footage that I've done

2:32 and compare it with new footage of, you know,

2:34 how much time was I in the air when I

2:36 landed the trick and when I didn't land the trick

2:37 and I can compare the two and that data is

2:39 really going to help me progress and obviously the next generation.

2:42 I am proud to announce our eighth generation TPUs.

2:46 For the very first time, we're launching two specialized platforms,

2:49 each built from the ground up for the distinct demands of training and serving.

2:55 TPU 8t is a powerhouse optimized for training.

2:59 We have redefined performance capability by moving

3:03 block scale multiplication directly inside the MXUS.

3:06 This native INFXU quantization eliminates VPU overhead,

3:10 delivering nearly three times the compute

3:12 performance per pod over previous generations.

3:15 This allows us to push the absolute limits of model flops utilization

3:19 at a massive scale and reduces the training time of Frontier models.

3:24 It leverages our breakthrough interchip interconnect

3:27 technology which now delivers twice the bandwidth

3:30 compared to Ironwood scaling up to 9,600

3:33 TPUs connected via our 3D Torus topology.

3:37 This is a 2.8x improvement over Ironwood to deliver

3:42 121 exoflops of FP FP4 compute per pod.

3:47 AT provides 2 PB of shared bandwidth memory in a single super pod

3:53 and utilizes the new TPU direct storage

3:56 to enable high-speed data transfers for managed storage.

3:59 We are extending our infrastructure leadership to general

4:02 purpose workloads as well with Google Cloud Axion.

4:05 Our Google Axion N48 compute instances powered

4:08 by a custom-designed ARM CPU deliver up

4:11 to twice better price performance and 80%

4:15 better performance per watt than comparable x86 instances.

4:18 This provides a sustained continuous operation designed

4:22 to eliminate cold starts and logic gaps,

4:24 ensuring your agents are always on and ready to respond.

4:30 We are the preferred cloud for Nvidia GPUs by some

4:33 of the largest scale and most innovative customers in the world.

4:36 Today I'm pleased to announce that Google Cloud will be

4:39 among the first to offer the Nvidia Vera Rubin NVL72.

4:43 Our new Agentic Workflow triggers have detected a tasty trend.

4:48 Midnight Swirl Froyo looks delicious,

4:50 but with any new flavor I need to know, is it safe?

4:56 Are there any hidden allergens?

4:57 Is there a market?

4:59 Where are my hungry customers?

5:01 And is it worth it?

5:03 What's the real ROI?

5:05 But first, safety.

5:07 We have thousands of PDF recipes.

5:10 I'm going to search for soy.

5:13 It's a top food allergen, and safety is non-negotiable.

5:17 Zero results.

5:18 Looks good so far.

5:20 But this recipe does contain an ingredient, base 204.

5:25 Let's check the supplier manual for B 204.

5:29 And here it is.

5:31 Base 204 actually contains soy.

5:34 A simple GEI search would miss this connection

5:38 because the information is trapped across two separate PDFs.

5:41 So, how do we digest all of this data to find those hidden connections?

5:47 Well, I need an agent that has the skill

5:50 to work with my PDF data combined with our knowledge catalog.

5:55 Watch.

5:55 Our agent helps us find hidden allergens in a recipe for Midnight Swirl.

6:01 And here it is.

6:03 Midnight Swirl contains soy.

6:05 And it's giving me a data citation, a schema called product specs.

6:10 Now this schema was generated by our knowledge

6:13 catalog working with Gemini to reason over our supplier

6:17 and recipe PDFs and extract entities like recipe ingredient

6:22 and allergen and mapping those previously invisible connections between them.

6:28 This is whis basically starts by automatically building

6:33 a dynamic inventory from your code and cloud completely agentlessly.

6:37 This means that you can see

6:40 every visibility into the environment into everything

6:43 that your teams are building using AI without

6:47 any friction with then builds the security graph.

6:51 The security graph think of it as a living map that explains

6:56 the architecture and the logic of any AI application it sees.

7:00 Here we see an example of an agent that is actually running on cloud.

7:05 It has tools to query a database and even execute code.

7:09 And you can see that whis actually flags that this agent

7:15 is internet exposed and also it has access to sensitive customer data.

7:21 Now this is security moving at the speed of AI.

7:25 You don't need any reviews with your development teams.

7:29 It just works.

7:31 To summarize, whenever you have an AI team that chips a new

7:35 product or even someone in finance that vibe codes a new agent,

7:40 now within minutes, security is able to one,

7:44 identify the agent and its architecture.

7:46 Two, automatically conduct a security review to find and validate the risks.

7:51 And three, automatically suggest the fix,

7:54 send it to the dev teams in their own native tools.

7:58 Tada!

7:58 YouTube TV customer support recently launched a voice agent to help

8:02 users with NFL Sunday Ticket and YouTube TV plan subscriptions.

8:05 It's currently live in production serving 100% of their users.

8:09 We're going to take a look at it in action today.

8:11 I'm going to call the support number now.

8:13 Thanks for calling the YouTube TV and NFL Sunday Ticket support line.

8:18 So, tell me, how can I help you today?

8:20 Hey, yeah, I've got a bunch of my friends

8:22 coming over tomorrow night for the NFL draft.

8:24 I don't have cable and to be honest, I really only watch sports.

8:28 Does YouTube TV have a plan that's just for that?

8:33 Yes, we do.

8:34 The YouTube TV Sports Plan has exactly what

8:37 you're looking for, plus over 30 other sports channels.

8:41 It's $18 less per month than the base plan.

8:44 Would you like me to text you a link

8:46 to the number you're calling from that will take you direct?

8:50 Go ahead and send me a link.

8:53 I'll take a look at it.

8:56 Great.

8:56 I just sent the link to your mobile phone.

8:59 Is there anything else I can help with today?

9:01 Uh, yeah.

9:02 I'm going to take a look at this now.

9:03 Um, you know, actually, my father-in-law is hanging out with me here today.

9:06 He's been thinking about getting the sports plan for his account as well.

9:11 I just wanted to see what he thinks about this.

9:15 Actually, could you just quickly summarize this plan for him in Spanish?

9:18 I think he just wants to know that he can get both football e.

9:23 Today we're introducing workspace intelligence.

9:28 With the advanced reasoning capabilities

9:32 of Gemini and state-of-the-art embedding models,

9:36 we're eliminating context fragmentation across the workspace suite.

9:40 Remember that furniture rebrand you saw earlier?

9:43 Let's say I'm the regional distributor and here I am in Google Chat.

9:48 This is where I collaborate with my colleagues

9:51 and now my agents all in one place.

9:54 I'm here in the regional manager chat space.

9:57 And look at this.

9:59 The regional operations agent which we built in Gemini Enterprise just

10:04 alerted the team that our new display kits are arriving early.

10:08 And as you can see, I have multiple chats blowing up.

10:14 What do I need to do next?

10:17 With Ask Gemini, Google chat becomes my command center.

10:23 Look here, Gemini tells me exactly what matters right now.

10:30 See, it surfaced an urgent task.

10:32 It linked the pitch deck that I need to localize,

10:35 and it flagged my 4pm deadline for the regional plan.

10:39 The gathering is done, and I haven't opened a single extra tab.

10:44 Now, to build this plan, I remember we had a great chart

10:49 that showed regional sales last quarter somewhere.

10:51 Normally, this is where the hunt begins.

10:55 I jump into a folder and stare at dozens of files.

11:01 Instead, I'll ask Gemini, find the merchandising playbook from last quarter,

11:08 the one with a chart showing regional sales.

11:11 This isn't just keyword matching.

11:13 Workspace intelligence understands the context of my meetings

11:17 and the content inside my files.

11:20 Here it is.

11:21 It pointed me directly to the doc with the exact graph I need.

11:26 Workspace intelligence is the end of the context tax.

11:30 It transforms how you work by turning

11:33 fragmented information into a clear path forward.

11:36 No complicated setup.

11:37 It's secure.

11:38 It's integrated.

11:39 It just works.

11:41 We believe the future of AI must be open.

11:45 While others want to lock you into Wall Garden that owns your models,

11:50 your data, and your agents, we offer you an integrated stack.

11:54 But the freedom to choose the world's best chips and models.

11:58 The freedom to run AI wherever your data may live.

12:02 The freedom to control your own destiny with deep governance features.

12:07 We scale this mission through our partners with whom we're building

12:12 a broad and deep network of forward deployed engineers including Accenture, BCG,

12:18 Deloitte and McKinsey who've announced major expansion of their Google,

12:23 Gemini AI practices along with AIEL service partners like Quantium,

12:30 Distill and Tribe.ai.

12:32 We're helping independent software vendors and SAS companies

12:36 transform their solutions with Gemini Enterprise Agent Platform.

12:40 And we're bringing AI to small to medium-size businesses by helping them adopt

12:47 Gemini Enterprise and our AI advances

12:50 in workspace that work so seamlessly together.

12:53 Half.