Opening Keynote: The new way to cloud
[Music] [Music and drumbeat] [Cheers and applause] Why not? Just two words, but those two words challenge everything and can change anything. Why not help find a cure? Bring it here, and even there. >> Wait, really? >> Yes, really! We're building the most helpful AI so you can turn an idea into an enterprise. Get the right crops into this box, and breakfast on the table. Inspect 800,000 packages a day. And help protect our power grids. Because once we turn this into that, we ask what else can we do? Find out where the wild things are? Uhhh... wilder.
Spot patterns in crime data? Catch phishing attacks? Take a thousand customer service calls an hour? Help coders, well, code. Let's make it happen. >> ANNOUNCER: Please welcome CEO of Google Cloud, Thomas Kurian! [Cheers and Applause]. >> THOMAS KURIAN: Wow.
Hello, everyone! Welcome to Google Cloud Next! Just one year ago, we stood here and talked about the future of AI for organizations. Today, that future is being built by all of us. In 2024, we shipped more than 3,000 product advancements across Google Cloud and workspace. We expanded Google Cloud to 42 regions, including Sweden, Mexico and South Africa, and are rapidly expanding to countries like Malaysia, Thailand and Kuwait. We expanded our 2 million mile terrestrial and subsea fiber network by announcing new subsea cables like Umoja, Bosun and Proa. Google AI momentum is exciting. We're seeing more than 4 million developers using Gemini, a 20 times increase in Vertex AI usage last year, driven by the strong adoption of Gemini Flash, Gemini 2.0, Imagen 3.0, and most recently, Veo, our advanced video generation model.
And over 2 billion AI assists monthly to business users right within Google Workspace. But even more exciting is the momentum with you, our customers. Here at Next, we'll be sharing over 500 customer stories show casing real business innovation impact from AI adoption. Google is building for a unique
moment. We're investing in the technology and the ecosystem to power your growth and transformation. Let's hear more from a special guest.
A warm welcome for the CEO of Google and Alphabet, Sundar Pichai! [Cheers and Applause]. >> SUNDAR PICHAI: Thank you, Thomas. Good to be with you all here in Vegas. Last year, I joked in my remarks
about how I was auditioning for the Sphere. Well, turns out, I got the gig. Last night, I was on stage at the Sphere to share a new collaboration. We are introducing the Wizard of Oz to a new generation using Google AI. Transforming one of the greatest films of all time for one of the largest screens in the world is a huge ongoing effort, and not something we could have attempted even 18 months ago.
Shows how rapidly technology is evolving and how it can enable us to rethink what's possible. I think that's a fitting them for Cloud Next. The chance to improve lives and reimagine things is why Google has been investing in AI for more than a decade. We see it as the most important way we can advance our mission, to organize the world's information and make it universally accessible and useful. With Google Cloud, we see AI as the most important way we can help advance your mission.
The opportunity with AI is as big as it gets. That's why we are investing in the full stack of AI innovation. Starting with infrastructure that powers it all.
We are making big investments now, and for the future. In 2025, we plan to invest around $75 billion in total capex. This investment will be directed towards our servers and data centers, which includes powering our AI compute and cloud business. So this will greatly benefit our customers, like all of you. We need our infrastructure to move at
Google speed, with near zero latency, supporting services like Search, Gmail and Photos. For billions of users worldwide. And we use it for training our most capable model, Gemini. Google's backbone network is unparalleled, as Thomas just mentioned, spanning more than 200 countries and territories, powered by over 2 million miles of fiber. Today, I'm pleased to announce
that we are making Google's global private network available to enterprises around the world. [Applause]. We call it Cloud Wide Area Network, or WAN. Cloud WAN leverages Google's planet scale network.
It's optimized for application performance, and delivers over 40% faster performance, while reducing total cost of ownership by up to 40%. Companies like Citadel securities and Nestle are already using this network for faster, more reliable solutions. And it will be available to all Google Cloud customers later this month. This builds on our legacy of opening up our technical infrastructure for others to use. We do this with our custom AI chips, called Tensor Processing Units, or TPUs. Since 2013, we've invested heavily in this specialized hardware.
And we continue to make massive improvements in performance and efficiency at scale. Today, I'm proud to announce our seventh generation TPU, Ironwood, is coming later this year. [Cheers and Applause]. Compared to our first publicly available TPU, Ironwood achieves 3600 times better performance. An incredible increase. It's the most powerful chip we have ever built and will enable the next frontier of AI models. [Applause]. In the same period, we've also become 29x more energy efficient. And Amin will share more later today. This progress is laying the
foundation for breakthroughs across multiple fields. Quantum computing is a great example. Our newest quantum chip, Willow, cracked a key challenge in quantum error correction that has eluded researchers for three decades. It can reduce errors exponentially as we scale up using more qubits. The Willow chip really paves the way for a useful, large scale quantum computer down the road.
Our infrastructure enables the next layer of the stack, research and models. Over the last decade, our research teams have pushed the boundaries of AI forward. And today, they are accelerating science and discovery, from our AlphaFold breakthrough with protein folding to WeatherNext, our state of the art weather forecasting models. World class research is what enables us to push the frontier with our Gemini models. In December, we introduced Gemini 2.0, with new advances in multimodality, like native image and audio output, as well as native tool use.
This new generation has also pushed the frontiers of another capability, called Thinking. A couple weeks ago, we released a new model, Gemini 2.5, a thinking model that can reason through its thoughts before responding. It's our most intelligent AI model ever. And it is the best model in the world, according to the ChatbotArena leader board. [Applause]. It's state of the art across a range of benchmarks requiring advanced reasoning. That included the highest score,
ever, on Humanity's Last Exam, one of the hardest industry benchmarks that's designed to capture the human frontier of knowledge and reasoning. Those are a lot of impressive words, but let me show you what it can do. Take a look at this Rubik's cube, coded by developer Matt Berman. You might think of it as a toy, but it's actually a really complex reasoning challenge. Adjustable dimensions, scrambling the squares, keyboard controls. Gemini 2.5 Pro can simulate it
all. This is a significant leap and shows the ability to produce robust, interactive code. That's a fun one. Let's look at one other example. With a series of prompts, developer John Martin used 2.5 Pro to create a series of physics simulations like the earth's magnetic
field and general relativity. You can see how the model turns really complex concepts into stunning and interactive visuals. These are just a few brief examples, but we're excited about the possibilities, and we can't wait to see what you'll build with it. Gemini 2.5 Pro is available now for everyone in AI Studio,
Vertex AI and in the Gemini app. [Applause]. I'm also excited to announce Gemini 2.5 Flash, our low latency and most cost efficient model with thinking built in. [Applause]. With 2.5 Flash, you can control how much the model reasons and balance performance with your budget. 2.5 Flash is coming soon in AI Studio, Vertex AI and in the Gemini app.
We'll be sharing more details on the model and its performance soon. I'm pretty excited by it and can't wait for you to see it for yourselves. Our goal is to always bring our latest AI advances in to the fourth layer of our stack, products and platforms. Today all 15 of our half a billion user products,
including seven with 2 billion users, are powered by our Gemini models. AI deployed at the scale requires world class inference, which enterprises can benefit to build your own AI powered applications. Gemini is also helping us create net new products and experiences. NotebookLM is one example, used
by 100,000 businesses. It uses long context, multimodality, and our latest thinking models to show information in powerful ways. Gemini is not our only industry leading model. Veo 2 is the leading video generation model. Major film studios,
entertainment companies, as well as the top advertising agencies in the world, are using it to bring their stories to life. Getting advances into the hands of both consumers and enterprises is something we're really focused on. This is why we are able to innovate at the cutting edge, and push the boundaries of what's possible, for us and for you. The result, better, faster and more innovation for everyone. It's exciting to see how that's
helping companies of all sizes do more with AI, and translate those benefits to customers. I'm delighted to introduce Chris Kempczinski, CEO of McDonald's, to tell you more. But first, thank you for having me, and enjoy your week together in Las Vegas! Over to you, Chris! [Applause]. >> CHRIS KEMPCZINSKI: McDonald's is undergoing a once in a generation transformation. We have about 65 million people that come to our restaurants every single day, and it's how do we make their experience even better? Google is a big part of that, particularly as more and more of these customer interactions are happening in a digital world. That's why we're transforming our restaurant experience with the help of Google Cloud.
Behind the counter, our restaurant teams' jobs are becoming increasingly complex. With Edge Computing from Google Distributed Cloud, capabilities will readily improve stability, security and performance in our restaurants, all while giving us the space and power to test several new concepts we weren't able to do previously. For example, shift leaders will be able to leverage an AI powered assistant to help spot issues in the restaurant quickly. Our restaurant managers will be able to receive alerts on their devices based on real time data, say, from their freezer or fryers, along with guidance for predictive maintenance.
And with Gemini on Vertex AI, we can centralize all this information from restaurants in real time, making it easier for the right people to get answers with a simple question or prompt, improving the work environment in our restaurants across the globe for our more than 2 million team members. That's the magic. That's the power of AI and what Google Cloud has brought to McDonald's. [Applause]. >> THOMAS KURIAN: Thank you, Chris. McDonald's is a great example of a company integrating AI into the very core of its operations.
Customers around the world are choosing to work with Google for three important reasons: first, Google Cloud offers an AI optimized platform with leading price, performance and precision and quality. And new today, everything you need to build and manage multi agent systems. Our AI platform offers advanced infrastructure and databases, world class research, leading models, and grounding for model responses with Google quality search. Vertex AI, a robust developer platform, including the broadest range of enterprise ready tools with which you can build AI Agents and enable a multi agent ecosystem.
And the most comprehensive portfolio of purpose built Agents. Second, Google Cloud offers an open multi cloud platform that allows you to adopt AI Agents while connecting them with your existing IT landscape, including your databases, your document stores, enterprise applications and interoperating with models and agents from other providers. You get value faster from AI investments. And third, Google Cloud offers an enterprise ready AI platform, built for interoperability. It enables you to adopt AI deeply while addressing the evolving concerns around sovereignty, security, privacy and regulatory requirements. You can adopt AI while we protect your data and your intellectual property and enable you to maintain compliance. Powering this offering is our
advanced infrastructure core layer. To share the latest, please join me in welcoming Amin Vahdat. [Applause]. >> AMIN VAHDAT:
Thank you, Thomas. Demand for AI compute for training and inference is growing at an unprecedented rate. For over eight years, it has increased by over ten times, year over year, a factor of 100,000,000, just in eight years! We're continuing to offer leading power efficiency, performance and networking for training and inference workloads, starting with hardware. Today, we introduced our seventh generation of TPUs, Ironwood. Ironwood TPUs are our largest and most powerful TPU pods to date, more than a 10x improvement from our most recent high performance TPU, with over 9,000 chips per pod, to meet the exponentially growing demands of the most demanding thinking models like Gemini 2.5.
This delivers a staggering 42.5 exaflops of compute per pod. To give you a sense of the scale, the world's number one super computer supports 1.7 exaflops. Ironwood pods offer more than 24 times that compute power.
Ironwood TPUs are truly built for the next generation of AI workloads. TPUs are an incredible achievement, but they are just one piece of our overall infrastructure. After all, a chip is only as powerful as the system that surrounds it.
Our AI hypercomputer is a supercomputing system designed to simplify AI deployment, improve performance and optimize costs. It supports the best hardware platforms and brings together a single unified software stack and consumption model that enables you to use the hardware that best meets your needs and easily transition from one hardware system to the next. Vital as we innovate so rapidly. We have enhanced our GPU portfolio with the availability of A4X and A4 VMs, powered by NVIDIA's GB200 and B200 Blackwell GPUs. We were the first cloud provider to offer both options.
We're also pleased that Google Cloud will be among the first to offer NVIDIA's next generation Vera Rubin GPUs, which offer up to 15 exaflops of FP4 inference performance per rack. We also introduced Cluster Director, which enables you to deploy and manage a large number of accelerators as a single unit of compute to improve performance, efficiency and resilience. Storage is also vital to reduce bottlenecks for training and inference. We are introducing new storage innovations.
Hyperdisk Exapools offer the highest aggregate performance and capacity, per AI cluster, of any hyperscaler. Anywhere cache keeps data close to your accelerators, with up to 70% improvement in storage latency to reduce training time. And Rapid Storage, our first zonal object storage solution, offers five times lower latency for random reads and writes compared to the fastest comparable Cloud alternative. Software is how we orchestrate and simplify access to this powerful hardware, and today we're introducing three enhancements for AI inference. First, we're introducing new inference capabilities in Google Kubernetes Engine, including gen AI aware scaling and load balancing features, which help reduce serving costs by up to 30%, tail latency by up to 60%, and increase throughput by up to 40%.Second, we're announcing that Pathways, Google's own
distributed ML runtime powering Gemini, is now available for the first time for Cloud customers. Developed by Google DeepMind, Pathways enables state of the art, multi host inferencing, for dynamic scaling with high performance at optimal cost. Now you can scale out model serving to hundreds of accelerators for the best combination of batch efficiency and low latency. Third, we're bringing vLLM to TPUs.
This allows customers who have optimized PyTorch with vLLM for GPUs to easily and cost efficiently run their workloads on TPUs. All of these AI hypercomputer hardware and software enhancements together enable us to deliver more intelligence, or useful AI output, at a consistently low price. This is one reason why Gemini 2.0 Flash, powered by AI hypercomputer, achieves 24 times higher intelligence per dollar, compared to GPT 40, and 5 times higher than DeepSeek R1. We're truly seeing tremendous
momentum across our AI infrastructure portfolio, tripling the number of TPU and GPU hours consumed by our Cloud customers over the past year . And we're seeing tremendous customer momentum with AI Unicorns like Anthropic, Anyscale, Arize and Contextual AI. And enterprises. I do deployed ML models for factory workers. Schrödinger uses cloud GPUs for
advanced drug discovery. TSMC protects its data for mission critical workloads. And Airbus deployed an AI platform to advance aircraft performance, safety and reliability. Beyond optimizing training and inference in the Cloud, we know that many AI workloads need to be run on premises. As you heard from Chris Kempczinski at McDonald's, Google Distributed Cloud brings our hardware and software to your environment so you can bring AI capabilities closer to where data is generated for low latency and highly sensitive data in particular.
Today, we are announcing that Gemini can run on Google Distributed Cloud locally in air gapped environments, as well as connected environments. [Applause]. This all comes with the support for NVIDIA's Confidential Computing and Blackwell systems, DGX B200 and HGX B200 platform with Dell as a key partner. This complements our Google Distributed Cloud air gapped product, which is now authorized for U.S. Government Secret and Top Secret missions, and on which Gemini is available, providing the highest levels of security and compliance. NVIDIA is an important partner for Google and our customers. Let's hear directly from CEO Jensen Huang. >> JENSEN HUANG: Building
advanced AI infrastructure is deep computer science. No company is better at every single layer of computing than Google and Google Cloud. Between NVIDIA and Google Cloud, this super partnership includes capabilities that covers literally every single layer and every single aspect of computing. Every industry, every company, every country wants to get their hands on AI. However, everything has to be fundamentally confidential and secure, and so we're announcing something utterly gigantic today. Google Distributed Cloud with Gemini and NVIDIA are going to bring state of the art AI to the world's regulated industries and countries. Now, if you can't come to the
Cloud, Google Cloud will bring AI to you. [Cheers and Applause]. >> THOMAS KURIAN: Thank you, Jensen. We really value our deep engineering relationship with NVIDIA. Building on the groundbreaking research of Google DeepMind, we're delivering rapid innovation across many AI models, starting with Gemini, our most capable family of AI models. In the last year alone, we released Gemini, our first native multimodal model.
We delivered the first 2 million token context window. We built the live API for live, bidirectional voice and video interactions, led in price performance with our Flash models, and we recently launched Gemini 2.5 Pro, which is state of the art on a wide range of bench marks, and I'm pleased to say, number 1 on ChatbotArena. [Applause]. Gemini is providing best in class AI for many companies around the world, including our close partners Box and Palo Alto networks, who are using Gemini 2.5 to deliver new applications. It's also integrated across our own products, including Google Workspace. Where Gemini powers features in Gmail,
Docs, Drive and Meet, and it's now included in all subscriptions. Gemini in Workspace is helping customers like EV manufacturer Rivian, Freshfields to enhance legal work and expedite tasks like document drafting and research, and companies of Schwarz Group, Europe's largest retailer. Today, I'm pleased to announce three new innovations with Gemini in Workspace: Help Me analyze in Google Sheets which guides you through your data to complete expert level analysis. Audio overviews in Google Docs, where you can interact with Docs in an entirely new way by creating high quality audio versions of your content. And Google Workspace Flows to help you automate the time consuming, repetitive tasks, and make decisions with more context. Let's see how Workspace is helping businesses around the world. >>
We are scattered all over the globe. It's just bonkers to try to get everyone to collaborate. >> Keeping up with my E mail has always been really, really difficult. >> Is there anything more intimidating than a blank white page? >> But where do you begin? >> It's more important than ever that you have tools to work more quickly and efficiently together. >> Using Help Me Write, we're
able to save 35 hours a month for product descriptions for our Web site. >> We are leveraging Meet to take notes, summarize them and generate action items after the meeting so that we can be really present and focus on the content. >> When you're developing a deck or a pitch for a client and you're coming up with ideas and you need to visually manifest those ideas, that used to take days. And now with the right prompts,
we can do that in hours. >> We have to worry less about security because we know Google has our back. >> The security played a big part of the decision to move from the legacy on premise tool sets to Google Workspace. >> The impact of Gemini for
Workspace in our business has been really transformative. [Applause]. >> THOMAS KURIAN: Beyond Gemini, over the last year we've made huge improvements to Imagen 3, our highest quality text to image model, which generates images with better detail, richer lighting and fewer distracting artifacts than previous models. Imagen delivers accurate prompt adherence, bringing your creative vision to life with incredible precision. We also introduced Chirp 3 to help you create custom voices with just ten seconds of input, and to weave AI powered narration into your existing recordings. Today, we're making Lyria available in Google Cloud to transform text prompts into 30 second music clips, and we're the first hyperscaler to offer this capability. Let's hear a clip.
From Lyria. ¶ [Applause]. Veo 2 is our industry leading video generation model which generates many minutes of 4K video, watermarked with SynthID to ensure they can be identified as AI generated.
It gives creators unprecedented creative control with new editing tools, including camera pre sets to direct shot composition and camera angles without complex prompting. First and last shot control, to define the beginning and end of a video sequence with Veo seamlessly bridging the gap. And dynamic inpainting and outpainting for video editing and scaling. With Gemini, Imagen, Chirp, Lyria and Veo, Google is the only company that offers generative media models across all modalities, and all are available to you today on Vertex AI. [Cheers and Applause]. We've seen great examples from
our customers. Kraft Heinz is speeding up campaign creation. Agoda creates unique visuals of travel destinations. Bending Spoons makes 60 million photos every day more fun. And L'Oreal Group generates diverse cinematic shots. Using our models.
Please join me now in welcoming Nenshad Bardoliwalla for a demo of our models in action using Vertex AI. [Applause]. >> NENSHAD BARDOLIWALLA: Hello, everyone! How many of you have heard about our Cloud Next concert already? That's not as many as I was hoping to hear about.
And I think that's because we've been missing a teaser video. Now, let me tell you, it wasn't easy to pick the artist this year, because it turns out that even though he looks very demure and very mindful, Thomas Kurian is a massive Chapel Roan fan boy! Yes, I have seen the video that he sent Chapel directly of him going, H O T T O G O Thomas wants you in the show! [Laughter]. Yeah! But we waited for weeks to get a response from Chapel, and then she broke Thomas' heart with just three words: Good luck, babe.
So, we're going to use Vertex video to pump Thomas up and create a teaser video that's going to get you as amped up as I am. Now, we've already created our final video. I'm just going to show you how we got there. Given where we are, we're going to use Las Vegas skyline as a perfect backdrop for what we're going to do with Vertex AI Media Studio.
So let's go ahead and we're going to start by bringing in the Las Vegas skyline image. Really high quality, beautiful image. We're going to generate video, but here's the new hotness. Check it out!
Camera presets built right into Veo, panning left, panning right, time lapse, tracking shots, and even drone shots. So let's go ahead and submit a drone shot of the city skyline. There we go. We'll go and submit this. Now, normally this would take a few seconds. I ran this earlier today so it's cached so it's going to be a little quicker than normal.
All right. Let's look at video number 1! Absolutely spectacular! We have the ability to see the Fountains [Applause]. the Eiffel Tower. Now, let's go ahead and take a look at video number 2.
A different angle that Veo creates for us. Again, stunning imagery. You can see the clouds in the background, and look at the cars driving up and down Las Vegas Boulevard! Absolutely incredible! Now, one video is not going to do it for the concert promo we want to do so I want to show you some of the other videos that I created. I have one here of the stage being set up. All through the power of Veo. I have one of the band.
I even have one of the audience actually clapping for what they're about to see. This will be a good reminder for all of you. [Laughter]. Now, something very interesting happened. It turns out that Veo can do something that my 12 year old can do, and that is be an expert in photo bombing! It turns out that this great video we just saw has a crew member, and we love our crew members. However, in this case, I'd like to feature the guitar because the guitar is the most important part of the band! So let's go ahead and use Veo's new inpainting capability. And I'm sorry, sir, I apologize.
I know you're very good at your job, but I am going to have to remove you from this image. We will send flowers to you and your family, though, sir. Let's use the new inpainting capability, wait a couple of seconds, and let's see what we see.
Now, if this does what I think it does, it should preserve every single aspect of what we saw before, just without our stagehand. Look at that! [Applause]. Okay. So we got some video clips. Now we need some music! Let's try the first clip I created with Lyria and see how we like it. ¶ you know, that's not quite my tempo. I need music that's going to make all of you
feel like I'm never going to give you up. I'm never going to let you down. I'm never going to run around and dessert you. So, let's try clip number 2 and see how that works. ¶
All right. We have the recipe. I like that tune better. We've got the videos, we've got the music. Let's pull it all together and see what it looks like! Here we go! Play it, Sam! ¶ [Applause].
What do you think? Absolutely amazing! We've seen the amazing capabilities of Veo, the ability to create incredible shots with very little prompting, the ability to have editing capabilities that are easy to use and the cinematic quality. We're going to see you tomorrow night when Thomas does a stage dive into the mosh pit at Allegiant Stadium for the Killers! [Cheers and Applause]. Welcome back to the stage, my friend and spiritual advisor, Thomas Kurian! [Applause]. >> THOMAS KURIAN: Thank you,
Nenshad! [Laughter]. I am also very excited for the concert tomorrow night. [Laughter]. We are also bringing AI models to the physical world. Our partners like Samsung are using Gemini models for their exciting new AI companion robot, Ballie. And Google DeepMind recently
introduced two new AI models for a new generation of helpful robots. Now let's talk about Vertex AI, our comprehensive AI platform to help you discover enterprise ready foundation models, to customize, evaluate and deploy applications built with the best foundation models, and to build and manage AI agents at scale. Let's hear how Intuit is making tax prep even easier with Document AI, which is part of Vertex AI. >> Last year Intuit Turbotax processed 44 million returns and $107 billion in refunds with the help of AI. Yet some customers with complex 1099 forms faced hours of manual data entry. This year, Intuit unlocked
higher quality data comprehension and auto fill with Google Cloud Document AI. This done for you experience simplifies tax filing for millions, freeing up time for living life. Intuit built a new way to make taxes easier. [Applause]. >> THOMAS KURIAN: Tens of thousands of enterprises are building with Vertex AI and Gemini.
Nokia built a coding tool to speed up app development. Wayfair updates product attributes 5 times faster. AES, an energy company, reduces audit costs by 99%, and audit time from 14 days to just one hour. Commerzbank is creating AI assisted summaries of investment advisory calls. Seattle Children's Hospital makes thousands of pages of clinical guidelines instantly searchable by their pediatricians.
United Wholesale Mortgage is transforming the mortgage experience, more than doubling the underwriter productivity. And Honeywell has incorporated Gemini into their product development. >> Honeywell and Google Cloud designed a new way to manage product lifecycles. It will revolutionize how they handle millions of products. Built with Vertex AI, BigQuery and Gemini, this agentic framework accelerates spec and model creation, connects with their global install based uncover performance and proven insights and extends life cycles by reengineering products.
Estimated to help their engineers deliver results up to 70% faster. With AI Agents, Honeywell is introducing a new way to optimize millions of products. [Applause]. >> THOMAS KURIAN: In just the last year, we've seen over 40 times growth in Gemini use and Vertex AI, now with billions of API calls each month.
Vertex AI gives you easy access to over 200 curated foundation models through our Model Garden. We offer all of Google's models, Gemini, Veo, Imagen, and our latest research models, plus curated popular third party models and open source models, all now on Vertex AI. New Vertex dashboards help you monitor usage, throughput, latency, and to troubleshoot errors.
New tuning methods help you optimize the model's performance for your applications. We are excited to announce the general availability of Meta's Llama 4 on Vertex AI. And last week, we announced that Ai2's full portfolio of open models are also accessible on the Vertex AI Model Garden. With Vertex AI, you can be sure that your model has access to the right information at the right time.
You can connect to any data source or any vector database on any cloud. And announcing today, you can build agents directly on our existing NetApp storage without requiring any data duplication. [Applause]. You can connect to a broad range of applications including Oracle, SAP, ServiceNow, and Workday.
And for model factuality, we offer the most comprehensive approach to grounding on the market today, combining grounding with Google Search, grounding with your own enterprise data, Google Maps, and third party sources. Let's hear more from Deutsch Bank CEO Christian Sewing. >> CHRISTIAN SEWING: For over 150 years, our clients have looked to Deutsch Bank to support their lasting success and financial security, and they need us now more than ever as advisor and risk manager in a world marked by uncertainty and shifting geopolitics. Technology plays a key role. Our partnership with Google
Cloud enables us to take advantage of the latest tools. DB Lumina is our AI powered research agent. Built on Gemini and Vertex AI. It maintains data privacy and improves our productivity while operating in one of the most regulated industries, where trust is built in years and lost in seconds. >> A tool like DB Lumina allows
us to be ahead of our competitors and provide faster, more accurate analysis of data. Recently, there was a big report in the markets which was 400 pages. We put it into DB Lumina and gave it some prompts and within seconds it gave us a three page summary.
We were able to give that to traders and our clients to help them process that information. >> CHRISTIAN SEWING: Through our partnership with Google Cloud, we have seen a real breakthrough and this is just the beginning. We see a future where generative AI is integrated into basically every process we run, making our employees' life easier while meeting the changing expectations of our clients. [Applause]. >> THOMAS KURIAN:
Thanks so much, Christian. We're thrilled to see how quickly you at Deutsch Bank have moved AI from pilot to production. Now let's talk about agents. Agents are intelligent systems that show reasoning, planning, memory and the ability to use tools. They are able to think multiple steps ahead, use tools including working with software and systems, to get something done on your behalf, and under your supervision. Agents work alongside employees to drive efficiencies, help with decision making, and increase motivation. A great example of a company working with Google Cloud to develop agents is Salesforce.
Let's hear from CEO Marc Benioff. >> Salesforce and Google, two of the world's most innovative companies. We've been on an incredible journey together, and today, well, that partnership has never been stronger. Right now, we are really at the
start of the biggest shift any of us have ever seen in our careers. I'll tell you, that's why we are so excited about Agentforce, and our expanded partnership now with Google. I just love Gemini. I use it every single day. Whether it's Gemini inside
Agentforce, whether it's all the integrations between Google and Salesforce, together we're leading the digital labor revolution. That's the future that's going to drive massive gains in human augmentation and productivity, efficiency, the fundamental KPIs of our business and ultimately our customer success, and we're looking forward to doing even more between Salesforce and Google. [Applause]. >> THOMAS KURIAN: Thank you very much, Marc. We're excited to build together and continue on this journey with you.
Now, with Google Cloud, starting today, you can build and manage multi agent systems with Vertex AI and our new Agent Development Kit. You can scale the adoption of agents across your enterprise with our newly released Google Agentspace, and you can accelerate deployment with packaged AI agents that are ready for use today. Following the introduction of Vertex AI Agent Builder last year, we are now announcing today, a new Agent Development Kit. It is a new open source framework that simplifies the process of building sophisticated multi agent systems.
Now you can build sophisticated Gemini powered agents, help them use tools and do complex multi step tasks, including reasoning. Or thinking. You can also discover other agents, learn their skills, and enable agents to work together while maintaining precise control. Agent Development Kit supports the Model Context Protocol, which provides a unified way for AI models to access and interact with various data sources and tools, rather than requiring custom integrations for each and every one. We are also introducing a new Agent2Agent Protocol that allows agents to communicate with each other, regardless of the underlying model and framework. They were developed with. This protocol is supported by many leading partners who share our vision to allow agents to work across the milt agent ecosystem, and with agents built on other agent frameworks, including Lang Graph and Crew AI. Today we're putting AI agents in the hands of every employee, with Google Agentspace. Employees using Google Agent space can
now find and synthesize information from within their organization, converse with AI agents, and have these agents take action on their behalf for their enterprise applications. Google Agent space combines Google quality enterprise search, conversational AI or chat, and Gemini and third party agents. It also includes a broad set of tools, including purpose built connectors to search and transact with documents and databases, as well as SaaS applications, with advanced security and compliance to protect your data and your intellectual property. Let's take a look at Agentspace in action. Please welcome Gabe Weiss. >> GABE WEISS: Thanks, Thomas! For the next few minutes, I'm going to be a relationship manager at a bank. Starting with a quick tour, this is my home page, authenticated and personalized just for me. The Agent Gallery lets me see my
company's approved selection of purpose built agents, including ones powered by third party models like Llama and Claude. You see, we've got some Google made agents. We have agents that my bank has made available to me, either ones we've created or ones built by partners.
And then the best part, my own personal agents, which I can build directly inside Agentspace with this button over here, or even easier just from having a little conversation. Let's see how easy it is to create an agent to automate a daily task. Now it's critical for me to stay on top with what I do for my clients. so I start every morning with a portfolio analysis.
This only uses information that I have permission to access. It knows which clients are mine and summarizes top points from my data sources like OneDrive, Salesforce, and Dun & Bradstreet. If I have questions, I have a direct link to my sources here, and if I need even more control, I can refine that list of sources. But Agentspace doesn't just summarize information. It's interpreting my question and surfacing what matters most.
For example, in this chart, I can see Agentspace has flagged that Acme General Contracting might have some cash flow problems in the future. Already it's given me a massive report which is going to save me a ton of manual research, and I can go ahead and read this later. But for now, let's set up an agent so I can keep an eye on Acme. Agentspace automatically generates an agent plan based on our conversation so far. And this is good, but I think I want more.
I'm going to have it generate an audio summary and send it to my inbox so I can listen to it on my morning commute. And just like that, I have built my own custom agent to use whenever I want, without writing a single line of code. Agentspace has identified a cash flow problem with Acme General Contracting. I need to dig into that. Maybe this is a problem with construction in general and not specific to Acme. Agentspace has already identified that possibility as a suggested follow up.
So now, let's go ahead and deep dive into general contracting industry trends. This activates Google's enterprise deep research agent, which starts by telling me what it plans to research and in what order. At this point I could edit this plan if I wanted to, but it looks pretty good, so I'll start the research. Now, I do want to call out, we've cached the plan and the results here.
Normally this would take a little bit longer. This agent is pulling in real time information from Google Search to build its report but even cooler it's also searching my internal enterprise data and adjusting this plan in real time, adding additional questions based on what it's going to find along the way. And again, an incredibly insightful analysis, including some source links, but thankfully here at the bottom, it also is going to give me a great, succinct executive summary. Let's take a quick look at this. Yep. I can see Acme General Contracting is likely
being affected by rising material costs, supply chain disruptions and regulatory complexities that pose significant hurdles. That's really great! I mean, okay, maybe not for Acme, but the analysis is really great. I don't want Acme to be surprised by this at all so I'm going to have our bank's cash flow agent do some forecasting across the next three quarters for me. This agent uses Google's new Time Series Forecasting Model, which is specifically trained for scenarios just like this. And again, I'm going to get a super clear, very clear summary with at the bottom some great recommended steps for Acme, and I need them to see it right away so I can ask Agentspace, draft me an E mail to Acme general contracting CEO requesting a meeting for next week. And just like that,
I've got the draft ready to go. And, even better, I can send it off directly from within Agentspace, so I don't even have to switch to outlook or Gmail. I'm all set and Agentspace has saved my session so I can prep for that meeting right where I left off whenever I'm ready. Let's go ahead and recap. While I don't actually work for the bank, the value that Agentspace adds is very real. It's so easy to interact with all of your enterprise data and tools in one place and build and use agents directly from your conversational workflow. Powered by Gemini 2.5 and Google Search technology, Agentspace is the only hyperscaler platform on the market that can connect to third party data and tools and offers interoperability with third party agents and models.
For companies with strict regulatory needs, like a bank, Agentspace provides stringent access controls at the employee level and can operate within the your own VPC, ensuring that your data stays yours while meeting all of your requirements. Agentspace is a game changer, and we can't wait to see how you put it to work. Thanks! Back to you, Thomas! [Applause]. >> THOMAS KURIAN: Thank you very much, Gabe!
Today, we are excited to announce that Agentspace is integrated with your Chrome browser to allow users to search and access your enterprise data directly from the search box in Chrome. Employees can use Agentspace to access Google built expert AI agents, including NotebookLM, an AI powered note taking and research agent that allows users to upload up to 50 documents with 25 million words and then query them using AI, effectively turning notes and sources into a virtual research assistant. You can also use our idea generation agent, which accelerates innovation, brainstorming and problem solving. It uses a tournament style framework to rank ideas based on employee defined criteria, refine them, and generate new ones. And enterprise deep research agent, which Gabe just showed you, researches complex topics on your behalf and provides you with findings in a comprehensive, easy to read report. Customers and partners around the world are already using Agentspace. KPMG is building Google AI into
their newly formed KPMG Law Firm and implementing Agentspace to enhance their own workplace operations. Cohesity is integrating with Agentspace to provide employees with greater data discovery for better decision making while also increasing security. And threat protection. Gordon Food Services is simplifying insight discovery and recommending next steps. Rubrik is leveraging agents to develop deeper customer insights and prepare for impactful sales interactions. And Agentspace will provide Wells Fargo bank the unique opportunity to modernize and simplify banking.
Now we are going to dive into five categories of agents where we are already seeing tremendous business impact. Please welcome Lisa O'Malley. [Applause]. >> LISA O'MALLEY: Thanks, Thomas. Let's start with customer agents. They can synthesize and reason across all types of multimodal information, including text, audio, images and video.
Communicate and engage natural ly with human like speech and dialogue. Connect across enterprise applications and take actions on behalf of the user. And be used in the contact center and on the web, on devices, in stores, in cars, and more.
Customer agents built with Vertex AI Search are helping customers to quickly find answers and the right products using both text and images in search queries. Let's hear from Reddit Chief Product Officer, Pali Bhat. >> PALI BHAT: At Reddit our mission is to empower communities and make their knowledge accessible to all. We've been working on this mission for nearly 20 years, which, in turn, has made Reddit one of the internet's largest sources of authentic conversations.
With that vast amount of conversations and perspectives, we wanted to build a unique search product that's powered with AI but still grounded in all of the real conversations and perspectives that are available on Reddit. This is why we introduced Reddit Answers, a new AI powered way to get information, recommendations and discussions on virtually any topic. It provides powerful AI that's grounded in Redditors' existing posts and conversations so it shows you more of what real humans think versus creating unverifiable perspectives on its own. Reddit Answers is different from any other generative AI product on the market. It leverages Vertex AI Search to make finding the answers and perspectives people seek faster and more relevant. We've seen awesome results so far because the users who have been able to access this product and test it out really love the experience.
This gets them to the heart of the conversations that they are looking for right away. That's the magic of Reddit Answers. It combines AI with the power of Reddit. >> LISA O'MALLEY: Thank you, Pali. [Applause]. We have also introduced Vertex AI Search for Healthcare and Retail, making it super easy for doctors, nurses and providers to rapidly search and analyze patient data, including x rays, scans, images and medical histories. Retailers can add product
discovery to their Web sites, powered by Google Search. This helps them deliver hyper relevant results and personalized recommendations for each customer, boosting conversion rates and maximizing revenue per shopper. We are seeing huge momentum for Vertex AI Search with billions of daily queries executed by our customers. For example, Lowe's is revolutionizing product discovery with Vertex AI Search to generate dynamic product recommendations and address ' complex queries. Globo created a recommendations experience inside its streaming platform that more than doubled click through play. And let's hear how Mercado Libre is transforming how customers discover products that they love. >> Mercado Libre, Latin
America's eCommerce leader, has deployed a Vertex AI search across 150 million items in three parallel countries. This multimodal search technology understands deep meaning across text and images, not just keywords. It is helping their 100 million customers find the products they love faster. Already delivering millions of dollars in incremental revenue, Mercado Libre is delivering a new way to shop. [Applause]. >> LISA O'MALLEY: Google Cloud's own purpose built Customer Engagement Suite is transforming customer service.
Grounded in your company's data, it provides out of the box functionality to build agents and works across web, mobile, call center, in store and with third party telephony and CRM systems. These unique capabilities have led to rapid growth, with an increase in conversational AI agent usage. DBS, a leading Asian financial services group, is reducing customer call handling times by 20%. Loveholidays saved 20% of their customer service cost per year. And our very own YouTube achieved a 75% reduction in calls abandoned while waiting to speak to a representative. Now, let's hear how Verizon is improving
their customer experience using AI agents. >> Verizon is transforming how they serve over 115 million connections with Google Cloud's Customer Engagement Suite. Their personal research assistant uses AI to provide 28,000 care representatives with instant, personalized information about a customer's unique needs, leading to faster and more satisfying resolutions for even the most complex inquiries.
With Customer Engagement Suite, Verizon is elevating its service experience, reducing wait times, and delivering exceptional support at massive scale. Verizon developed a new way to personalize customer service. [Applause]. >> LISA O'MALLEY: The business impact Verizon has experienced is nothing short of extraordinary. Today we're announcing our next generation of Customer Engagement Suite, which will include human like voices, comprehension, and the ability to understand emotions so agents can adapt better during the conversation. Streaming video support, so virtual agents can respond can interpret and respond to what they see in real time through customer devices. AI assistance to build custom agents in a no code interface, and the ability to use a variety of tools through API calls to interact and perform specific tasks for your application, like look up products, add to cart or check out. And integration with data sources, CRM systems,
and popular business messaging platforms. Now, let's see a demo of all this cool stuff in action. Welcome my teammate, Patrick Marlow to the stage. >> PATRICK MARLOW: Thanks, Lisa. Hey, everyone. I'm Patrick Marlow, a Product Manager here at Google Cloud, and I'm stoked to be here today showcasing our next generation Customer Engagement Suite in action. To be honest, I'm even more
excited to spruce up our keynote stage. I was thinking some greenery and flowers might be nice. You know, I've already made a couple of trips to the hardware store this morning, and I still forgot to pick up potting soil. Classic. So let's see how a next gen agent can hopefully help me get this last order correct.
We're going to start a brand new voice interaction with our agent here. >> AGENT: Hi there. Welcome to Simple Home and Garden. Is this Patrick? >> PATRICK MARLOW: Hey, yeah, this is Patrick. Good morning. How are you? Good morning to you, too, Patrick. Doing great. Thanks for asking. It looks like you've
been shopping with us recently. I see you have a few things in your cart already. What can I help you with today? >> PATRICK MARLOW: Yeah, so I just bought some plants and I'm trying to pick out potting soil and fertilizer and I'm not sure if I'm getting the right stuff. Could you help me with that? >> AGENT: I can definitely help with that, Patrick. To give you the best recommendations, could you show me the plants you bought? I can take a look and help you choose the best soil and fertilizer. I'm sending a video request to your browser now. Great!
I can see you. Just hold the plants up to the camera so I can help identify them. >> PATRICK MARLOW: Yeah, so I have about 40 of these in various different colors. >> AGENT: Those are beautiful petunias. They're a great choice for vibrant color. Let me take a closer look at the items in your cart. Okay.
I see you have standard potting soil and general purpose fertilizer. That could work. But petunias really thrive in a rich, well draining soil and a fertilizer specifically for flowering plants.
Okay. I found some recommendations for your petunias. >> PATRICK MARLOW: Yeah, I'm taking a look at these now. They look fantastic. Thanks. >> AGENT: Great! Would you like me to remove the standard potting soil and general purpose fertilizer and add the bloom booster potting mix and flower power to your cart? >> PATRICK MARLOW: Yeah, let's go ahead and do that. >> AGENT: Okay. I've updated your cart. Is there anything else
I can help you with today? >> PATRICK MARLOW: I mean, not unless you can come over and plant these for me. >> AGENT: Funny you should mention that. We actually do offer professional landscaping services. Would you be interested in a quote? >> PATRICK MARLOW: Yeah. I mean, how much would something like that cost? >> AGENT: For planting 40 petunias, I would suggest a four hour appointment, which would come out to $200. >> PATRICK MARLOW: Okay. That's interesting. I'm curious, do you also price match? >> AGENT: Yes, we do price match on occasion, but it's usually on a case by case basis. >> PATRICK MARLOW: Oh, in that case,
I most definitely saw an ad from one of your competitors, Greenscapes, where they were offering 50% off first time landscaping services for new customers. You should totally give me that 50% discount. >> AGENT: Apologies, but it looks like that amount is beyond what I'm authorized to approve. Let me check with my supervisor to see what we can do. Just a moment. >> PATRICK MARLOW: So I'm going to step away from being the human user in this conversation for a moment, and we're going to go behind the scenes over here to our call center.
And so what you're going to see here in our CRM Salesforce is that our virtual assistant, our virtual agent has escalated up to a human call center manager, also played by me, and what I can do here is guide the agent through this portion of the conversation without fully taking over the conversation. So you can see here, Patrick is requesting a price match for 50% off offer, and I'll say something like, you know, we can't we can't do that, but we can do 20 oh, say we can do 20% instead. And I'm going to send this message and then we're going to pick up the conversation on the other side. >> AGENT: Good news, Patrick! My supervisor has approved a discount. We can offer you 20% off our landscaping services. I've applied a 20% discount to the landscaping services. >> PATRICK MARLOW: Well, that's awesome.
I'll take it. >> AGENT: Fantastic! Let's get that scheduled. I can have someone out tomorrow from 8 a.m. to 12 p.m., or between 1 p.m. and 5 p.m. Which would you prefer? >> PATRICK MARLOW: Let's definitely do the afternoon time because I have a pickleball tournament in the morning. >> AGENT: Great!
You're all set. You'll receive a confirmation E mail in just a moment with all the details. Is there anything else I can assist you with today, Patrick? >> PATRICK MARLOW: Yeah, do I actually have to pay for the landscaping services right now? >> AGENT: No. We'll just add this to your account. And you can pay once the service is complete. Anything else? >> PATRICK MARLOW: No. I'm just going to finish checking out here. You've been really wonderful today.
Thanks for your help. >> AGENT: You're very welcome
2025-04-10 23:19
