Fortune partners with global people analytics firm Great Place to Work to analyze survey feedback representing employees from companies across the U.S.

GPTW Millennial Small & Medium 2023-Gem

Employees say:&nbsp;"I have had incredible managers who are very invested in my career goals and growth. When I share my goals, I feel heard and supported. My managers give me clear expectations and opportunities to grow my skills to up-level myself—and advocate for my promotion and compensation changes."

100 Best Small and Medium Workplaces for Millennials

Fortune partnered with Great Place to Work to analyze survey feedback representing nearly 80,000 U.S. employees.

GPTW Bay Area Small Medium 2024-Gem

Employees say:&nbsp;"There is an incredibly uplifting, kind, and supportive culture across the organization. Leadership prioritizes work-life balance by example, from direct managers to executive leadership."

75 Best Small and Medium Workplaces in the Bay Area

See which workplaces made this year's list.

GPTW 2023 Best Workplaces tech small:medium GPTW 2023 Best Workplaces tech small:mediumGem_5

Employees say: 
 
"Gem has an incredible culture where leaders live our values, and those behaviors spread through the company. It is a very welcoming place where everyone is encouraged to think and act like an owner."

Best Small and Medium Workplaces in Technology

Fortune partnered with Great Place to Work to analyze survey feedback representing nearly 100,000 U.S. employees.

GPTW Bay Area 2023-Gem

Employees say: 
 
"There's a genuine feeling of symmetry and courtesy that can be felt across the organization into leadership. This is the first time I've worked at a company that didn't have drama. People respect each other and try to celebrate each other through various channels of conversation, whether through Slack, live chat during Zoom meetings, etc. It generates a unique feeling of positivity that I have never felt in my career before. I feel lucky to have landed here.

Fortune partnered with Great Place to Work to analyze survey feedback representing nearly half a million U.S. employees.

Employees say: 
 
"There's a genuine feeling of symmetry and courtesy that can be felt across the organization into leadership. People respect each other and try to celebrate each other through various channels of conversation, whether through Slack, live chat during Zoom meetings, etc. It generates a unique feeling of positivity that I have never felt in my career before. I feel lucky to have landed here."

These are the best workplaces in the U.S. that employ between 100 and 999 people.

GPTW Medium 2023-Gem

"The people at Gem are wonderful. I have a diverse team of talented, thoughtful, supportive coworkers who make it so enjoyable to be part of Gem. I appreciate that our hiring process identifies candidates that are a good fit for our culture. Doing great work with a stellar team is very valuable to me as an employee."

100 Best Medium Workplaces

GPTW Information Technology Small Medium 2022-Gem

Employees say: 
 
“There's a genuine feeling of symmetry and courtesy that can be felt across the organization into leadership. This is the first time I've worked at a company that didn't have drama. People respect each other and try to celebrate each other through various channels of conversation, whether through slack, live chat during Zoom meetings, etc. It generates a unique feeling of positivity that I have never felt in my career before. I feel lucky to have landed here.”

Based on employee reports, Bay Area-headquartered companies excel at fostering wellness.

GPTW Bay Area Small Medium 2022-Gem

Employees say: 
 
“The company culture is built around diversity and inclusion. Employees are all made to feel at home here at Gem. Transparency is also one of Gem's core values, and the management does a great job of making sure everyone in the company is aware of the company-wide decisions, goals, and strategies.” 
 
Read the Great Place to Work review .

To determine millennials' workplace preferences, Fortune's research partner Great Place to Work analyzed feedback representing more than 6.1 million U.S. employees.

GPTW 2022 Millennials List Small Medium-Gem_7024105-2

Employees say: 
 
“Amazing level of transparency which enables an effective work environment and great cross-functional partnerships. Gem genuinely allows everyone to be their true selves. Most companies say so, but Gem actually does it."

The companies on this annual list make a conscious effort to be supportive, inclusive workplaces.

GPTW Women Small Medium 2022-Gem

Employees say: 
 
“The transparency at Gem is unlike any other company I have worked for. From board decks to decisions on my team, I feel like I am given visibility into why we do what we do, and what is going on at our company on every level. It makes me feel very connected to our company mission, and gives my work more meaning, because I know what we are driving towards.”

Best Small and Medium Workplaces for Women

These are the best workplaces in the U.S. that employ between 10 and 99 people.

GPTW Medium Companies 2022-Gem

Employees say: 
 
“Gem is the most value-driven company I've ever worked for. Our culture is incredibly inclusive &amp; people are very genuine. There are more cross-functional working relationships and friendships here than any other company I've worked at, and that has led to a truly outstanding product that customers LOVE.”

These companies empower employees to bring their whole selves to work, prioritize their well-being, and grow in their careers.

Best Workplaces Bay Area 2021-Gem

Employees say: 
 
“During this 'shelter in place' time during the 2020 pandemic, Gem has done so much to enable us to work from home well. The company shows it cares about our mental health by continuing 'off site' team building events.”

GPTW-MED-92-Gem

Employees say: 
 
"The company culture is built around diversity and inclusion. Employees are all made to feel at home here at Gem."

Best Workplaces Small Companies-2020-Gem

Employees say: 
 
"The company treats you like family and everyone is willing to go out of their way to ensure you succeed. We stand firmly behind our values of being customer-focused, being transparent, celebrating diversity, and moving with velocity."

100 Best Small Workplaces

Company Info

Founded

Industry Vertical

U.S. Employees

Sites

HQ Location

Web Address

Date info provided (companies may opt to keep some data confidential)

Best Medium Workplaces

Industry

Best Workplaces In The Bay Area

Best Workplaces For Millennials

Get insights into Gem including business profile, stock information, news, and rankings from Fortune.

Gem Company Profile, Stock Price, News, Rankings

legacy

Companies

Gem | Fortune

Gem Company Profile, Stock Price, News, Rankings | Fortune

A generation of Gen Z founders is redefining how intelligence shows up in the world. They’re advancing the science behind how machines perceive reality while reimagining how humans connect, create, and share meaning. By blending research, storytelling, and design, these emerging leaders are shaping how artificial intelligence moves through everyday life. As AI pierces the cultural zeitgeist, they share how they’re helping to shape the future of technology, creativity, and human connection.

The Future Makers

Conferences

Welcome to Fortune, Tiara. Thank you so much for having me. I am so excited for this conversation. When we were putting this agenda together, someone on the team suggested that we should have the Gen Z perspective, and I was like, yes, I want to be the one to moderate this conversation. So, um, I am a big believer in including youth voices in, um, shaping the future of AI, and you are right there making it happen. So I I'm guessing a lot of the audience are not familiar with who you are, so I would love to give you an opportunity to share a bit of your story. But you are so impressive, you actually, when I was prepping for this, you all, I was like, oh my God, I feel like a loser. You've been named to the Time's most influential teens, Forbes 30 under 30. You're a Peter Thiel fellow, a YC Combinator, uh YC alum, Y Combinator alum, and you're the founder, co-founder and CTO of Chima. So just give us a little bit about your story and how that all culminated into you starting Chima. So I did computer science at Stanford and before that I actually have always been very passionate about research and algorithmic development. So pre um college, I won the Google Grand Prize for seminal research in AI and applying that to climate change. You, you were like 16. I was 1617, OK. And I, I've always been really fascinated about the advancements that we can make with AI models, but also applying them to real world use cases, very much with what we're seeing right now with large language models and generative AI being applied to real world use cases. So I've always been really passionate about this application layer from sort of the model layer to real world use cases and Stanford, a lot of my time was spent at the human-centered AI labs where we actually saw this rich amount of data being applied in different use cases and that really was some of the founding stories of how we started and we operate as an applied AI lab and we are backed by YC, General Catalyst, incredible angel investors, and really Focus on what we call Sam Altman as an investor. So, so what we really do with an incredible number of supporters is, um, focus on how we take these incredible models from different, uh, sort of AI labs but really unlock it in application for different enterprises. Yeah, I would love for us to just spend a few moments, um, double clicking on that. What, who's your target customer and Bring this all to life. Give us an example of what you would do for one of your customers. And so it's really interesting because what we call our target customer can range from medium to a large scale enterprise because what we really see is that as AI models right now from let's say Opus 4.5 to a GPT 5.1 get really smarter, the amount of use cases they unlock exponential. increase, but even if the models stopped increasing in efficiency and improving in the benchmarks right now, the amount of economic unlock and efficiency unlock that they provide is still incredible. And so there's still so many ways that they can be applied that they haven't been applied that even if they infiltrate every part of the enterprise is really incredible in terms of what we really do. I really understand from a very clear use case how a certain function or operation occurs in the enterprise. Think about it as a full back office process and we take it all the way from actually automating that full back office process to end to end redoing it with AI from the ground up. Interesting. So you would come into a potential enterprise and you would like inspect. Workflow and then reimagine that with AI and and it really ranges from being able to apply not just basic large language models through coding agents to also computer use agents. So everything from how do you take the best and most capable computer use agents that can actually automate RPA typical work to actually being able to go fully multimodal. So if you think about this as a Full go to market function being able to understand from let's say the research perspective having deep research agents unlock and fulfill that full end to end function to actually going fully fully multimodal. So what does that mean for uh an enterprises' full campaign as they expand it and enter an entire new market? Yeah, very cool. So you are building in an incredibly fast moving AI landscape and you're doing it super early in your career, um. Without decades of legacy kind of patterns to unlearn, so what advantage does your youth, um, give you as a co-founder? Is it your mindset, your intuition, your hustle, something else? I love that question because I think specifically being part of Gen Z and the sort of younger generation being builders is what we're actually seeing is that most young people that are building these models have an inherent fluency with how they interact with them. We're not. The truth is the younger generation. Isn't adopting AI. We're growing up fluent in AI, so we aren't thinking about coding from scratch. We're thinking about coding with a coding agent side by side, and that fundamentally changes how you write, how you take tests, how you apply to jobs or different applications, because it's not from the ground up. It's actually being able to do that. different models or agents side by side and I think what that really means is that this broad level of use cases and applications we're seeing is really being pioneered by the younger generation. So we didn't have the mobile era as as being sort of builder first, but we do have the AI era which is going to be, if not bigger than than the previous one. I'm gonna go to questions next, um. But I want to talk about the future of work. So yesterday we had a really robust conversation around is AI killing entry level jobs, and it sparked a lot of debate, and I actually want to make it personal for a second. So I have two Gen Z kids. My daughter is 22 and she just graduated from college. He's a food anthropologist. It's a thing. And my son is almost 17. He's a junior in high school. He's my AI guy. He's AI forward and, but he's trying to figure out what to, you know, what to study. Um, so what advice would you have for both of them? Well, I think your daughter is incredible and probably it's not going to take her job because I don't know, but I also think that a lot of the conversation and the anxiety around AI is because often we haven't sort of embraced that it is coming, right? The actual sort of nature. of the future of their work is going to change and what that really means is things like customer support or typical back office jobs are going to change whether they change now or whether they change in a couple of months or years. What that means with the new types of jobs that they take on is something that I think The best positioned to think about very deeply now um but with that being said, I also think that them being able to build in the space is incredible, right? I think they're well positioned to not just join other companies but also contribute to the space because. Never before has there been so many unexplored ideas and use cases with AI that they obviously we've all heard about the one person billion dollar company, um, but what that really means is they can be that, right? And so being a builder or being somebody that's excited to contribute is incredibly exciting right now with AI. Yeah, amazing. Um, do we have any questions? I can't see. Is there a question here? Yes. Charlotte. Hello, my name is Charlotte. I am with Fortune. Um, Kara, I have a question for you that I want your perspective on. We hear a lot in the media world from a lot of CEOs. Like I feel like one of the main talking points is AI is not gonna take your job, but the person who knows how to use AI does. Um, my question for you from your POV is like, where, where does that statement sort of stop and end? Like what is the meaning from your perspective behind that statement and how do you see that like in the future? I love that question because I think. Coming from some of my earliest research has been in climate change, and I think a lot of people are very familiar with the term climate anxiety, which is this idea that there's this movement of climate change coming up and we don't really know what to do, but we know it's coming and nobody is moving as fast to solve the problem. There's a similar thing with AI which is AI anxiety, which is there's literally just In the past couple of weeks been two model releases that have engulfed the benchmarks in such an enormous way that pretty much everything you've ever used AI for can now just be 10x right before Gemini 3 we didn't know what was possible. Now there's a ton of other use cases. So there is a sense of AI anxiety because if you're not consistently ahead of it or on top of it, you're kind of left behind. And the truth is that If that is the case, how can we all stay informed or use these tools every single day so that we do feel as up to date as possible. Now what that means is I'm not saying everybody has to sort of keep up to date with every model release, etc. but that does mean sort of arming yourself with a couple of different tools, whether that be I I'm, I'm not gonna. Sort of name specific ones, but if it's sort of Chat GBT or Gemini, because ultimately there are sort of some main model players are coming out and being able to use them as co-pilots and tools in your everyday life, right, because when Gemini 4 comes out, it's still going to be sort of in the Google interface, right, when the next open AI model comes out, it's still going to be in chat GBT being able. To understand and use these different tools as they expand are really key to staying ahead, right, because being able to use them in different interfaces allows us to become really familiar with them, become really comfortable with them, but also use them as they continuously push different interfaces out, um, but also not having that sense of anxiety. About it because you're sort of comfortable with them. Kiara, um, one of my investment thesis is human centric AI, so AI that creates massive economic opportunity but also unlocks human potential. You did your masters at the human centered AI lab at Stanford. What are some of the principles or values that you, um, kind of explored at Stanford that you've brought with you to the company? I think what's really interesting about this concept of human-centered AI is that pretty much every Interaction with a model, whether that be an image model like nano banana or um a sort of large language model is really rooted in human centered AI. And so we see incredible benchmarks, we see um sort of the next model improving on the on the SWI benchmark and coding agents, but really if you take a step back, what makes specific models feel different than the other ones you see this. Very clearly in image models, right? People see and and work with nano banana and multi-turn edits, and it really is because they see something in the image quality that's very different to a seed dream, right? Um, it's really diff difficult to measure the quality differences between these models and, and why that's the case is that as models and use cases and efficiencies change, the the the key differentiator. is taste. Gary Tan from which is not captured in the benchmarks around accuracy, right, right, and we see this with design. If you've ever sort of coded with a coding agent, you know it loves a couple things. Sparkle emojis is one. It puts it all over front end UI. It loves sort of shades the end sort of components in a way that um you, you know something is vibe coded if you've ever sort of worked with a coding agent and the. The the key sort of differentiator is what makes it taste and feel different, and that is human centered, which is what do people actually want to see in an AI application or an AI use case. Yeah, I would love to see more benchmarks that kind of capture that aspect of these models like the design principles and kind of the the human aspects of it and I'm involved with this MIT group. It's called Advancing Human. With AI and they are actually trying to collaborate with some of the big kind of the open AIs of the world to build these benchmarks. So it'll be exciting to see, yeah, I, I think as models code more on the front end and, and less on the back end, that's, that's going to be so important. Do we have more questions in the audience? OK. What is, um, the biggest misconception about the Gen Z, um, usage of AI? Like how do you guys use AI? I think that or think about AI. I think that the biggest misconception is that young people are using AI to not think things through. I think that like as a shortcut, yeah, I, I think that really intelligent Gen Z, um, individuals are using it to think even deeper because you can run. A deep research report on let's say parallel AI that goes on for 34 hours and what are you doing when you actually have that report on your fingertips? What does that unlock for you if you're looking at the financial market, if you've done a deep dive on your favorite sort of article or topic, what that unlocks for you and I. I think people think that instead of you doing all that work, you've now sort of handed off the cognitive load to an AI model, but really what does that allow you to think differently about and think even deeper on and I think that's the biggest misconception that people have about young people and their use of AI. I, I know, so impressive, right? So cool. So cool, um. I have to ask you this. You're a Peter Thiel fellow. Did you consider dropping out of school to start your company? Please say no because my son may listen to this conversation and. And, and I do want him to, no judging, but yeah, but I, how did you think about this? I, I think what I would say to, to, to that point is that the community that, that brings people together in places like Stanford is incredible, right? Just because, um, as we're seeing sort of model or AI applications proliferate, where it really comes back to our previous point is how people come together and build really unique things. And, and I think that was key to what is a sort of Stanford college experience. And definitely through my experience I think for his use case or his application, I think that's really dependent on on sort of what how he sort of sees himself contributing to the space building or even working with different AI models. I, I think this is such an important point because in this age of AI, I think human connections are going to be even more important and, and, and experiences and spaces that allow you to build these authentic human connections are going to be so important. OK, so you have a lot of older people in the audience today. Um, what's your advice to all of us? I would love to say that the sort of you can be honest I'm OK when we sort of have a conversation about younger people using AI models and and being really comfortable with them, it's just as important as the older generation using it because AI fluency is just as important for people that are already in the workforce. not going to teach it to younger people to use AI. And the truth is, if, if there's this sort of anxiety about it, we are just at the beginning. The models right now are as dumb as they are ever going to be. What that really means is that when you look at a couple of months down the line, we are going to be in a very Different landscape and so being able to be really comfortable with that and having your core tools that you use um and and and get comfortable with that is really important because we're sort of at the beginning. It is only going to get faster, more advanced and more intelligent each and every model from from here on out. So being. Really comfortable with that is, is no AI anxiety, you guys. Kira, thank you so much for this wonderful conversation. It was great. Thank you.

The enterprise AI boom is moving from proof of concept to proof of profit. The chief of Google’s fast-growing cloud computing business explains how Fortune 500 companies, startups and public sector organizationsare building foundational AI platforms that blend intelligent agents, multi-cloud ecosystems, and next-generation infrastructure. He shares what’sworking, where adoption is accelerating, and how leaders are making the leap from experimentation to measurable business transformation.

Leading the Enterprise AI Revolution

Thomas, thank you for joining us. Thank you for having me. Um, we're gonna talk about agents, but not right away. Um, this is, uh, uh, a delight for me because, um, almost a decade ago in 2016, I was able to sit down just like this with your predecessor Diane Green and talk about Google joining the cloud computing business. And the whole vibe of that conversation was we are trying to transmit to the market, communication. To the market that we are serious about business now. This is Google, right? This is the consumer company, the company that we all know that it was a formidable thing and you know, as an Oracle, a longtime Oracle veteran that you know enterprise companies are built a bit different than consumer companies. So Today, Google Cloud is, by my understanding, you'll tell me, the fastest growing major cloud service, but it's off of a slightly lower base than your competitors, 13% of market share, uh, versus Microsoft's 20% and Amazon's 30%. Does that speak your truth? Yeah, OK, um. Why today are you winning more business compared to these other companies? Both because we have great products, people are finding them very easy to use, and they're getting real value from it. And we offer a very diversified product portfolio. Um, you know, our focus on cloud has always been, take amazing products, make them super easy for people to use, take care of the complexity that companies want us to deal with. Things like security, compliance, ah, but differentiate from everybody else across the strength of our portfolio. And so that's been why. And we've been, you know, we've had amazing growth rate. Um, Over the last 16 or 18 quarters, I think we've been the fastest growing. So if, if Amazon is the, you know, the the choice of, of the little guy, and that's how they get their volume, and Microsoft's supposed to be the choice of business, what is Google Cloud? We have a very diversified base. So we've done really well with startups, uh, particularly as startups shift to doing more AI stuff. We've taken a lot of share and. We're now the number one choice of AI startups by all independent reviews. Ah, we've done really well in traditional enterprises, banks, insurance companies, telecommunications companies, hospitals, ah, pharma companies. Um, and we've done well in not just The US, but in global markets, our revenue is highly diversified globally, which also people are surprised by how much strength we have in overseas markets, and that's reflective of the we've got a very, very focused plan on how we build and take our solutions to market. We've We also got great partnerships with many independent software vendors. We've got huge practices. I think there's 700,000 consultants who implement Google Cloud globally, which is a huge number. And so we've all that's come because of the demand we see from customers. Sounds like an enterprise business. One of the other things that struck me about what you're doing, and this is very on trend last couple of years, is allowing the use of any model, right? So when you choose to, when you, when you go with Google Cloud, you don't just have to use Gemini, you really have a, a wealth of options. That is that right? Yes. OK, so how in that circumstance, which is great, how in that circumstance do you competitively differentiate, right? Because you're not the only one who, who, who makes that, that offer. Yeah, so we, we offer. Our differentiation when you look at AI is to do AI really well, you need a stack of technology. And a stack is energy, chips or systems, infrastructure, models, tools and applications. And we're the only ones who offer all of it. We also offer a choice at any layer. So for example, we have our own amazing silicon called TPU as some people call it. Ah, but we also work very closely with Nvidia because depending on the kind of model you're running, you want that choice. The same thing, if you look at how people use our platform, ah there are customers who use a combination of models. Uh, Intuit, for example, the customer, they use some of our models, Gemini and our diffusion models, Imagine and VO, but they also use other models from other people for tasks that they want to use. I would not say that because you offer a stack, it's closed. I think many other companies don't have parts of the stack, so they depend on others for silica and other people for models. We have our own, but it doesn't mean it's closed, and so you get the best of both worlds. You said closed, reminding me of the open versus closed debate, which is usually in the context of the models, but you're, you're speaking more broadly than that. Is that debate dead? We've had it here before on this stage. Is the open versus closed dead? You're a tech guy. In enterprise, you have to have, for example, if you look at our models, they can access information from any application, database, collaboration tool, because in enterprise, you have to have it work with everybody. And so we've always had an open philosophy, uh, back in 2019 when we said, Companies will want to have cloud technology from multiple providers. The incumbents said no, they would not want it. Today, 95% of large companies use cloud from many, many different organizations, and so that gives them choice. And we've always felt for enterprises, they want choice, and choice means open. So I'm gonna ask you a, a big question and we're gonna have to unpack it a little bit, so stick with me, um. When you got started, we were calling this cloud computing. AI was not quite on the tip of our tongue the way it is today, um. We were calling it machine learning and deep learning I think back then, um, solely um. How have you had to rethink this service given these massive waves that have happened since for, and this is the unpacking part, stick with me, Thomas. The AILLM arms race, the AI demand boom from your own customers who suddenly all want to be AI companies or use it, and then the actual capabilities of AI itself and how that's changed what you do. So I'll, I'll keep you on track, but Let's start with the LLM race. How has that changed what you offer and how you do what you do? So we, we've been working on AI a long time before LLMs came along. That's right. And we've always designed for the long term. So if you look foundationally, when you look at what we were trying to solve for cloud computing, it was take technology and make it accessible to everybody by making it super easy for people to use it. And with AI we've always had the same philosophy. So, at the foundational level, we said people want amazing systems to train models, but we at the same time, we also realized you can't just train forever, you actually need to do inferencing well in order to do, to make money as a company. So we built systems for both training and inference. We also knew that The the most problematic thing that was going to happen was going to be energy, because energy and data centers were going to become a bottleneck alongside chips. So we designed our machines to be super efficient in delivering the maximum number of flops per unit of energy. So that's one example. When we built flops per unit of energy, yes, it sounds like a soccer metric, but I, I digress. So, the second thing we felt was, when you look at LLMs, you can't just hand over an LLM, you need a tool set around it, uh, in order to make the LLM really work for organizations. So that toolset needed to secure the LLM, it needed to access information, it needed to help you tune it. Uh, it helped needs to remember what you're asking it and handle long-running things. So we built an entire tool chain around it, and we gave it to developers. And to give you a sense, when we launched Gemini 2.5, uh, we went from 0 to 1 trillion tokens in like 7 hours. Because it's so easy now to get it and scale it. Uh, so that's the 2nd thing. Third thing, we always felt that along with those who want to roll their own as developers, there's always going to be parts of the organization that don't have the capability to roll their own. For example, a legal department that wants to use our models to understand uh contracts and which contracts have indemnification and some warranties. Uh, marketing people who want to create advertising and really tailor the advertising to the segments, and they want to create the advertising content using a model. Uh, for all of them, we built a portfolio of applications. Uh, we took Gemini, which is our app. And made it available for companies by integrating it with all the enterprise systems they have and we built some domain specific ones that have gotten great, great traction. For example, we built one to help people find products online, uh, because today when you search, I have never found any product online. Never. No, I'm kidding. I'm kidding because you're not used to it yet. I'm kidding. The internet is an addictive thing. Well, look, if I may distill, um. A lot of iterative building for to to to get what your your business ready for these successive waves. I wanna, I wanna go back to something you said earlier about power so. We are all well aware of the finite amount of power that we can generate on this planet. I'll go so far as to just let leave country, you know, international boundaries behind. Um, are you concerned about that? Yes, you've been building for efficiency, but the adoption curves have been. Massive and fast. I mean, we've never seen anything pick up like this before. Are you concerned about your ability to power your business, your ability to keep, you know, pace with the, the, the, the amount of demand that you and your competitors are seeing? Take us into your, uh, uh, your brain as far as the, uh, power side of thing is concerned. We are working on a variety of new technologies that would you like to share any here? Well, I. Tell you that there are 3 parts to what we're working on. One is to be as diversified as possible in the kinds of energy you can use to power computation for AI. Uh, it's, most people say, can you use any form of energy to power AI. That's actually not true, because if you're running a cluster for training and you bring it up and you start running a training job. Uh, the spike that you have with that computation draws so much energy that you can't handle that from, uh, some forms of energy production. So one is diversifying those. Second, being super efficient in how we reuse energy within the data center itself, so that you're not just bringing energy once in the data center, but the thermodynamic exchange can be super efficiently done and used. And third, some new fundamental technologies to actually create energy in new forms, and we're working on all three of them. Since you often ask, where are we applying AI, we're using applying AI, the control systems that monitor the thermodynamics in our data centers are all governed by our AI platform. The dog fooding, as we like to call it, right? Um, let's talk about the, the TPUs, the custom silicon that that Google is working on. Again, another trend, all the big tech providers, the hyper scales are very into the custom silicon thing. What role does that play for your business forward? You're already using it. You have been for a while. Um, are you trying to, you know, eliminate all of the chips from other vendors or is it in certain circumstances, where does the, the TPU roadmap go vis a vis your business? We've, I mean, we've worked on TPUs since 2014. That's right. So a long time before AI was fashionable, we were working on, on the silicate. And we kept investing in it because we fundamentally believed that you can change the design of an accelerator, which is what a TPU system is in fundamental ways. Uh, we offered that to our customers through our cloud. Ah, many companies are using it. Many of the top labs use it, many of the unicorns use it. And we are now broadening it to support new forms of computation. So for example, capital markets firms are switching from numerical computing to inference, and they're using it. There's people in high-performance computing, there's people in molecular simulation, energy simulation. So there's a variety of different domains. And at the same time we worked very closely with a number of companies on adopting and making our technology work amazingly well on GPUs, most importantly Nvidia. We've optimized Gemini and our models on it, and so that's been a long term partnership as well. Um, I wanna ask you one more question, but I, I, I do want to come to the audience as well. So, uh, just give me one more and then I'll come to you, um. I was going to ask you about that very, that very company in fact. So, uh, Jensen, Jensen Wong, the CEO of Nvidia, um. If he were sitting here right now, would he say that he was Excited and thrilled about the TPU roadmap that Google has, or is he concerned that it's going to harm his business? You should ask him. Oh, wow. No, but truly, I mean the, the big push for custom silicon. The big push for custom silicon threatens the business of, of these, of, of the, of the other chip makers. Obviously you need a chip maker to make the TPUs, a contract chip maker. Jensen is a great partner of ours. We've done work for a very long time. You know, I'll, I'll give you an example of the kind of work we've done. For a long time, if you wanted a, uh, a model that was uh an LLM that had a closed weights and didn't share the weights. But you had information in your data center that you wanted to hook up to it, and you wanted to bring the model into your data center, you could not do it. He asked, he gave us the idea and we worked for 3 or 4 years now to really optimize Gemini. And you know, earlier this year, we announced that you can take Gemini and run it on an Nvidia GP cluster anywhere. And we have to do it while protecting our model weights and our intellectual property, and that's an example of a joint project. We work with them on compilers, workload management, there's many, many things. Because I think there's there's way too much in the press on. It's a zero-sum game, when in fact, for those of us who have been working on AI infrastructure, there's many different kinds of chips and systems that are optimized for many different kinds of models. And as the market grows, we're creating opportunity for everybody. Yeah, yeah, understood. All right, let's go to the audience for questions. Forgive me, it is rather bright. Yes, right up here in the front, wait for a mic, and then your name and who you're with. Thank you. Hello. There we go. Hi, I'm Ash from Adeline. Um, Google works with a lot of customers through solutions engineers and getting them to implement LLMs within their organizations, customers, enterprises. And some of these customers, I'm sure they go through the journey of trying to build something but don't come to the other side, meaning it doesn't work for some reason or last minute it gets pulled back. What are the reasons from your perspective that enterprises today still struggle to push something out, either as an internal tool or to their customers, and what will we need to do as a market to kind of get them over the hump? Thank you for that briefly, Thomas. Typically, it's one of 3 things. They didn't design the architecture right. Second, they didn't clean up the data. Third, they didn't test the model and think about things like security, and if the model gets compromised, ah, how to handle it. Uh, and lastly, we've seen often, they didn't, because they didn't think of the problem the right way, they didn't think about how to measure the return on investment on it. There you go. Check yourself before you wreck yourself. One more, please. Your name and who you're with. Hi, Julie Morris, um, thought leadership for cybersecurity, etc. Um, thank you so much for this. I'm curious, so Google Cloud acquired or Google acquired Wiz, um, biggest acquisition, um, but there's still this big gap around model security and Uh, I'm curious, is Google Cloud going to try to step into any of that to do more? We, I think we are working on two different dimensions on, uh, with AI. One is how to protect against threats that are created using AI models, and second, how to protect AI models themselves. Uh, because a lot of company-specific data is now inside these models, and people are always worried about somebody hijacking them and taking access to the data. We have a lot of deep science work going on at Google on both dimensions, and you'll see us present more and more of it as we go forward. There you go. Thank you very much, Thomas. Thank you so much for coming back to Brainstorm. A round of applause for him. Thank you. Thank you so very much.

Adding AI to the enterprise can feel like reassembling a plane mid-flight. How should corporate chiefs strike a balance between reshaping their business for an intelligent era with maintaining the quality of operations that their customers and investors demand? And what parts are worth keeping proprietary? AI leaders from a trio of Fortune 500 companies discuss.

How to Rebuild Big Business for AI

AI is only as powerful as the stack beneath it. Compute, chips, data, and energy are becoming the decisive factors in who can build, scale, and sustain intelligent systems. But how much infrastructure will we really need—and who’s prepared to run it? As build-outs accelerate, operators face tough choices about capacity, efficiency, cooling, carbon, and even something as fundamental as cable management. In this session, technology and infrastructure leaders explore what it takes to keep pace with runaway demand while keeping costs and complexity in check.

Gem

Gem in Fortune Rankings

Videos