Transcription
Heat. Heat. N. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. [music] Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat. Heat.
The future isn't waiting. It's being written by every decision we make today. Technology is the pen. Humanity is the author. The story is still ours to write. And that story can be a positive one of a brighter future.
[music] Small actions [music] at scale don't just change industries. They change [music] history. We don't inherit the future. We co-create it with courage, [music] code, and compassion. Every startup, every breakthrough, every bold [music] idea is a vote for the world we want to live in.
[music] [music] [music] >> [music] [music] >> Heat. Heat. >> [music] >> The greatest disruption of this century won't be in technology. It will [music] be in how we choose to treat each other. Heat. [music] Heat. Heat. Heat. >> [music] >> Please welcome to the stage founder and CEO of WebSummit, Patty Cosgrave.
[music] Good evening everyone. Welcome to Vancouver. What an incredible city. This week you'll meet attendees from over 100 countries from Brazil and China to Nigeria and India. We're also proud to welcome official trade delegations from across the world all eager to build new relationships and create new alliances for the future. Canada is asserting itself as a global leader at a time when leadership quite frankly is in short supply. Tonight we welcome the Honorable Evan Solomon, Canada's first dedicated AI minister, and the Honorable Gregor Robertson, Canada's Minister of Housing and Infrastructure.
We meet at a critical moment in the history of technology. A battle is raging for the future of AI. A battle in a sense between open-source AI and closed AI. On one side, more than a trillion dollars has been bet on an almost singular belief that a small number of mostly American firms will provide proprietary AI services for a fee to billions of individuals and of course businesses all over the world. On the other side are open-source AI models freely available to anyone in the world with Chinese open-source models dominating the rankings. There are speakers you'll hear from over the coming days who'll tell you US closed AI models will win. And you'll hear from others who will tell you it's already over and free Chinese open-source models have won. And there are some who say it's still all in the balance. They'll be dotted across our stages over the next three days. Also dotted across our many stages will be lots of people with very different opinions on the future. I hope I think we hope as Websummit that you disagree fiercely with some of those speakers, but we also hope you leave websitummit with your views on the future challenged and perhaps changed and perhaps changed for the better as well. So to get our conversation started, I'm delighted to welcome to the stage representatives from the government of Canada, the province of British Columbia, and the great city of Vancouver.
[music] I'd first like to call in Canada's Minister of Housing, Infrastructure, and Pacific Economic Development, the Honorable Gregor Robertson.
>> Thank you very much. [applause] Thank you, Patty. Uh, thank you, everyone. Welcome to Vancouver. Welcome to BC. Welcome to Canada. We're thrilled to have you here. Also, I want to give uh proper respect to the land and the waters that we're on. These are the unseated homelands of Musquam, Squamish, and Slavetooth peoples for countless generations, and we are ever grateful for their stewardship and they're welcoming all the rest of us from around the world to be here. I am honored to be with you on behalf of our prime minister and uh all of us in the government of Canada. Over the next few days, this city will become a hot meeting place for so many of you innovators, investors, creators, uh leaders from across the globe who've come here to join us. Websummit Vancouver brings the global tech community together in a place that has long been uh a gathering for many across the world. between our continents, the Pacific Rim. Here, this is a city built on connection, on creativity, and on ambition. BC is home to multinational anchor companies, uh, worldclass universities, more than 12,000 technology firms, about 180,000, uh, tech workers at this point. We have two of Canada's five supercomputing sites and uh we are the Asia-Pacific gateway frankly for North America. Seth uh thank you for first and foremost uh for making time to be here join us in our city as innovators and investors and leaders. We have to push the envelope. As Patty said, we're at a critical time in our world right now. It's an essential that all of us take it to the next level. We have a world to take care of, a legacy to leave. Here in Vancouver, we built legacy companies from the early days of Electronic Arts, Sierra Wireless, Slack to Stem Cell, Hootsweet, Cleo, uh, Svante. We've got amazing companies that have been built and grown up here and scaled globally. Uh, that is the kind of momentum that this week is all about. and having you here to build on that momentum to create the companies, the successes and the world of our future. It's a great honor for us uh to have you here. Last year at WebSummit 2025, we had incredible attendees from around the globe. I know we're pushing it to the next level this year. Uh thanks to all of you being here and let's go forward. Let's build on the momentum. Take it to that next level. Strengthen, create jobs, create innovation. uh most of all build great relationships with each other. Thank you so much Meli Atus for for joining us here and have an incredible time in Vancouver. Mei [applause]
Hello everyone. Welcome to Vancouver, to British Columbia, to Canada. We are so delighted to welcome you here. This is our second WebSummit Vancouver. I want to thank our colleagues at WebSummit and all the many volunteers who've helped us uh get this to fruition. Websummit is an incredible opportunity for British Columbia companies. It's an opportunity to shine the light on the incredible talent in British Columbia, the companies that are building such extraordinary innovations, to look for partnerships and collaboration from around the world to attract investors to talk to international media. We pursued WebSummit to come here. Patty remembers it well. Ken remembers it well. We pursued WebSummit to come here for you, for the technology companies in British Columbia to shine, for you to find the partnerships that you need and show off everything you have. We're so proud of our British Columbia technology companies, the innovation that happens in our province. Take advantage of this amazing opportunity and have a wonderful web summit. Thank you. [applause]
Thanks, Brenda. Good evening, everyone. How's everyone doing today? All right, welcome to Vancouver. Look, it's an honor to be here with Patty to open the second Web Summit Vancouver and to welcome thousands of founders and in investors and innovators across our city. uh you're arriving in a place that has always been shaped by people who choose to build something new, people who saw saw opportunity here and people who decided to go for it. So my parents, they immigrated to Vancouver in 1967. And like so many others, they didn't arrive here with much, but they found something incredibly powerful. A place where hard work, risk-taking, and ambition could turn into real opportunity. That story isn't unique. That's the story of this city. And it's the same spirit that brings all of you here today. Because Vancouver isn't just hosting this moment. We're helping define it. We are a city of builders. Builders of companies, builders of communities, builders of the future. And this is personal for me. Before public office, I started a career in accounting, investment banking, private equity, focused on tech, and as an entrepreneur starting and growing businesses here, including a company now that we have a global platform, and we're actually a tech platform. And so, I've personally been on both sides of the table, building companies from the ground up and investing in the ideas and teams that have the potential to scale. So, I do truly understand what you're looking for. You're looking for talent. You're looking for access to capital. You're looking for a place where great ideas can grow quickly and grow globally. And that's exactly what Vancouver has to offer. Today, Vancouver is home to nearly 100,000 tech workers and one of the fastest growing uh talent pipelines in North America. Our strength uh it starts with our people. exceptional talent coming out of world-class institutions like UBC, SFU, UIC, and BCIT and continues with a globally connected ecosystem that makes it easier to scale here more than ever before. But talent alone isn't the story. What makes Vancouver stand out is how ideas turn into impact. Across this city, companies are solving real world problems at a global scale. In life sciences, companies like Abselera are developing next generation medicines right here in Vancouver, building not just breakthroughs but and an entire ecosystem around them. And in AI and advanced uh computing, local innovators are pushing the boundaries of what's possible. In clean techch and critical minerals, companies are tackling some of the biggest challenges of our time while creating new industries in the process. And importantly, we're not just cheering these companies on, we're working with them. At the City of Vancouver, you're we're using our buying power to support local innovation, partnering with companies like industrial to deliver world Yeah, you can clap. industrial, [applause] you know, helping to develop world-class technology that will help us uh, you know, host global events like FIFA World Cup and leave a lasting legacy for our city. Because when local companies succeed, Vancouver succeeds. And when Vancouver succeeds, those solutions scale worldwide. That's why over the past five years, more than $30 billion has been invested in tech companies headquartered right here in Metro Vancouver. And the momentum is real and it's growing. But here's what matters most. The next wave of companies, the next breakthroughs, the next global success stories, they're in this room today, right now. Web. Yeah, you can clap about that. Yeah, there's there's one right there. We already know there's like a there's a unicorn. uh happening there. Uh, it's where you know um web summit isn't just a conference, it's a catalyst. It's where ideas meet capital. It's where talent meets opportunity. It's where conversations turn into companies. So if you are here to invest, look incredibly closely to what's happening in Vancouver. If you're here to build, know that this is a place where you can scale, attract talent, and access global markets. And if you're still deciding where your next chapter begins, I'll make it very simple. There has never been a better time to choose Vancouver. But while you're here, don't just stay at the conference. Get out into the city. You, if you look outside, you can see how beautiful this place is. But it's not just the natural beauty. Experience the energy, the diversity, and the creativity that make this place so unique. Because Vancouver isn't just about somewhere where you do business. It's somewhere where you can build a life. And so to Patty and Casey and the entire WebSummit team, thank you for believing in Vancouver. And to our partners across government and industry, thank you for helping make this moment possible. And for everyone else here this week, welcome to Vancouver. Welcome to Web Summit. Have an incredible, incredible time. Enjoy our city. And I hope you guys are incredibly successful. Thank you very much. [applause]
Thanks, guys. All right. Follow the petty [music] [music] heat. >> [music] >> Please welcome to the stage co-host of Websummit Vancouver, Casey Laauo.
>> [music] >> Uh oh, here comes another speech about Vancouver, but that's it's my turn. As a born and raised Vancouverite, it brings me an immense sense of pride uh to have WebSummit here in Vancouver, and this second year fills me with an ineffable amount of honor being Canadian as well. Last year, we as a country faced immense challenges, but it was amazing to see the solidarity that swept across the country. Even though it was an extremely turbulent time, to me, it was a historic moment for Canada. Our first year here in Vancouver turned out better than even I hoped for. And a lot of you know, I I worked a lot to hype up the event. And this year, we're welcoming more than, as you've probably already heard, 20,000 attendees from over 100 countries over the next few days. Wow. That's just amazing. So exciting. >> [applause] >> As we kick off year two tonight, I can clearly see Vancouver in its pre-websummit era and its possummit era. It also hope happens to coincide with the enormous rise in Gen AI and how it's changing the tech landscape and the way the current Canadian administration is handling the geopolitical changes that are happening. Trust me, it is making an impact if my conversations around the world are any indication. Now, it's time for Vancouver to make sure it's seen and heard on the global tech stage. This week is one when the world's eyes will be on Vancouver. And I know from companies like Cleo, Photonic, and Gum Loop, and fresh rounds of investment into companies like Moment Energy, Juno Industries, AK Robotics, and PH7, as well as the triumphant return of Ryan Holmes to the CEO chair, Hoopuite, that the city is in good hands. So, I I just want to quickly shout out the new stages we're including this year. Games Summit, Film Summit, and Planet Tech. Exactly what they sound like. Three new stages we are launching this year, as well as the very first WebSummit hackathon on Thursday. And then also look to our socials this week for our joint crunchbased report on funding funding raised since last year's event. I'm sure a lot of people want to know who who raised money. This report will surprise you. Many people ask me if Vancouver can be the next San Francisco or Senzhen. I always answer it doesn't have to be. It has its own strengths, talent, and creativity and its own leaders in verticals like hard tech, quantum, and clean tech. And unlike in the movies where Vancouver always stands in for another city, I can definitely say in the tech world that Vancouver is Vancouver. Here's to an amazing Web Summit Vancouver 2026. All right, let's get things started. Our first talk of opening night. If you're tracking the bleeding edge of AI agents and open- source disruption, you already know the name Sigrid Jin. A software engineer and harness engineering expert. Sigrid achieved legendary status by executing a lightning fast clean room rewrite of the clawed code source, which hit an unprecedented 100,000 GitHub stars in a single day. From utilizing a staggering 25 billion tokens to automate complex engineering workflows to pioneering how agents orchestrate tool execution loops. Cigarette has become a central figure in the movement to build open-source alternatives to proprietary frameworks from companies like OpenAI and Anthropic. Whether presenting on the future of tech or pushing code that defines the next decade of infrastructure, Sigret represents the new era of the architect creator, building the massive IP engines that power the modern developer ecosystem. So, please welcome in conversation with Madison Mills from Axios to talk about the intricacies of AI and open source development, Sigrid Jin.
We are witnessing history right now of GitHub repo with 100,000 stars. The fastest thing that I've ever seen in the history of GitHub. [music] >> Anthropic accidentally uploaded a file to a public repository and exposed some of the source code of cloud code. Being a token maxer just means you're using as many tokens as possible. The term is also just a reflection of this moment we're in where token usage has come to become a a key metric for productivity. Sacred, thank you so much for your time. Uh, I can tell how popular you are based on how many people were trying to get your number in the speakers lounge. So, I feel like we're going to bring that energy into the stage. Uh, you've been called one of the most legendary AI coders in history. Uh, cementing that title on March 31st after a source code leak allowed you to replicate Claude's codebase into what is now the fastest growing GitHub repo in history. A huge accomplishment. Uh you also did it while you were on an airplane which I don't know if a lot of people know. Take me into that day and how you made that happen.
>> Okay so um um like I was in um kind of uh in in in in a team of roughton which is one of the um the famous AI hackathon uh like on top of the rough group. So for so for those who have never heard about rough root, you just try to make sure your AI agents to write your code more than um three hours and four hours just let your agents to do uh autonomous work instead of like waiting the humans uh to the prompt in the uh in the in the between the middle of the um the agent group. So there was a huge success of the team attentions uh the hackathon roughton and after like I flying back to Vancouver um I just got uh the my friends call uh the saying that the entropics code the code source code was in like uh like like available in online on Twitter. So what I did like just to try to download it on my local computer but my girlfriend is really get worried about it because like when it comes for traditional ways of thinking so the um the copyright is they always give me an issues when it comes for licked like software materials. So what I got to think is like what if I can rewrite the whole software code into using the um the AI agents to into different languages between uh because like the same specifications the same specs but in a different languages and the AI agents can take autonomous loop for more than a three hours and a four hours to rewrite into different languages different structure but the same uh specifications. So like it went super viral uh what I take and try to bypass its cop uh copyright materials. So like it was like super uh kind of went viral than that that and um I was in an air airplane at the time. So I was an Air Canada flight. So I took a like inflight Wi-Fi to do so because like yeah even like closed code they're like in conversating between in in the text. So although like inflight Wi-Fi is super slow even I can't let my own agents can like work in a continuous loop for more than three hours and four hours. So it was very hectic that even you can like rewrite uh the specific software uh into the um other ways other the medium of software within hours even if you have a lower low latency of speed of internet. So I think it's really um next level of the software engineering I think.
>> Well, it also speaks to the systems you already had in place.
>> Yeah, of course. Yeah. So basically like um I always uh try to say that the um uh like software engineering style uh the way of working is really changing a lot. So for example uh from some years ago we are just trying to edit the files on the uh on the specific some text editor or some even a notepad when you go get to a very long ago but like we are now trying to have the new ID new ways of uh integrated development environment as um kind of messengers. So you can just invite multiple agents in a single messenger. You can just invite them in the discord. You can just invite them to Slack and let them agents to assign to be assigned with the specific personas and specific roles of the jobs and then they can make them like communications in anonymous way. So try to push the AI agents to the level of they have never been to so far. So uh that is the my older systems that I try to have in these days
>> and and we'll get into how that could impact the future of work but I want to pause on the copyright uh of it all that sounds [clears throat] like part of your mind as doing this. Uh were you concerned about anthropic reaching out with a takedown?
>> Uh well I think like it's bit kind of in the gray area because um when it comes for the uh the uh the foundation labs uh they also have in elegant to um kind of try to bypass the copyright ros when it comes try to get um a pre-training data because like well it's really difficult to like get a like appropriate like abundant amount of the data although we know the scaling law works. So it's always been a problem to like get scrape beyond like those qualified you know quality data in the first place. So always like it's been a like kind of gray area like what I want to say is that we are trying to make sure how we need to adjust the copyright ideas because the copyright ideas is been uh been implemented really long ago like 100 years ago like where it stands like uh the everything the um rewriting of materials would be very slow it takes some time so there's such a kind of premises but when it comes for the agentic era like we are now taking the new stage because the speed is like unpreently fast and even they're really accurate enough so that we need to reconsider the what kinds of the copyright material kind of kind of doctrine should be rewritten and reconsidered in the first place. So like I think like there's lots of the uh uh ways of thinking ways of the um kind of discussions like amid the days uh the lot of industries but like we are now like just trying to make a new models new a great models in the forward. So like those um like voices are really being unheard in these days. So I think like what what I can do is to try to raise those like uh questions into the front by just rewriting this specific very um like uh uh like software into the different languages so that you just try to like spike those kind of conversations in the front.
>> Well, and it's interesting because you're using a similar logic that the AI labs have used to train their models on copyrighted material as well. Um were you surprised that Enthropic did not reach out to you?
>> Yeah, so true because like it's really become a gray area because I think like the um the mod of the um AI frontalter labs is not on the specific harnesses because they they are they are mod is in the um data pipeline the pipeline they try to make a mod in the first place because the harnesses is the code even that generated by the other AI. So even the Borist he just mentioned that the code is being re like written by the code itself. So basically the code is not a mod anymore. So at least like um although like I can just try to copy the um specific popular software but at least like they have their own like marketing team they have their own like uh like like folks to really enjoy such a vibe of the team. So it's really difficult to not copy on on their the fan base on their own like um like culture their own like folks right so like it's a matter of like moving the mod between the code itself to the culture to the um something the environments they the omit of the software I think
>> yeah and what does that signal to you about just who kind of owns the code if an agent is reading proprietary code and then you're able to rewrite it perfectly in a different language than who does own it.
>> Okay. So I think like um code itself is becoming more public goods because like software is now getting really cheaper and even the um the the the Mark Anderson he posted on Twitter saying that today's software is tomorrow software's um the prompt. What I mean is that um like software is getting more easier to getting a like kind of copy getting to like like kind of like kind of kind of getting a reproduce in the first place. So that we just try to make sure to focus on not just on the line line by line of code but to impact of the codes can generate in the first place. But I think like those idea of ways of thought like not to focus on the lines of the code that much but to try to focus on what kinds of impact the software can make is always like applied to today's like yesterday's software industries as well because there's a lot of theum intelligent like engineers are there but the thing is like how can we monetize those software in the first place and even like AI or even like other some big deep tech like blockchain for example they're always getting like in in in in the challenge like how you guys can to like monetize how can you make sure to have their own use cases of their own software. So like it's like a more than an idea of discussions of the lines of lines of code. So like the um the uh discussions of the owners of the code, discussion of the copyright of materials of line by line of code is getting much not important in these days. I think
>> it it's such a valuable question though because when you think about Anthropic, this is a company that is reportedly valued at nearly a trillion dollars now.
>> So true. and in a couple hours you were able to rip their highest revenue product.
>> So true. >> What does that tell you?
>> Okay. >> About where the value is with software.
>> Okay. So I think like value itself is that they have their own models, right? They have their opus. They have their mythos. This really amazing model. I really do use a lot every day. And the thing is like they can make a more better model because they have their own life cycles. they they have their own harnesses. They have their own model pipeline. They have their own like training like kind training data. They have their own training code. So basically these are the um all all the components are not irreplaceable in in terms of their mod systems. They have their own ecosystem. They can generate a much better model for every month. So like it's not a thumb something that it cannot be like cop in the first place. It's like a infrastructure. It's like a life cycle of the ecosystem. So like it's it's just much beyond just like coping the line by line of code. So they do have their own mod on there and open AI they have their own mod on the um like kind of generating the model pipelines and making a better better models based on the hu like users feedbacks on the on their harnesses which is called codecs. So like they like have their own fly. So that's the mod. So it's just beyond line by line of code that's that's why trying to like persevere.
>> Yeah. And you've talked about how you are personally agnostic. You don't have a favorite AI model, but what have you noticed is the key distinctions between them and what makes you go with one over another in
>> so I think I really love the idea of neurodyiversity because like every models they have their own different philosophy of being trained, right? They do have their own different data. they have their own different ways of training and they do have their own different ways of like um getting um kind of like interest from the foundational models right so if the um models labs try to focus on more ethical ways then the model itself is trying to focus on those implications and if the founders labs try to focus on more on programming work then uh the models getting to learn more about some the how to code in the python or different languages so basically it's like a more neuro deeper It's difficult to like converge into a single models to work at best to rule the all the world. So I really love the idea of trying to diversify the models portfolio as you like invest in a different portfolios in your stocks. I think like if there something like if you have your own um like for example specific models getting upgraded but like when you come to looking at the benchmark of your own your like for example your writing task is getting enhanced while the coding task is been like inferiorated then you can just have some room to uh make sure to replace in different models and even so like models is like also they have their own opinion and ideas of what they have generated. So try to make sure to um other models to like criticize at first and try to make some consensus as like um like like carpassy try to like focus on like giving a more ideas of the Adam council. I really love the idea of that because like model itself treat them model as human beings they also have their own opinion ideas and try to make sure to like debate each other to have consensus and I think like it's a way of like approaching to a better agent engineering these days.
Have you run into problems where you have trouble getting a model to believe what you are telling it about your work?
>> Of course, because the model itself is like getting trained to follow instruction of users rather than uh giving referral of the users. So that I really want to to make sure the u models to have their own ideas because like I do not have this clear instructions when I uh even working with um specific ideas where I have know that idea that much much clear about that. So I just want to make sure the agents can rebut on myself and even can claim on myself. So in that way I just need to give a full autonomy just treat the models as like co-workers the co- follows of me. So that is the one of the systems I just try to like focus on these days.
>> And you've talked so much about the importance of input and being kind of a manager of agents. How do you maintain that creativity and executive functioning when you're using so much AI in your day-to-day?
>> Uh so my creativity. >> Yes. >> Okay. So basically like um I I do not want to fully rely on the AI itself because AI is just just treat AI as a human beings like AI is always like prone to some errors. Uh AI can make a mistake while AI is really smart by the way but they have their own personalities. So I think like I can join on the AI's discussions rather than being a spectators of some somewhere outside of the circle. So, I just want to find a way to intervene on their loop or at least I can just like kind of cast my opinions to my agent. And sometimes the agents can like agree on me. Sometimes agents can rebot on me. But at this I just want to make sure that I know the the timing where I can jump in or where I can just hand off so that I can just make sure they um the agents can work autonomously. agent work can I just want try to make sure to enhance the environment where agents can work they're in the know full um the the um potential and possibilities I think
>> but ultimately you're still in control.
>> Yeah. So true. Yeah.
>> So one of the things that's interesting about hearing you say you still want this human in the loop is that at the same time you are kind of the poster child of token maxing. You used 25 billion tokens in one year. Um how did you finance that?
Okay. So basically like it was basically um kind of a bit of like um some limitations between uh yes last year because like uh when you look at yeah last year the the close code beta was released um last Februaryish stuff and um there was no such like weekly and daily limit till the um late of the August and early September. So um so like close code was released on the last fee uh the 2025 February and when I just tried to use in so net for 3.5 with um clo code and um I thought thought like it's a way of the um working for the future. So I just tried to get a multiple account by myself and try to subscribe the so I have at that time I had a 10 clothe pro subscription at the time. So it
>> how many do you have now?
>> Uh well I do not need to and I have those one because there's lots of the um um like alternatives at the point like I really love the codex subscriptions. I really love the kiny as well. I really love the um GRM too. So I just want to make sure that to give us some some like attempts give us try for the different models to the specific task and if the certain task been passed for the certain models then they can like join my systems. always like try to make us a benchmark like just try to like so there's new grass here new grass from the college and you you gave us some exam if they pass it you can just have some chances to hire them right so just think in that way
>> so when you had your 10 claude subscriptions I assume you also had subscriptions to other models
>> yeah so true
>> and you were using so many
>> sonet opus so what was what was the final number how much
>> uh well honestly I do not take the kind kind of the make make a like calcul ulations on that but like I think like it's more than like 15 billion because like there isn't such a like uh vibe rank dashboard which is unofficial on the Reddit so I just like posted on my like to like uh the entire usage on the on the dashboard and um it was ranked in the first so I was in a viral in the fails uh day of the uh Twitter's uh the access the one of the like um the uh popular podcasters and he's YouTube that I was in there so I just try to get some virus from that So it was in the last Augustish uh Septemberish I think.
>> So 50 million.
>> Yeah. 50 billion I think.
>> And so you were not paying for all of these tokens yourself.
>> Yeah. Because like there was some Yeah. There was some sort of subscriptions for unlimited usage. Yeah.
>> And um after the um early of September weekly limit has been introduced. So that uh Yeah. So I cannot like let my cloth to use like like a pouring um pouring of water anymore. So try to find the other alternates from the Chinese models or open the open open AI codeex. So try to diversify the models portfolio from that one.
How hopeful are you that some of the changes recently announced their partnership with SpaceX to get additional comput?
>> Yeah, of course. So basically I think the mod like results and electricity and the data center itself because like well we know that um uh well mod is getting better and better. They have foundation labs they have their own um kind of the life cycles to make sure that model getting better but things like we are now still living the days of the scaling laws. So scaling law always works and scal law is beautiful but we are lacking the data we are lacking uh the electricity we are lacking the uh abundant number of the data center. So so I think like the mod resides and the electricity self. So I think like the the movement of the law office is is so much like kind of um respectful I think and even for like I know I know like some of the uh uh regions like Dubai or Middle East they have the plenty of the electricity data centers there. So um I'm also like participating as the hash to one of the um famous the global VCs from the uh the focusing on AI issue stuff. They also like try to have their uh headquarters in the Dubai and um uh uh I forgot the number the city of of there of Davi Abdi there because like they feel like there's lots of chances because like at the end of the day it only remains the data center it only remains because I training data training recipes they are like getting like publicized then you're open source but who gonna runs in the first place if you if you do have there such like basic materials basic law raw of like some in like infrastructure you cannot even reproduce in the first place so like it's getting important I think
>> but it sounds like you don't think it will be as expensive as it is now forever
>> yeah so true so it's getting much expensive more expensive more expensive more expensive because like model itself we need to put more scalings infrastructure to make sure the model is getting better is it has never been re rebutted in the first place. So scaling law works well. It's really beautiful, but we need more like resources to to make sure the life cycle happens.
Do you still feel pressure to use more tokens? Do you worry about that?
>> Oh yeah, of course. Because like um are there anyone who have not used clothes here or codex? I think not. Right. I think not. So just imagine the um any AI frontalter labs just rock pulled so that you cannot use any models in the first place then everyone going to like a panic at the panic I think. So even for any sing of software engineers who are really like pessimistic on AI they even relied on at least basic chatbot right. So we are now trying to make sure that we know that what AI can do have the more chances for the like some opportunities. So like at least that we are we only use AI on software engineering at most or some like creative writing work. But what if we can just like um like re-imagine the whole infrastructure of government for example. What if we can just transform like we can just say okay we can just take like say as like AI transformation AX of these traditional companies traditional enterprises to reorganize its corporate structures into like co-working between humans and agents then we need to want to push more tokens to to use right so it's getting more important right
>> what is your advice though to other people who talk to me about you in my interviews with them as someone who stresses them out because they want to make sure they're using as many tokens as
>> Okay. So, basically like I know like someone is like talk like pushing some of the ideas of like making your harnesses to be a more token efficient and um I I know it's really like getting more importance topic but I think at least like majority of people majority of the folks they do not uh experience the um kind of the the how much AI is powerful in transforming your or like your or or structure. So what I try to advise is to try to ignore your own thought about AI, try to push on your personal project to make sure to use a lot of tokens as much as possible to try to do some token maxing and to like look at what AI can do instead of you, right? So I know like some of the um the folks they just try to like ignore the um the the how much AI is powerful because they just try to only use a free subscriptions only 20 bucks subscriptions because like intelligence itself is getting more um kind of like like get getting into a more uh trying to put the more importance of giving a more like high intelligence into um like expensive subscription plan. So I think like at least two like use around 200 B subscription plan at least to try to like kind of paying for the um expensive plan you can like make your own like taste of how AI would be um powerful now or for future. So I always try to persuet my uh my friends to try to spend your AI uh pricing as much as you pay for your own house rent because like it's really getting yeah it's really getting importance because like you can get a like return on investment because you have such an investment there
>> but what is the return on investment
>> return on investment you can just run multiple businesses right you can just like delegate your like routine life routine life of work. For example, I do not like manage my uh the calendar inside of me. I just try to like leverage my own Hermes agent. I really love Hermes agent and open class to like manage my own schedules on behalf of me and even the open clause my own hermes agent which resides on my discord. They just try to like kind of purchase my own um daily goods instead of me and they just try to ask my permissions to like proceed to check out or not. So it's getting more like important to like I think because like you can just hire your own agents everyday life for token for the GPD 5.5 like like X high for every single single days of task then they just like hiring for part-time job is much cheaper I think. So like it's better for like kind of making a like kind of balances but I think like if you want to know the future of AI try to do token maxing and if you know the taste and if you going to believe have your own like thought and you have believe on the AI then you can just try to like optimize your own infrastructures or harnesses at least I think.
So the advice is to continue to token max it sounds like. Uh but for business leaders especially in the room um Uber CTO recently talked about blowing through his 2026 IT budget in the first quarter because of token maxing uh what is the actual metric of success when you are using tokens so much how do you actually measure it's a really good question because like um when it comes for AI engineers u machine learning engineers that they just relied on the uh the um uh like old-fashioned benchmarks and the benchmarks is now getting obsolete. I think for like evaring the models um I think that everyone needs to have their own benchmarks. Everyone needs to like create their own evaluation criteria to like evaluate their model. So I think like well you are the domains expert in certain task or certain industries. So I think that you can just have some um chances or time to look at what you have done so far for in a regular work and you can just make sure you can just have your own um for example exam questions if you have chances to hand over your work to your junior or junior engineers or your new newcomers right so try to make sure to have your own benchmark like it should be like but it should not be on because it's kind of like complicated Like it should be like very bareium questions or bareium like assignment and make sure you're to see how the models behaves on the task and you can just have your time to um modify the certain models behavior by changing your prompt or changing your own harnesses or sometimes you can just change your entire model so that you can just find your best fit on your harnesses. So I think like each different harnesses would be evolved for different task for different industries. So I think there's no such god harnesses in the front but everyone has their own pain pains. Abraham has their own task. So that you are the expert. So you are responsible for supervising the AI models I think.
And in our final minute sacred uh have all of the AI labs tried to hire you at this point or who has reached out?
>> Okay. So lots of folks haven't reached out to me. uh but I'm working for basically like I'm working for psionic AI which is the one of the AI infrastructure company in ape region so we have our Asia Asian tractions in Japan Korea which and Thailand and we are now trying to like serving they called storm platform which is the enterprises allin-one platform for um serving for like uh retrieval solutions and parsing solutions and um I have worked for them two years and a half uh for as as like founding member a very early member of that and um yeah I'm rather than just getting accept the offer from the um lot of from left that I got got I just tried to start my own new um startup starting next month so yeah I'm just going to found with my own co-founders thank you
>> thank [applause] you >> well that's all the time we have so thank you so much thank you yeah thank you thank you >> thank Yeah. [music] Heat. Try not to be boring. >> Hi. Oh, there we go. Okay. Canada is currently executing. Thank you. Thank you for that. Canada is currently executing a transformative national vision to secure its position as a global AI superpower through a multi-billion dollar commitment to sovereign AI. Central to this promise is the Canadian sovereign AI compute strategy, a historic investment exceeding $2.4 4 billion designed to build domestic supercomputing infrastructure and ensure that Canadian intellectual property remains protected on home soil. Today, we are at a unique crossroads of policy and power. On one hand, we
Have a government that has moved beyond talk, appointing a dedicated minister and investing billions to ensure sovereign AI is a reality with a clear assignment. Turn AI and emerging technologies like quantum into the engine of Canada's next chapter of growth.
On the other hand, we have a Canadian-born global champion helping lead a 20 billion enterprise AI company following its recent merger with Germany's ALF Alpha, creating a transatlantic AI powerhouse. To talk about how we as Canadians scale this innovation without losing our ethical compass. Please welcome in conversation with Bloomberg's Thomas Seal, Canada's Minister of AI and Digital Innovation, Evan Solomon, and the Chief AI Officer of Coher, Joel Pino.
>> The map of the future is not drawn in ink, but in algorithms. For a modern nation, AI is the new electricity, an invisible force that illuminates every industry and empowers every citizen.
>> There's an aspect of sovereignty that's really about building a robust strategy, having control over a technology.
>> Canada named its first ever Minister of Artificial Intelligence.
>> We got to get it right because this has got to be one AI for all. It's coming and our job is to make sure it serves people that people don't serve it.
Hi everybody. Welcome to Vancouver and welcome to Websummit. My name is Thomas Seal. I'm the Vancouver bureau chief for Bloomberg News. Really delighted to be here with this uh panel that is really the center of Canadian AI. Evan Solomon, Minister of Artificial Intelligence and Digital Innovation for the Federal Government, and Joel Pino, the chief AI officer at COH here. So, really excited to get into some of the issues here. And I think the title of the of the panel uh means we should probably define our terms. So, what is sovereign AI? What does that mean to both of you?
>> Well, first of all, great to be here. Uh, great to be with Joelle, incredible pioneer. and Thomas, thanks to you and great to have Web Summit in Canada. Also, great to be the only person in this huge room wearing a tie. Uh, this is loser move, but that's it. Uh, my former self would be like, why is that guy wearing a tie? Uh, look, so we have a lot to talk about and I want to welcome people who are from this ecosystem here in BC, which is amazing, and around the world. Um, we have this is an amazing time to be in Canada for AI. Um, and as the first minister of AI, it's because we understood that this we are at this unique moment. And the moment that we're in is really a moment of dramatic political realignment that's happening at the same time as a technological revolution. And because this this political realignment around the world that our prime minister has talked a lot about and the advent of AI as a transformative general purpose tech this has created an enormous opportunity and one of those opportunities is sovereign AI and what does it mean and in some ways it's the most challenging policy idea for any country around the the world right now to define it. One definition and there's political, technical, legal, operational definitions of it, but the political definition of sovereign AI is to make sure that you have the AI ecosystem and the infrastructure to protect your government and your citizens most sensitive information, data, and workloads under Canadian law, not subject to the coercion of an outside force. Okay, this is really important and every country is making sure that they have the infrastructure and the capabilities to protect sovereign data and most important Canadians or citizens data and privacy under the laws of their own country. And that means to do that you need to build up the entire stack from the ecosystem of the talent and the researchers like Joelle to the energy infrastructure and we announced uh moving forward with a massive AI factory from a company here called Telus. Three new data centers, two of them here in uh Vancouver. Closed loop uh heat reuse really brilliantly high sustainable data centers that are built the right way we think. So you got to build that infrastructure right up to the large language models and Joel will talk about cohhere and and our government support of cohhere uh and the application. So building that ecosystem. So that's sovereign by the way that applies to quantum too. So really challenging an important moment. What does sovereignty mean? And the last thing I will say, sovereignty does not mean solitude. We are not trying to put a digital moat around Canada. We are open for partnership. It means resilience and optionality, but not solitude, the best innovation, American, European, Asian countries, Brazilian companies, South American companies. We want partnerships. And that's why we've made sovereign technology alliances with Germany, with Europe, with India, with Australia because we want partnerships with like-minded share valued to to build the scale we all need. And this is a massive opportunity with this geopolitical realignment to make new partnerships and build new alliances to make sure that the resilience we have is part of our sovereign and secure uh AI.
>> Thanks Evan Joelle. Does it mean something more technical to you? I mean you've been an eminent researcher in the field for many years still are. Um, so yeah interested your take on how you would define sovereign AI.
I'm happy to chime in and I I'm at least glad we're not starting by like defining AI at least you know we've progressed since past years hopefully. Um, I I'll maybe share the point of view from from an enterprise perspective. You know Coher builds AI solutions. We build our own models. We build a Gent AI platform. We also build sort of a governance platform and in many cases I think what we hear most from enterprises of what they're looking for especially in the last few months is the ability to control their stack to really have the ability to choose what they are building on whether it's the infra layer whether it matters that their data is housed in a particular geography what's going into their models having the ability to choose whether their models are going to reflect the languages that they need, the types of cultural information that's relevant to them. And so all of that control gives them the ability to really build the products that they want to build and the services that they want to build. And so I would say that's the what comes back most often >> uh in our conversations with customers around the world.
>> Can I just ask Joel a question? uh because sovereign digital sovereignty means control and choice.
>> Yeah.
>> You need choice and you need some control.
>> But one of the things that we've worked on as the government with cohhere is this sovereign technology alliance and um with Germany and Canada both the countries are interested in in the question about sovereign control and this merger between cohhere and alfalfa has been a really important moment. So can you know we want to make sure that people have choices between hedgeimones and hyperscalers not that you don't want to use hyperscalers there's a role for that but choices and control and we believe that and we've encouraged and very much played a role in the cohhere alalfpha merger and there's more to come on that maybe it it might be worth talking about from your perspective I'm curious Joelle what that contributes to the sovereign capability and why that merger and partnership is so important.
>> It's it's been really interesting, you know, having these early discussions with Halifa over the over the months and and we have many more to come. I would say in many ways, you know, we we started on on similar trajectories building out models. We both, you know, teams have incredibly talented people and and over time the the ability to to build models, it's an expensive business and and frankly, you know, having partnerships around the world is beneficial for us uh in order to benefit. You know, we do need government support in terms of building up the right compute infrastructure to do that. Germany was investing in that. Canada has invested and will continue to invest in that. And so there's definitely some some ability to to build bigger better models when we work together. Um, the other thing for cohhere is, you know, we've we haven't been shy about our ambition to build a global company. We have our roots in Canada, headquartered in Toronto, co-founders Canadian, but but we've been very open about our our goal of building a company that provides solutions globally. Well, if you're going to go build solutions for Europe, Germany, as well as other places, you need partners. You need partners who understand the needs, partners who speak the language, who really understand how to bring that technology in a way that that meets the moment. And so, partnerships that we seek around the world,
>> which was the big news two weeks ago and and continues to be. So, no, it's great uh because I wanted to ask about some wording in in the release about that deal, which was describing the combined company as a sovereign alternative in an era of AI concentration. Now, we've been kind of getting at this idea already, >> framing it as a problem, a thorny problem, but also an opportunity.
>> Now, specifically in terms of like a market opportunity, our customers now coming to you, our companies, our countries coming to to Canada's government and saying, "Okay, yes, obviously there are a couple of superpowers, but we want more like what's the customer reaction and the demand like on that front for alternatives?"
>> I I would say the the signal has been incredibly positive. There there are a number of customers across the world but a lot in Europe who are looking for good sovereign options and who are looking for models that um reflect their language models where they can understand where the data is going to be used in terms of training where it's going to be used where the inference is going to happen and and they're very interested in this in this type of in this type of solution of what we're building with alphalfpha
>> you said in the last few months which I guess we can kind of all understand what that means but that that interest has ramped up in 2026, has it?
>> I mean, we we've had that interest in terms of what Coher is offering, which has probably contributed to this deal, certainly in the last couple of weeks. It's been paying off.
>> I I would add like the reaction from investors has also been super exciting to see. I think, you know, they're looking for these kinds of partnerships who who have, you know, an ability to to grow what we're building together. And so, that's been really encouraging to see. Training models is an expensive business. Yes, absolutely.
>> Yeah. A huge reaction um to the sovereign technology alliance from other countries who are very keen. Uh, we have a G7 tech meeting at the end of this month in Paris and there's an enormous amount of conversation about the Coher Alfalpha deal partly because it was a pioneering deal in sovereignty that this is going to have a Canadian headquarters, Canadian IP, Canadian innovation and a German headquarters and it allows those countries and others who want a platform that allows them to uh build their sovereign stack with others. and you know we need scale like this is a competitive business that's the large language model and and there's only four countries that really have big large language models for this Canada's one obviously France Europe or France China and uh the United States so you know Canada's role is to build on some of its major competitive advantages we our role as a government is to boost those cohhere is one of them and now you know they've got the scale and the velocity to keep growing. But you know here in this audience um there's lots of application models. So so the layer on top of the large language model is kind of where the macroeconomic action is. You know where we're going to see the big lift for our lives and what AI does not what AI is what it does for us. the genius innovators here in healthcare, in life sciences, in agriculture. That's where you really see the big impact on the adoption. But you know the populations have to citizens have to trust where their data is, where their jobs are going to be and how this is going to grow in a reliable proworker sustainable way. So we know AI is growing but you know we're in char we have control over how to do this people first innovators first and summits like websummit where people can connect and the innovators build on top of large language models like cohhere and there's going to be open models as well. Um, that that's where the action is. So but we've got to make sure that we're supporting that as well. So there's, as you say, there's a lot of work to do as we're as we're building this industry.
>> Yeah. I mean, it's an unenviable task and I'm just thinking about the the stack up to the usage and the adoption, which I know you're working on as well supporting that. But, you know, you described it from the the power to the the data centers, the large language models, the applications. There's also the the hardware of the chips as well. And so I mean what is is it possible at this point to sort of categorize which bits are are sovereign and non-negotiably so and which bits you can have more flexibility on. How are you thinking about that question at the moment?
>> Great question. So again you can't build it all alone. By the way, no country can do it. not the United States is partnering with. Everybody needs to work together. Which is why, you know, here in Canada and I think around Europe, I think in many countries around the world, we want to fight for a rules-based order where we can make trustworthy long-term alliances where we work together to promote the kind of values that we all share. I mean, in the end of the day, whatever technologies we built, we want to reinforce the kind of society we want to live in. I mean, we're here in Vancouver. We want the nature to be here. We want people to come from all different cultures and and live happily and grow happily. We want academics to share information in a way. And we want citizens to trust that they are not going to have their information and their privacy abused. So, so there's a value in in in making these partnerships. Part of the stack. The best way I think to frame it is um build partner buy. Some things we're going to build here and we're going to build because we have the competitive advantages and we can build them ourselves and we want to press on that. So that's >> where where are Canada's competitive advantages? Jump >> in. What we have we have massive competitive advantages in clean energy availability. We have massive competitive advantages in researchers. We you know we have our three national AI institutes. You know Yosua Benjio in Mila in Montreal, touring prize winner, Jeff Hinton at Vector in Toronto and Rich Sutton the touring prize winner in machine learning at Amy. So we have three national AI institutes that are churning out talent. some of the best university systems in the world here in BC and around churning out talent. I mean Waterlue graduates more engineers a year than Stanford. So we have massive talent, huge tal uh access in data and in some applications but there's weaknesses that we have. We are having to bolster access to capital commercialization and you asked just specifically about chips to finish it off. Look uh Canada has an worldclass advantage in photonix. So this is photonic chips and semiconductors. We have the only end to end photonic fabrication lab in can in in North America and one of the best in the world. So we are going to double down on that. We've just decided to open our Canadian photonic fabrication center up for uh private investment. So we are going to crowd in capital on that aspect of semiconductors. But there are some parts that you know we will encourage but we are going to have to either partner or when we buy because we just don't have it. We're just going to make sure that what we buy is reliable and safe and and is helpful for us. So there's going to be a build part, there's going to be a partner part and there's going to be a buy part.
I I think we also have to be mindful of just how fast things can change, right? Like you you don't have to look too far back before, you know, LLM's rushed onto the scene, agents and so on. And so there's still a lot of uncertainty in terms of what's going to be the value across the whole stack where the actual value is going to be created versus which part it's going to be commoditized. Um, even with respect to models, you know, a little while ago, there was definitely a sense that you need to have the ability to train your own models. It turns out to be very expensive endeavor and few are left, but nonetheless, you want to have optionality in terms of which models you're going to use. Now, a lot of the value is moving up the chain, whether it's in terms of developing platforms or even in terms of the service layer. We talked about, you know, building large people are worried about employment, but we're seeing the opposite effect of building large teams of forward deployed engineers, reskilling workers across different segments of the um industry who have that ability to be AI native. And we're seeing value also sort of pushed down to the lower layer in terms of the chips, in terms of the GPUs and so on and so forth. And these things can change very quickly especially >> well we all remember the deepseek panic >> as a function of exactly you get these kind of disturbing events these events that disturb the status quo a lot of it comes from the research community but some of it also comes from the developer community and and the community at large
>> on the on the question of sovereignty in terms of keeping talent you mentioned Waterloo I think as famous as Waterloo is it's most as famous for being a pipeline to San Francisco. So, is there as is there more that the government can do to like keep that talent here? I mean, Coher is a great example. Um, I'm sure the government would like there to be more. So, what can you do there?
>> Joelle came home. We just >> I am a Wateroo graduate, by the way. [laughter] >> Uh, first of all, you're 100% right. Um, I just want to acknowledge something Joelle said about the candid concerns about jobs, privacy. you know, we when we're building this uh AI, what in Canada we call AI for all to make sure that there's equitable access, we should be open to all these opportunities and triple down on them. But we got to be candid about the concerns. Like we can't be too far ahead. Innovation, I always say, uh, you know, technology moves at the speed of innovation, but citizens move at the speed of trust. So, we've got to get the trust up. uh we have low levels of adoption of AI here in Canada, some of the lowest in the OECD. So part of that is trust related and we should be open about the concerns about sustainability, about water usage, heat usage, power usage, uh job training, security issues. These are very very real and present issues that we should be, you know, you know, like my joke is there's team pompoms and team pitchforks. The pompom teams are like, "Yeah, AI is going to change the world." Fant poms here and then team pitchforks are like stop building. And you know my job and I think a a good entrepreneurs's job is to be sort of a pragmatist. We know this is coming. We've got to get it right and build pragmatically. The big pro one of the problems Canada has always faced is we plant the seeds, we water the seed, we grow the plant and then someone comes along and harvests it. And so we lose a lot of the brain drain. Uh so we have policies here now and I think there's an opportunity with this geopolitical realignment to keep people here because this it's a stable environment from a social cohesion point of view that matters. We're welcoming to talent. We have a $1.6 6 billion talent attraction strategy in our last budget, the largest in the G7 to attract a thousand top talent and their labs, top researchers and their labs to Canada. Huge pickup on that. We also have to make the investment climate and the capital climate and the risk climate um attractive which we have. I mean things in our last budget that now I'm going to make use of my tie. The marginal effective tax rate in Canada, there it is, is lower than the United States even after the one big beautiful bill. That is a way to attract investment here to make sure that you don't have to leave to bring your capital and keep the talent here that you can build in Canada uh and still have access to global markets. more trade deals than any other country in the world, free trade deals, um 20 new trade deals in the last year. So, you can trade with the world here, live a safe life, have an attractive investment climate, and have the talent to innovate. And we hope that is part of the magnet to keep talent here without uh any sacrifice on innovation ability or the ambition to grow globally.
>> And I suppose returning to a similar point, you know, sometimes watching AI from the outside and the announcements from companies and some states, it's just the numbers are eye watering. Tens of billions being invested, uh, gigawatts of compute, power sources, who cares, you know, just, uh, throw up some more gas turbines. How do you compete in a sort of values first way and yet at the same time urgently? And I guess I I would also open that one to cohhere.
>> Look, you know, in many cases, these resources are not always deployed in the most effective possible way. I think you mentioned the deepseek moment before. No one expected it and and yet you know by by all means they they were absolutely able to produce a very high quality model with with relatively modest compute capacity and so there's a lot that can be done. Um, I think what we need more than anything in terms of competing is on the one hand more appetite for risk. You know you go down to California you talk to professors at Stanford every professor has 10 different startups with their students. You come to some of our Canadian universities and you know it's one professor out of 10 who has a startup going and on the go right like that ability to take risk that ability to try and fail. We want to have communities where we have second time third time fourth time founders who are bringing all of their knowledge into the ecosystem.
I think that's also what we need a lot more of. Of course, you know, investments would be very useful, in particular, high-risisk capital. Of course, more compute capacity will be absolutely necessary. But I think there's a lot of things that we can do that matter from a also from a sort of a cultural point of view in terms of wanting to innovate, bringing them out into essentially building new solutions.
>> How is Canada and Coher actually specifically Coher sort of playing smarter and not just, you know, bigger is better, numbers aren't everything. I mean, you know, what's an example of that that you might be able to help help us understand that?
>> I think there's two differentiators. If we talk specifically in terms of like the compute level of investment that where coher is on sort of a different path than some of the the hyperscalers you're seeing. Um, number one is a lot of the models we deploy for enterprise are actually deployed on premise in enterprise. So on their own compute in many cases for security and privacy reasons. So they're deployed completely airgapped to provide very high guarantees, huge benefits, especially for regulated sectors, financial services, healthcare, government services. You want to make sure that that data is captured and isn't going to start traveling all around the world. Um, and so what happens in this case is the customers essentially have to find the capacity in terms of all the inference capacity.
>> So are they investing also co-investing?
>> So coher builds the model, builds the solution. We need the capacity to train those models in some cases to customize the model. But in terms of running them at steady state for tens of thousands of employees and all the costs for tokens that is on the customer's side that it happens. So I think that's one major difference. The other major difference is in some sense I think we've been very smart about how we scale up the models that we train and we've trained models at large scales hundreds of billions of parameters and so on but in many cases what the customers want isn't actually that huge model. What they want is a model that has a much better performance price trade-off. It's nice to run the best model for our models when you're running for just like a pilot a few dozen employees. When you start scaling this up to 30,000 employees, 100,000 employees, giving them access, especially with agents that consume huge amounts of token, what you actually want is a model that's much more efficient for a decent performance and you also want the ability to track that usage and to make sure that you're getting ROI
>> and that proliferates very quickly with the Gentic. And so we've been able to craft models that have a very sweet trade-off in terms of performance and efficiency which makes them really attractive to enterprise customers that are a lot more costconcious
>> on just on how do we compete look we have a role as a federal government and the provinces we have you know billions of dollars invested in our compute challenge so just today's announcement with Telus on their AI factory that we're moving forward on towards theou with them to support um you know their three data centers. There's others there's over 160 proposals for data centers in this uh what we call our enabling sovereign AI data center program. So that's part of it and the reason we're involved is not that there's not a commercial appetite but we want to do it right. We want to do it according to key criteria. We want to enable our own ecosystem to grow. We also have a compute access program to fund and subsidize access to compute and then access to compute if they're using any compute or Canadian comput even higher subsidy. So we we can compete by supporting our ecosystem. Um, so that's part of it. We're building an AI supercomputer. So we put out an RFP on that. It will close in June. that will be a a worldclass top 15 maybe top 10 we'll we'll see because it we'll find out how big it is but it's a billion dollar supercomput that we'll build so uh public access for our research
>> so you've got your own billions announcements but and that's before we launch our strategy
>> which will have other aspects to build
>> we will invest in skills we think we are a big believer in investing in skills training to make sure we have literacy and skills training for workers. That's really important. Um, and the protect model. So, look on the compete, we believe we can compete and scale because we have advantages. But I would also say to all of us here because you're, you know, I was an entrepreneur for 10 years. Um, you're risktakers, you're innovators, you're builders. Like, we need you. This is the moment to build because this building is the response to how we want to control and build the world we want to live in. And you're at this kind of unique moment. But when we talk about competing, we should also talk about competing for the worldview we want and how you build your companies and who you partner with and how you treat the workers and how you treat the data and what kind of companies you build and where you want to build them is a way of competing as well because you you are the builders of this economy and the you're writing the instruction manuals for the future. And when we talk about competing, we've got to compete but not copy. Compete based on the world we want and the world that we want to see. And I don't think we should disagregate building successful companies, worldclass companies, being entrepreneurs, being innovators and disagregate that from leading a kind of values-based world where our democracies are under threat. Many of the communities that we believe should have access don't have the right access. And we got to build urgently, but we got to build right. And and I actually think we're in competition for that as well. And I think that's part of how we compete. And I think governments have a role. I think leaders have a role. And I think all of us kind of have to think about that at, you know, two in the morning when we're worried about getting access to compute. We also should worry about, you know, I I want to build the kind of company that, you know, is seven generations. So it it's it's lasting because the society we're building it in is lasting. Thank you so much. Speaking of urg Well, exactly [applause] speaking of urgently, I think unfortunately we have to wrap it up there. There's loads of things I would have loved to dive more into, especially this talk of future partnerships. So maybe in a year's time we can we can return to that point. So thank you again both very much, Evan Angel.
>> Good luck. Have a great summit. Welcome everybody. Fantastic. Yeah. All right, that's it for opening night, but it isn't the end of making connections at WebSummit Vancouver. Come and get involved and explore Vancouver with Night Summit. We are visiting three different neighborhoods and each night is different. We have meetups, speed networking, games, mini golf tomorrow. Breathtaking views from North Vancouver on Wednesday, but tonight we kick off night summit right now at Bentall Center, just a 5-minute walk from here. So, I hope to see you there. Just all you have to do is just walk up Bard Street and you'll see it. Only badge holders will be admitted. And for the first time, we are having a new event called After Night Summit called After Hours, which will be on Tuesday and Wednesday night and will feature off thereord no holds bars talks with surprise guests that you won't want to miss. Check the Websummit app for all the details, maps, and times. Have fun and see you back here tomorrow morning for day one of WebSummit Vancouver 2026. Good night. Heat. Heat. Heat. Heat.