Ship 26 NYC - Think Bigger: From Prototype to Global Scale with Vercel and AWS
VVercel
Computing/SoftwareInternet Technology
Transcript
00:00:00Hi everyone. It is so great to be here at SHIP today. Thank you for joining us in New York City
00:00:09in the middle of all this crazy World Cup action. I suggest no one leave this building for a couple
00:00:15hours because it's traffic out there. Luckily, everything lines up nicely. I want to really
00:00:20thank Guillermo and his team for hosting AWS today. We have such a great partnership between
00:00:25AWS and Vercel. It's an exciting time to be a builder. We are going through one of the biggest shifts of
00:00:34our lifetime and it's really changing what's possible for everyone in this room. For most
00:00:40in community history, the hard part was everything between the idea and a working product. It took
00:00:46specialized skills, months of setup, and the budget and time to be wrong a few times along the way.
00:00:54AI has been removing all those constraints. It's enabling us to build at an incredible unprecedented
00:01:01pace, which means for the first time you can build almost anything that you can imagine. A lot of ideas
00:01:09start small. A prototype for a specific pain point, an improvement to a process, but the right idea
00:01:16that doesn't stay small. It catches and it reaches escape velocity and it becomes the product everyone
00:01:23is using. The one that no one saw coming. The thing that changes how industries work. The distance between
00:01:30I had an idea and I changed the world in some meaningful way has never been shorter. And it's time for all of
00:01:36us to take advantage of that. The most transformative inventions didn't come from people playing it safe.
00:01:42They came from people chasing ideas that felt too ambitious, too complex, or too early. And now AI has
00:01:50helped us catch up to that kind of ambition. That's why AWS and Vercel are working together to take the
00:01:56friction out of building your biggest ideas from the first prototype all the way to getting to global scale.
00:02:03Now, we know Vercel gives builders a fast path from an idea to something deployed and live. A developer
00:02:11experience designed so you spend your time building and not configuring. When you ship a change, you see
00:02:17it instantly. That's what lets the team move at the speed of its ideas instead of the speed of its
00:02:23infrastructure. For the value of AWS, the speed only matters if you can build what holds up when it
00:02:29works. That's why AWS matters. When your idea finds real usage, lots of traffic, real data, security
00:02:37compliance requirements, you need what AWS can bring. Global scale, resiliency, enterprise-grade security,
00:02:46and, of course, the deepest and broadest set of cloud and AI capabilities. Whatever your app needs as you
00:02:53scale, it's there. And you don't have to choose between the two. Vercel runs on AWS. It uses more than
00:03:0230 AWS services under the hood, including AWS Lambda and Amazon Bedrock, deployed across 20 of our global
00:03:09regions. So you start with Vercel's ease and speed on day one, and as your app scales, you
00:03:16can adopt more and more of AWS's depth as needed without re-platforming. That's the path we're going to
00:03:24follow today. How you build, how you add production capabilities, and how you operate at scale using
00:03:30Vercel and AWS together. Let's start with how you build. For some of you, building starts with VZERO,
00:03:40Vercel's AI assistant for design and iteration. You describe what you want, and VZERO gives you a fully
00:03:46working, full-stack application in minutes. That's incredible for speed. But speed by itself has a
00:03:53catch. For anyone who's lived with an AI-generated app for a few months, you know there are some
00:03:57challenges. Inconsistent patterns can emerge, there can be duplicated components, there's often thin test
00:04:03coverage, and if you're not careful, you face architectural drift that can make change more
00:04:08difficult as your app scales out. That's where Curo comes in. Curo is AWS's spec-driven agentic IDE
00:04:16built on Amazon Bedrock. Instead of jumping straight to code, it works spec first. It turns your prompt
00:04:23into requirements, user stories, acceptance criteria, technical design, even a task list. And then Curo writes the code,
00:04:31and tests against those original stories. While VZERO gets you building fast, iterating on an idea and
00:04:38testing a functional app with a focus group of users, with Curo, you can build things that are
00:04:44so durable they can really scale. We see teams combining VZERO and Curo in a bunch of ways.
00:04:50The most common is VZERO to Curo. You prototype fast in VZERO, and then you bring it into Curo to generate
00:04:58those requirements, planning the new features, refactoring, testing, and you can manage any
00:05:03technical debt that shows up, which you don't want. The reverse also works. Curo to VZERO. When requirements
00:05:09are complex, you define those specs and architecture in Curo first, and then you hand them to VZERO to
00:05:16generate the app. And of course, many teams are running both in parallel. VZERO is the UI factory for the
00:05:22frontends, design systems, and dashboards, and Curo is the engineering operating system for the backend,
00:05:29the APIs, the data models, and your test strategy. With both, you can build fast, and you can build right.
00:05:36Now, I know you all love the frontend development VZERO is so well known for, but as you heard Guillermo
00:05:43talk about, VZERO is getting into full stack development in a big way. Now you can deploy
00:05:48backends on VZERO like FastAPI, Flask, Express, and others. This morning we ran a joint AWS and VZERO
00:05:56workshop for many of you, where we connected straight to Amazon Aurora using VZERO's AI Gateway
00:06:01and Amazon Bedrock for AI inference, and shipped a full stack app live, frontend, backend, and database,
00:06:09all in a single workflow. Check out this QR code, you can get a recap of the workshop. It'll take you
00:06:14through all of what we did and how you can do it yourself. Now, I want to root this in a real
00:06:19customer example. One of those is the Weather Company. They've built full stacks apps at scale. The Weather
00:06:26Company serves real-time forecasts to over 350 million monthly active users, and they're backed by more
00:06:34than 100 meteorologists who calculate 2 billion coordinates every 15 minutes. That's truly scale.
00:06:41The Weather Company runs on AWS, and using Vercel, they rebuilt their entire web serving stack
00:06:47and their content managing system, something that previously would have taken months. They did,
00:06:52or taken years, they did in just months. That change meant that a web page can go from design to being
00:06:59published in a few hours, an 80% increase in velocity.
00:07:07So, you're used to V0 in Kiro. You've used V0 in Kiro to build your prototype. Now, let's talk about
00:07:14production capabilities. There are three powerful AWS capabilities available in Vercel to take your app
00:07:20further. I've already referenced a bit about AWS databases, there's Amazon OpenSearch, and access to
00:07:26the broadest choice of models through Amazon Bedrock. Let's go a little deeper in databases. Every app,
00:07:33we know, needs a way to manage data. And this is the first place where prototype to production gaps
00:07:39can be the most difficult. The lightweight data layer that's perfect while you're prototyping and
00:07:44scaling early users is the first thing to buckle under heavy use. When users show up, you have concurrent
00:07:50connections climbing, your working set can outgrow memory, and read traffic can really pile into a
00:07:56handful of hot rows. The product is fine, but the data tier becomes the most likely to fall over,
00:08:02and standing up a reliable production-grade database is unfamiliar and slow to many people.
00:08:08To address this, we worked with Vercel to make AWS databases native in the Vercel marketplace.
00:08:14Aurora PostgreSQL, Aurora DSQL, and DynamoDB. These are the same database services that run mission-critical
00:08:22workloads at some of the largest companies in the world, including, of course, Amazon.com,
00:08:26and they're available to you instantly. You provision them right from the Vercel dashboard,
00:08:32and VZero can create and connect AWS accounts and databases to your projects as it builds.
00:08:38There's no manual IM setup, no passwords and environment variables, authentication uses short-lived
00:08:44credentials, and those details are injected right into your project automatically. And Vercel tested this,
00:08:51and by a wide margin, Aurora is the fastest database available compared to competing options,
00:08:57because there's no network hop between your code and your data. You get under a millisecond of latency,
00:09:02twice as fast as the next best option. I want to show you an example of using VZero to build an app
00:09:08on an AWS database. Let's say I want to build a food delivery app. Traditionally, I'd create that app,
00:09:15provision a database, configure credentials, wire the services together, lots of separate steps. Instead,
00:09:21with Vercel, I just described what I want. Build me a food delivery app with restaurant listings,
00:09:26menus, a shopping cart, order tracking. VZero starts generating, and it recognizes the app needs a
00:09:32database. So right inside the flow, it recommends Aurora Postgres SQL. With one click, I've got a
00:09:38production-grade Aurora database, even with $100 in AWS credits if you're new to the platform. And Vercel
00:09:44handles the rest, connecting that database and wiring it all together. If I ever want to take a look at
00:09:51that config, it's just a few clicks away, in Vercel, I can even jump into my AWS management console without
00:09:58having to log in separately. It's just that easy. And builders are already running with this. We just
00:10:05wrapped up a joint hackathon with Vercel. More than 8,000 builders took part, competing for $160,000 in
00:10:12prizes and AWS credits, going from a prompt to a deployed application on the infrastructure I just
00:10:17showed you. These same data services are the ones that startups and enterprises run in production.
00:10:24Now let's talk about search. Increasingly, that means retrieval for AI. And here's the challenge. An AI
00:10:31feature is only good as what data it can get to. To give a model accurate current answers, you have to
00:10:37ground it in your own content. And that means running real search infrastructure, things like vector
00:10:43indexes and embeddings that hold up under unpredictable loads. Standing that up yourself usually means
00:10:49stitching together several systems and often paying for capacity you don't end up using. That's why we
00:10:56worked with Vercel to add Amazon OpenSearch serverless to the Vercel marketplace. Once again, provisioned in
00:11:02just a click. It has unified support for vector, lexical, hybrid, and agentic search. It's the ideal
00:11:08foundation for RAG, grounding your AI in your own content. And it auto scales up to 20 times faster,
00:11:15but maybe more importantly, it can scale to zero when not used. So you pay nothing when idle and can cut
00:11:21your search costs by up to 60% because you're not paying for peak capacity, you're paying for real
00:11:27usage. The result is a retrieval layer behind production-grade AI features without dealing with
00:11:33the operational burden of running it yourself. The third capability I want to touch on is access to AI
00:11:40models. No single model wins at everything. Accuracy, speed, cost, and context length all trade off against
00:11:48each other. And the right pick for the job is somewhere in front of you, but you have to know what it is.
00:11:54Having those right models on hand and being able to switch as your needs change matters a lot.
00:12:00That's where Amazon Bedrock comes in. Bedrock is a model provider in the Vercel AI gateway. So by
00:12:06configuring Bedrock you get one endpoint to over a hundred foundation models, including the latest
00:12:10Frontier models from Anthropic and OpenAI. And you can swap between them without rewriting your app.
00:12:16Instead of wiring up each model provider one by one with separate accounts, separate keys, separate
00:12:21policies, everything routes through Bedrock. It enables that security and governance, every call runs
00:12:27under your own controls and logging, and Bedrock guardrails can apply consistent policy checks across every
00:12:32model. And because it all runs through Vercel's AI gateway, it's one endpoint, no separate accounts to
00:12:38manage and usage tracking built right in. We've touched on how you build and how you add production
00:12:46capabilities. Let's look at how you operate at enterprise scale. And for a lot of builders operating
00:12:51in enterprise, it very painfully starts with how you buy. Today, enterprise users can procure Vercel
00:12:58directly through the AWS marketplace, which means it runs through the procurement and billing you already
00:13:03have set up, and it can apply Vercel workloads against your existing AWS commitments. Vercel's been
00:13:09an AWS partner since 2023, and we've seen more than 300 enterprise customers use this path to adopt
00:13:16Vercel quickly. For larger enterprises, the details of where your workloads run, which account owns them, how they
00:13:24reach your private internal systems. And if you have to deal with auditors and regulators, tons of really
00:13:29complex things come up. And these concerns have all been addressed when you have an existing cloud
00:13:34provider agreement. The tricky part is how do you get the benefits of Vercel in a way that deals with all
00:13:39those concerns? With Vercel's new bring your own cloud capability, it's now possible to run Vercel functions
00:13:46inside your AWS account. Your compute, your build artifacts, your data, all run on your AWS instance,
00:13:54and Vercel runs the rest. Your apps, your agents reach private backends the same way anything else in
00:14:00your account can, with your own CI and pipeline. It's a win-win. Your engineers keep the Vercel developer
00:14:06experience they love, while your platform security and compliance teams keep what they need. The network
00:14:11controls, full IAM, audit evidence, data residency, if they're already consolidated the AWS accounts with the
00:14:17accounts, with the agreements you have in place. Bring your own cloud is in private beta. So if you're interested,
00:14:23reach out to your account team, and they can take you through details. So before I leave you, I wanted to give
00:14:28you one more example. Fanatics. They're a great example of combining the power of Vercel and AWS. Fanatics does
00:14:38so many things for fans, and one of them is pretty fun. Making and selling trading cards.
00:14:44That part of the business is built around connecting the physical and digital collector experience,
00:14:49and when they drop new cards, there are massive tracking spikes. Staying fast and available in the
00:14:55moment is everything to them. To make it happen, they use Vercel and AWS together. Vercel's serverless
00:15:03deployment running on AWS used Lambda, EC2, and S3. And their teams, both engineers and non-engineers
00:15:10alike, used V0 to turbocharge new front-end experiences so they can constantly wow those picky collectors
00:15:17with the exciting next drop. I started today by saying all of us have the potential to build something that
00:15:24changes the world. And I know that at first, those ideas don't always feel safe or obvious or even possible.
00:15:31That's exactly the moment to push through. Don't accept the limitations of today as the boundaries of tomorrow.
00:15:38Our goal at AWS has always been to remove every barrier standing between your imagination and reality.
00:15:44That's why we build. That's why we keep pushing more services, more capabilities, more ways to help you
00:15:50move from idea to impact. And that's why we've been deepening our collaboration with Vercel. So when it
00:15:55feels like you have an idea that's too big, the technology is never the thing that holds you back.
00:16:02So here's my parting challenge. Think even bigger than you think is possible. Be audacious, be bold,
00:16:08and build with Vercel and AWS. That next great invention that changes the world might just come from you.
00:16:14Thank you. Have a great ship. See you soon.