# The Infrastructure You Choose Is the Company You Become > Source: > Published: 2026-07-07 14:32:04+00:00 ## Key Takeaways - Choosing a database for scale is a ten-year decision that caps how far you can grow. - As capabilities converge, buyers choose on taste and judgment. - The primary user is now the agent, not the human. Hence [one agent, one database](https://www.pingcap.com/blog/database-consolidation-for-ai-agents/). - The audacious requests are where the next product directions start. - What you believe determines what you [design](https://www.pingcap.com/tidb/). ## Believing in Growth: A Hard Question From a Distributed Systems Veteran A few years ago, I was talking with the chief architect of a very well-known global company that offers a home sharing marketplace. He opened with a genuinely tough question. “Siddon, I have spent decades building distributed systems. Google Bigtable, for one. I have probably eaten more salt in this field than you have eaten rice. So tell me: why should I choose TiDB?” That stopped me for a moment. Reciting how advanced [TiDB’s architecture](https://docs.pingcap.com/tidb/stable/tidb-architecture/) is, how fast it runs, or how well it scales would not land with a world-class distributed systems expert. He already knew all of that, probably better than I did. So I offered a point of view that felt a little grandiose even to me at the time. “Choosing a database for scale means choosing your database infrastructure for the next ten years. More precisely, you are choosing how far you want this company to grow.” He clearly did not expect that answer. He smiled, then lit up, and told me how they built Bigtable back in the day, and how it carried Google through its years of explosive growth. Then we got into a serious discussion about TiDB’s architecture and its tradeoffs. In the end, the company chose [TiDB](https://www.pingcap.com/what-is-tidb/). It resolved a series of problems they had with HBase, Aurora, and other systems, and supported the rapid growth of their business, especially the surge after the pandemic. The part I find most interesting is what came next. When they design new systems now, they actively think about how to get the most out of what TiDB can do. ### Choosing a Database for Scale Let me be clear about one thing. This customer did not choose TiDB because of one conversation. Selecting core infrastructure is like buying a house or getting married. No one signs off on it because of a single inspirational line. That conversation did crystallize an idea that has grown clearer ever since. The infrastructure you choose is the company you become. For TiDB, the expression is direct: Choosing a database for scale means betting that your company will grow 10x, even 100x. The same story has played out many times: - A payment company in Japan that crashed during its earliest promotional campaigns grew into one of the country’s top payment platforms. - An e-commerce company in India that could not survive its own version of Black Friday now runs bigger sales every year, and has become the largest e-commerce platform in the country. - A photo-based social platform in the United States that once buckled under advertising traffic now helps its merchants generate 600 million dollars in annual revenue. Which is why I believe this line more and more. Choosing TiDB means believing in growth. ## Having Great Taste: A Bracelet, and What It Signaled The first time we had dinner with the CTO of a leading large language model company in China, our CTO Ed Huang had a set of bracelets custom-made to match the company’s aesthetic. That gesture stunned them, and moved them. The bracelets were not worth much money. What they conveyed was taste: An attention to beauty and to detail. At one point I spoke with an engineer at that company who was born after 2000. Every generation produces its own talent, and this young engineer was already responsible for critically important work. He told me something I have not forgotten. “If what you build cannot make people under 25 think it is cool, you are already obsolete.” Hearing that, I felt the deep melancholy of age. I also felt a certain pride, because TiDB is being adopted and endorsed by exactly this generation of engineers. They are not using it out of nostalgia. They use it because they think it is genuinely cool, and because it matches their vision of what future systems should look like. Again, no deal was closed over a dinner and a bracelet. Choosing a database for scale is not like ordering bubble tea. You cannot just say “30 percent sugar, less ice” and check out. ### Shared Judgment About Databases Built for Agents One important reason this company chose TiDB is that they share our judgment about where databases are heading. TiDB is no longer just a database for humans. It is becoming a [database for AI agents](https://www.pingcap.com/ai/agentic-ai/). Their own product logic is equally direct: Let the agent decide for itself whether to use a database, and how. Treating the [AI agent as a first-class citizen](https://www.pingcap.com/blog/rethinking-scale-tidb-evolution-ai-agent-database/) is a conviction we hold in common. I also visited the business lead of another prospective customer. Over dinner, he told me, very directly: “Siddon, we are not your customer yet, but I will do everything I can to bring you in. A lot of what I know about distributed systems, I learned from the articles you publish online. Your latest [TiDB X architecture](https://www.pingcap.com/press-release/pingcap-launches-tidb-x-new-ai-capabilities/) in particular, building a transactional system on object storage, is genuinely impressive. I trust your technical taste, and I believe that taste can carry us to our next stage.” He paused, then added: “Also, could I get a few of your company T-shirts? They are really well made.” Everyone at the table laughed. We have joked that if TiDB ever stops selling, we could open a merchandise store instead. T-shirts, caps, mugs, backpacks, pens, mouse pads. The margins might beat the database business. ### Choosing a Database for Scale: When Capabilities Converge, Taste Decides Behind the joke sits a serious observation. Databases have matured to the point where, in many scenarios, the capability gap between vendors has narrowed. I would go further: The performance of many database systems today is more than most workloads will ever require. So when customers choose a database, beyond raw capability, they increasingly weigh deeper and longer-term questions. - What is your aesthetic? - What is your technical taste? - What is your judgment about where systems are heading? - Are you a team that can keep producing great work? Hence the second line. Choosing TiDB means having great taste. ## Building Beyond Imagination: An Absurd Request From a Customer Building With AI Almost everyone building software today is anxious, developers most of all, because AI is eating the software. With AI, anyone can build and ship software at remarkable speed. The barrier to software development has collapsed. Yet the software that breaks through and earns wide adoption is still rare. When everyone can write software, it is hard to build a world-class product without a slightly crazy idea, and without the nerve to change something. So when a customer came to us with a request, we were stunned. “Can you sell us one million databases, very cheaply?” Honestly, my first reaction was: Is this a requirement, or a stress-test script? The reason turned out to be simple. This customer is doing something audacious. ### One Agent, One Database Building a website is one of the most common needs on the internet. Whether you want a personal blog, a product homepage, or a SaaS system serving real users, you need a site, and an entire industry of website builders exists to serve that need. Today, generating static pages is trivial for an agent, so many companies now offer static site hosting. The problem sits one layer down. The database at the core of the site is still something users have to buy, configure, and operate themselves. That model works, but for a large share of users the barrier remains too high. So this customer set out to build a fully managed site-building platform: Not just hosting the pages, but hosting and operating the user’s database as well. The idea is audacious, and as TiDB’s builders, that is exactly the kind of audacity we like. Every audacious attempt carries real possibility. What looks absurd today can become the breakthrough for an important direction tomorrow. So we said yes. That request is where a new principle of ours took shape: [One agent, one database](https://www.pingcap.com/blog/database-consolidation-for-ai-agents/). It is what pushed us to treat the agent as a true first-class citizen of TiDB. ### The Next Audacious Question We visited this customer again recently. In their original mental model, these sites belonged to individuals and small businesses, with limited business criticality. They have since discovered that large customers are now using these generated sites to deliver serious production services. AI-generated sites are already going straight to the enterprise, far beyond what they first imagined. Teams building at this scale increasingly rely on [database branching and per-agent isolation](https://www.pingcap.com/case-study/manus-agentic-ai-database-tidb/) to run thousands of environments without chaos. Even better, once TiDB solved their near-million-database problem, they found the next requirement. Agents do not just need database access. They need cloud storage for all kinds of unstructured data, which led to the next question. “Can you sell us one million cloud drives, very cheaply?” Well. Audacious enough. Challenging enough. The answer was simple. Yes. Let us get to work. Hence the third line: Choosing TiDB means building beyond imagination. ## A Closing Thought: Beliefs, Not Specifications This blog is a checkpoint in my thinking after more than a decade of building infrastructure software, and TiDB in particular. It is not a typical technical article. It is closer to a set of philosophical notes. I am increasingly convinced that infrastructure software has never been a mere collection of technical specifications. Underneath it sits a deep system of beliefs, and what you believe determines what you design. If you believe the business will grow, you do not design only for today’s scale. If you believe engineers should have taste, you do not settle for features that merely run. If you believe the future is full of uncertainty, you do not build a system that serves only one scenario. Many of TiDB’s design decisions trace back to these beliefs. - We believe in growth, so we committed to horizontal scaling, elastic capacity, and cloud-native architecture. - We believe in taste, so we keep refining the developer experience, the agent experience, and how the product expresses itself. - We believe in unlimited possibility, so we join customers in pursuits that sound a little crazy at the start. ### Choosing a Database for Scale: Design for the Future You Believe In Great systems rarely grow out of a tidy requirements document. They tend to start with a simple question. If the future really will be 100 times bigger than today, how should we design for it now? Which brings me back to the thing I believe more with each passing year. The infrastructure you choose is the company you become. Choosing TiDB, in a real sense, means choosing three things: Believing in growth, having great taste, and building beyond imagination. *Curious what this looks like in practice? Explore how teams build on **TiDB for agentic AI**, or read how **Manus scaled its agent data layer** on TiDB.* Experience modern data infrastructure firsthand. ## TiDB Cloud Dedicated A fully-managed cloud DBaaS for predictable workloads ## TiDB Cloud Starter A fully-managed cloud DBaaS for auto-scaling workloads