SQLite Just Beat MySQL by 4.9x, And Hosts Are Noticing

Look at the average small business hosting plan for 2026 and you’ll soon notice something missing. There’s a PHP runtime, an NVMe drive, an Nginx configuration, and a certbot script. But there’s no MySQL server, the “M” in the LAMP stack that defined web hosting for two decades.

Instead, more and more, there’s a single .db file sitting next to the application. No daemon, no port 3306 and no tuning of a connection pool at 2 a.m. when something locks up. The industry is beginning to call this “edge-native.” Engineers who’ve been doing it for a year already just call it SQLite.

And new 2026 benchmarks confirm what engineers have been reporting anecdotally. On sites with fewer than 10,000 concurrent connections — basically every small business site on the planet — read operations with SQLite are 4.9 times faster than with a comparable MySQL or MariaDB setup on the same VPS. The only difference between the two configurations is where the database lives.

It turns out that where the database lives accounts for most of the cost. The way LAMP worked required the application server to make a TCP connection to a separate database server, usually on a different host or as a separate container on the same host. Data came back, got processed and triggered another connection. That overhead compounded fast for pages joining multiple tables.

SQLite eliminates the trip. The database is a file. The application reads the file directly, with bytes going straight from the SSD to application memory.

This is an old idea that has now been made practical. SQLite has been available for more than 20 years. What changed is everything else. NVMe-class storage became cheap enough that even budget VPS plans ship with it as standard. PHP 8 and modern Node runtimes got fast enough that the application is no longer the bottleneck for small business hosting.

And SQLite’s engineers quietly added every feature the database had been missing over the past five years: full JSON support, write-ahead logging that handles concurrent writes without the database locking up, partial indexes, generated columns and even a vector search extension for AI-native applications.

The “toy database” reputation never quite disappeared from the public eye. But behind the scenes, the project had spent 20 years preparing for this moment.

For Tier-2 and Tier-3 providers selling shared and entry-level VPS plans to small businesses, the shift to SQLite has two opposing effects.

The first effect is entirely positive. Server resource consumption drops substantially, the set of failure conditions that would trigger a support ticket shrinks, and by far the most common cause of “my site is down” tickets disappears with the database connection.

The second effect is less desirable. The chance to sell a separately managed MySQL or PostgreSQL service for an additional fee — often anywhere from tens to a few hundred dollars per month — is one of the most profitable lines on the pricing pages of a small host.

If small business customers no longer need a separate database server, this revenue line goes away, and the upgrade path from shared hosting to “shared with database” to VPS to “VPS with managed database” is compressed.

Providers reducing the emphasis on the managed database tier of their pricing pages are doing the math correctly. The ones still leading with it will have a significantly harder conversation with customers in 12 months.

The shorthand for this new philosophy: local-first data storage, file-based databases and single-machine application units easily moved between regions or providers by copying a directory.

The components aren’t new. NVMe drives, file-based databases and single application runtimes are all well-established technologies. What’s new is putting them together as the default rather than as a fallback.

A small business hosting plan using SQLite for the database, NVMe for storage and a single application runtime delivers page load times that would have required a load balancer, two application servers, a managed database cluster and a Redis instance five years ago.

As a hosting provider, the question is whether your current product page reflects this. Plans that lead with “free MySQL database” as a primary selling point read as if they were written in 2018. By contrast, plans that lead with “NVMe storage” and “local-first runtime” read like they were written for the customers placing orders this year.

A specific class of company is almost certainly not going to pivot toward this. Managed database revenues are a major component of the pricing strategies of each of the hyperscalers (AWS, Azure and Google Cloud). Products such as RDS, Aurora, Cloud SQL and Cosmos DB aren’t technologies that benefit from encouraging customers to put a SQLite file on local storage.

So the migration to SQLite is, in a real sense, a wedge for smaller providers. A Tier-2 provider that offers an edge-optimized plan based on SQLite delivers a price-performance ratio an AWS reseller can’t match because the reseller is paying AWS for a database tier the SQLite plan doesn’t need.

Will small business customers perceive these advantages? Probably not. Most won’t care about the architecture. They’ll care about the speed and the price. And on both, the math currently favors the file.

Conditions over the next 12 to 18 months — maybe sooner than that — will probably look like this: Shared hosting plans stop providing MySQL by default. The strategy for selling managed database services shifts toward a niche market of customers who specifically require it. New low-cost VPS offerings start advertising “SQLite-ready” as a feature. And the “M” in LAMP becomes optional.

The acronym never quite recovers. Eventually, someone comes up with a new one to describe the resulting stack. LASP doesn’t quite sing. Neither does LANP.

Whatever it ultimately becomes, the database server — the workhorse of every small business web hosting plan for the last two decades — may be headed for obsolescence.