The archive · Developer & Business Tools · Product decision · 2011
MemSQL raises $2.1M to bet in-memory SQL can replace disk databases
YC Winter 2011's MemSQL raised a $2.1M seed from SV Angel, First Round and Ashton Kutcher on the bet that MySQL-compatible SQL in RAM beats databases on disk.
MemSQL
What the business is
MemSQL builds an in-memory relational database: it keeps data in RAM for faster reads, writes and realtime analytics while speaking the MySQL protocol, so applications already talking to MySQL can switch without code changes.
Starting capital：$2.1 million seed round announced on 12 July 2011.
How it started
MemSQL was a Y Combinator Winter 2011 company founded by Eric Frenkiel (formerly Facebook) and Nikita Shamgunov (six years as a senior database engineer at Microsoft), surrounded by engineers from Facebook, Oracle and SQL Server backgrounds. Frenkiel's thesis was that datasets at major web companies were doubling every 18 months, cloud I/O was sluggish, and memory had become cheap and abundant enough for large data sets to live in RAM — Google, Facebook, Twitter and Zynga already kept their datasets in memory backed up on disk. He sized the online transaction processing software market as growing from $20 billion to $60 billion over five years.
What happened
MemSQL raised $2.1 million in seed funding, announced on 12 July 2011, from First Round Capital, New Enterprise Associates, Start Fund, SV Angel and Y Combinator, joined by Ashton Kutcher, Paul Buchheit, Sam Altman, Guy Oseary and others. The six-person team planned to use the money to complete a move into new offices in downtown San Francisco and to hire. MemSQL's first customer was fellow YC Winter 2011 startup LikeALittle, which had hit 20 million pageviews within six weeks of launch and, per TechCrunch, relied on MemSQL's technology to keep performing through the influx of traffic.
How it ended up
By 12 July 2011 the seed round was closed, MemSQL was running a private beta with LikeALittle as a live reference customer, and the six-person company was hiring as it moved offices — the in-memory-SQL bet had passed its first commercial test.
Background
MemSQL is an in-memory database startup that came out of Y Combinator's Winter 2011 batch. Co-founders Eric Frenkiel and Nikita Shamgunov came from Facebook and Microsoft's SQL Server engineering respectively, and their core argument was simple: web-scale datasets were doubling every 18 months while disk and cloud I/O stayed slow, so the database belonged in memory.
The product bet was deliberately conservative on compatibility. Instead of forcing users into NoSQL or a caching layer, MemSQL was built from the ground up to speak the MySQL protocol while keeping SQL and ACID semantics, claiming to run up to 30 times faster than relational databases on disk — with no code change required for applications that already talked to MySQL.
On 12 July 2011 the six-person company announced a $2.1 million seed round from First Round Capital, NEA, Start Fund, SV Angel and Y Combinator, with Ashton Kutcher, Paul Buchheit, Sam Altman and Guy Oseary participating. The money was earmarked for new downtown San Francisco offices and hiring.
MemSQL's first customer was fellow YC Winter 2011 company LikeALittle, which had reached 20 million pageviews within six weeks of launch and was reportedly being valued at $35 million; TechCrunch credited MemSQL's technology with helping it scale through that traffic. A fast-growing peer paying to run on the engine was the seed of the sales story.
What has to be true
- Frenkiel and Shamgunov attacked the incumbents on their own protocol: keeping MySQL compatibility meant zero application rewrites, removing the switching cost that kills most infrastructure products.
- The memory-cost trend gave the bet a clear clock — RAM had become cheap enough to hold live datasets, and web giants already kept data in memory backed up on disk.
- The team's database pedigree made a from-scratch engine credible to investors: a Microsoft SQL Server engineer with a PhD and patents alongside former Facebook and Oracle engineers.
- A live reference customer, LikeALittle at 20 million pageviews in six weeks, showed the performance claim holding up under real traffic rather than on benchmarks.
What can be applied
A performance product wins on compatibility, not speed alone: MemSQL sold 30x speed by promising MySQL apps zero code changes, removing the migration cost that blocks most infrastructure switches.
Aftermath
As of 12 July 2011 MemSQL was a six-person startup in private beta, planning to use its $2.1 million seed to finish moving into downtown San Francisco offices and to hire. The company was inviting users to its homepage for private-beta access, and its public proof point was LikeALittle scaling through heavy traffic on MemSQL technology. The verdict on the broader in-memory-SQL bet was still ahead.
Sources
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card