- Startups Databricks, Opaque, and Anyscale raised massive amounts of funding this past year.
- Intel also reportedly offered to acquire the chip startup SiFive.
- And they all emerged from UC Berkeley labs, which focus on innovative open source software.
- See more stories on Insider’s business page.
This past year has marked the monster success of four open source startups.
In July, the AI startup Databricks announced $1.6 billion in funding at a $38 billion valuation, and the data security startup Opaque announced $9.5 million in seed funding. Intel offered to acquire the chip startup SiFive for more than $2 billion, Bloomberg reported in June. And last October, the distributed computing startup Anyscale announced a $40 million Series B round.
These companies all started at University of California Berkeley’s historic computer science research labs.
These labs have an unusual tradition in that they only stay open for five years.
Founded in 2017, RISELab is the newest of this series, giving students just five years to solve a specific problem within fields like AI using entirely open source technology.
The self-imposed time limit clearly works.
The labs also garner millions in funding from tech companies like Amazon, Google, Intel, Facebook, and Microsoft.
“It is the combination of the talent in the labs, and the format, and the track record of the labs that keeps on continuing,” said Raluca Ada Popa, RISELab and Opaque cofounder.
A profitable way to open source
These labs show how open source research can translate into commercial success.
The startups release their software as open source, meaning it is available to anyone to use for free and modify. This allows their software to gain massive followings as developers and even big firms like banks can easily get their hands on it.
For instance, Databricks emerged from the open source project Apache Spark, a data analytics tool built in Berkeley’s AMPLab. Anyscale is based on Ray, an open source project from RISELab that distributes computing power across multiple computers.
The five-year format primes students for success because it discourages complacency, says Robert Nishihara, CEO and cofounder of Anyscale.
“The benefit of the five-year timeline is that it pushes researchers to reevaluate what are the most important problems to solve and adjust the direction they are moving in every five years,” Nishihara said.
“People there are ambitious – yet the lab is very creative and multidisciplinary,” Nishihara added. “If you have a culture that’s only focused on getting immediate results, it will not give you the space to tackle big problems.”
An unusual collaboration
The labs have an unusual collaboration with the tech industry, too. Researchers visit companies like Facebook, Google, and Twitter to identify what problems they have and then work to solve them.
“Everything we get from the industry is gift money,” said Ion Stoica, RISELab director and cofounder of Databricks, Anyscale, and Opaque. “There is no obligation from us to deliver something. What they are actually looking for is to see and be a part of the next trend.”
Some lab professors move from the labs to the startups and back, which is also rare in academia. Most PhDs “go straight into teaching and research,” said Alfred Chuang, general partner at Race Capital. “You rarely see people like Ion that go back and forth.”
Students move from the lab into the startups, too.
For example, Nishihara and cofounder Philipp Moritz teamed up with their advisor Stoica to found Anyscale. SiFive cofounders Asanovic, Yunsup Lee, and Andrew Waterman developed open source chip hardware that would form the basis of that company.
“When I was a student, I just remember how fun it was because we would get all these people coming through from all the leading companies,” Asanovic said
Venture investors also keep an eye on projects emerging from these labs, looking to buy into the next hot startup as soon as they are launched, says Chuang, who participated in Opaque’s most recent funding.
RISELab is about to come to a close as it hits the five-year mark, and two new labs will launch: one focused on cloud computing and new concepts like “sky computing,” and the other on programming languages.
As Databricks’ Ghodsi describes the experience: “There’s a sense that you can change the world and you can change anything and do whatever you want.”