Supercomputing Startup Conduit Launches From Stealth
What if there was a way to utilize unused or excessive computing power from all computers and put it to use? What would happen? Well, if this startup gets it right, Amazon Web Services would be obsolete.
Cambridge-based Conduit claims that it is building the first collective computer that can supply computing power faster than Microsoft Azure and Amazon Web Services can. As it emerges out of stealth mode, the startup founded by MIT graduates will use the collective grid computing solution derived from different institutions across the country starting with its alma mater MIT. What does this mean? Simply put, it directs supply sources — unused or excess computing power from dormant computers and towards demand sources. If this sounds like an Airbnb for computing — it is.
This can be done with essentially any computer but the startup wants to harness the power of computer clusters at universities.
In-depth news coverage of the Greater Boston Area.
“At first, we had the idea to do this with anyone who has a computer on the conduit network but we are starting with universities,” said Taylor Caforio, chief business officer at Conduit. “Universities have supercomputing power that is just sitting there.”
For now, Conduit is bridging the demand-supply gap between excessive computing power in universities to pharma and biotech companies who need it for running drug discovery simulations through a process called parallelization.
It works like this: Parallelization means dividing up a computational task to such granularity that it can be done by multiple computers at the same time. “Think of it like cutting a sandwich in two and giving it to a friend so you eat the sandwich in half the time,” Caforio explained. “Drug design is an extraordinarily complex process, computationally speaking, and takes a long time.” Parallelization helps by providing a very secure platform as it makes data granular and parsed out.
Conduit is going after clients in pharma and biotech because when it comes to drug discovery, time is a crucial element. “Companies running drug discovery simulations spend twice as much time as they need to,” Caforio said. “For such processes, the earlier you fail, the better.”
The computing power that the Conduit network can provide is up to 100 petaflops. For scale, an iPhone operates around 1-10 gigaflops. So that makes 100 petaflops equal to 100 million iPhones. Time wise, computing power of that scale can turn a job that takes 10 hours in approximately less than 10 minutes.
To start, the two-year-old startup will use excess computing power from MIT’s Lincoln Lab and pay the university for it. Going forward, the startup’s business model will be to pay the people who supply computing power.
Founded in 2017 by MIT mechatronics and information theory graduate Ryan Robinson, Conduit wants to compete with behemoths like Amazon web services and Microsoft Azure. Currently, it has a staff of eight full-time employees.
The startup is currently working with mid-sized pharma companies in the Boston area but in the future plans to add clients in fintech, supply chain logistics and quantum computing.