SambaNova Systems Inc. reckons it’s the best-funded artificial intelligence systems and service platform startup in the world after landing a massive $676 million late-stage round of funding.

Today’s Series D round was led by the SoftBank Vision Fund 2, together with participation from new investors Temasek and GIC, plus existing backers including funds and accounts managed by BlackRock, Intel Capital, GV (formerly Google Ventures), Walden International and WRVI.

SambaNova had already raised close to a half-billion dollars, including a $250 million Series C round in February 2020. The company said today’s cash infusion brings its total funding amount to more than $1 billion, placing its value at more than $5 billion.

Founded in 2017, SambaNova has designed and built an integrated hardware and software platform that it says can run AI and data-intensive applications from the data center to the edge. By integrating both hardware and software, the company’s chips offer reconfigurable “dataflow” architecture that enables applications to drive optimized hardware configurations. In addition, software is said to be no longer confined by the constraints of fixed hardware.

The company announced general availability of its DataScale hardware platform in December. It uses custom seven-nanometer chips that the company says are more attuned to machine learning and deep learning processes than the general-purpose microprocessors and graphics processing units widely used for such tasks today.

The platform’s Reconfigurable Dataflow Architecture runs an integrated open-source software stack called SambaFlow that helps run each machine learning model optimally. It does that by minimizing the need to interface with memory, eliminating the main bottleneck in AI, which is the interconnect between chips and memory.

Companies can buy the SambaNova Systems DataScale hardware or rent it through the company’s Dataflow-as-a-Service offering. Through that, SambaNova installs and maintains the system on the customer’s site and charges only for usage, with prices starting at $10,000 per month. The offering is designed to jumpstart enterprise-level AI initiatives by augmenting a company’s existing AI capabilities and allowing them to focus purely on their business objectives rather than the infrastructure.

SambaNova co-founder and Chief Executive Rodrigo Liang said his company is aiming to “revolutionize the AI market” and today’s funding will accelerate that mission.

“Traditional CPU and GPU architectures have reached their computational limits,” Liang said. “To truly unleash AI’s potential to solve humanity’s greatest technology challenges, a new approach is needed. We’ve figured out that approach, and it’s exciting to see a wealth of prudent investors validate that.”

Analyst Holger Mueller of Constellation Research Inc. said AI needs a new generation of hardware, networking and software and that SambaNova’s funding round today shows that the industry is well on the way to reaching that.

“Startups that have a great pedigree through their founders have no problem picking up new funding,” Mueller said. “That’s the case with SambaNova, which is raising its market capitalization to five unicorns today. Clearly, SoftBank founder Masayoshi Son is confident that the team of former Sun and Oracle engineers will be able to deliver.”

One of SambaNova’s main tasks going forward will be to try and make its AI hardware especially, stand out amid a wide field of competing startups and more established players in the space, said Patrick Moorhead of Moor Insights & Strategy.

“SambaNova is one of a handful of companies, along with Groq, Graphcore, Cerberus, Intel and Qualcomm, trying to challenge Nvidia in machine learning inference,” Moorhead said. “Its biggest issues are the crowded space and differentiation.”

Strangely, for a company that has attracted so much investor capital, SambaNova Systems is rather secretive about who is actually using its platform. Its customers are entirely unknown, though VentureBeat reported last year that it’s said to be working on specialized “software-defined devices” inspired by research from the U.S. Defense Advanced Research Projects Agency.

Photo: SambaNova