Rivos made headlines in 2022 after Apple filed a trade secrets lawsuit against the corporate, accusing Rivos of hiring dozens of Apple engineers and using confidential information to develop chips just like those of the iPhone manufacturer could compete.
The company denied the allegations and sued Apple for unfair competition. Apple is finished settle his lawsuit in February. Around the identical time, a separate lawsuit involving several Apple engineers hired by Rivos ended.
Now that the courtroom drama is behind it, Rivos is redoubling its efforts to bring its chipset technology to market, CEO Puneet Kumar told TechCrunch.
“Rivos was founded with a mission to construct industry-leading, high-performance, energy-efficient chips,” said Kumar. “We are excited to interact customers who’re developing data-driven solutions.”
A major recent tranche of funding will help fund these efforts.
On Tuesday, Rivos announced that it had raised over $250 million in an oversubscribed, expanded Series A led by Matrix Capital Management with participation from chip giants including Intel (through its corporate VC division) and MediaTek. Other backers included Cambium Capital, Hotung Venture Group, Walden Catalyst, Dell Technologies Capital and Koch Disruptive Technologies.
That's quite a turnaround for Rivos, which was founded in 2021 and struggled to boost money from investors and recruit employees a few 12 months ago within the shadow of the Apple lawsuit. In August, Rivos laid off nearly two dozen employeesor 6% of its then-current workforce, and was forced to postpone a planned $400 billion Series A funding round, The Information reported on the time.
A custom server chip
Rivos' long-term goal, Kumar said, is to construct chips primarily for servers that may handle intensive data evaluation and AI workloads, including generative AI workloads.
“We goal customers who’re constructing data-driven solutions (e.g. people who use generative AI and data analytics to drive decisions),” said Kumar. “There are many firms targeting such markets. Streams supports the high hardware requirements of the AI models and analytics that may reshape the enterprise.”
Rivos' first chipset relies on RISC-V, the open standard instruction set architecture (ISA).
ISAs are technical specifications underlying every chip that describe how software controls the chip's hardware. For general computing applications, chip design teams typically license an existing ISA from a longtime vendor (e.g., Arm or Intel). But RISC-V represents an open, royalty-free alternative.
Rivos' chip features what Kumar calls a “data parallel accelerator” to speed up AI and large data-related calculations, essentially a GPU designed for purposes beyond graphics processing. It was manufactured using TSMC's 3nm manufacturing process. In chip manufacturing, “process” refers to the dimensions of the smallest component that may be embedded on a chip.
This 3 nm is taken into account near the innovative. While they include Qualcomm, MediaTek, Nvidia and AMD, amongst others expected to make use of TSMC's process for his or her upcoming chip families, Apple was the one company to make use of it in its M3 chipset series in 2024.
In addition to constructing the chip, Rivos is working on standalone data center hardware based on the Open Compute Project modular standard, which effectively serves as a plug-and-play chip package. And a “firmware-to-app” software stack is being created for programming the chip, Kumar said.
“Customer workloads may be easily deployed on our more efficient hardware while still leveraging their existing models and databases, giving them immediate value,” Kumar added.
Rivos, which currently has no revenue, plans to earn a living by charging its customers – primarily large data center operators – for its hardware and complementary software solutions. David Goel, an early investor, said Rivos' “low-friction” launch pipeline was a key differentiator within the highly competitive chip market.
“The Rivos team has cleverly integrated the groundbreaking recent RISC-V architecture with an revolutionary accelerator, effectively bringing this vision to life,” Goel told TechCrunch. “Their prototype chip is a compelling demonstration of their unique capabilities.”
But is it differentiating enough?
Tough competition
Big tech firms – considered one of Rivos' potential customer segments – are scrambling to develop their very own chips for AI and large data analytics because the generative AI boom continues.
Google is on the method to its fifth generation TPU and recently introduced Axion, its first dedicated chip for running models. Amazon has several custom chip families in the marketplace. Microsoft jumped into the fray last 12 months with the Azure Maia AI Accelerator and the Azure Cobalt 100 CPU. And Meta continues step-by-step along with his own designs.
There at the moment are dozens of startups on the lookout for a share of the custom data center chip market That could reach $10 billion this 12 months and double by 2025.
Groq, an organization that develops chips to run AI models faster than traditional hardware, recently launched a brand new business unit focused on enterprise applications and use cases. The AI hardware startup Tenstorrent, led by engineer Jim Keller, wants to put in its chipsets in data centers. And Riotsa South Korean Fables The AI chip company has raised a whole lot of thousands and thousands of dollars in capital to ramp up production of its data center-focused chip Atom.
But Nvidia, currently the dominant force in chips, is proving difficult to topple.
Nvidia briefly became a $2 trillion company this 12 months as demand for its GPUs for AI training soared. Wells Fargo Equity Research Estimates that Nvidia has a 98% market share in data center GPUs, and was the corporate's data center business increased by greater than 400% within the fourth quarter of 2023 as Nvidia builds a brand new unit to design custom chips for cloud computing firms and others.
In view of the tough competition – and the deterrent effect Nvidia's dominance has also impacted funding for potential competitors – making it difficult for some newcomers to the custom server chip space.
Just a few months ago, Graphcore, whose value had reportedly fallen by $1 billion after a take care of Microsoft collapsed, said it was planning to chop jobs on account of the “extremely difficult” macroeconomic environment. Habana Labs, the Intel-owned AI chip company, laid off an estimated 10% of its workforce last 12 months. Also last 12 months, SiFive — like Rivos, a RISC-V startup — laid off 20% of its workforce and discontinued its core product line.
So will Rivos get well? Perhaps.
Kumar wouldn't speak about customers and Rivos' chip isn't expected to enter mass production until sometime next 12 months. But with 375 employees and a whole lot of thousands and thousands of dollars within the bank, Rivos is well positioned to expand production and double down on platform and software development, Kumar said.
“Due to the rapid changes in generative AI and the fusion with the information analytics stack, it’s critical that accelerators are easy to program and debug and that data may be transferred seamlessly between CPU and accelerator,” said Kumar. “Rivos addresses this need through our 'recompile, not refactor' approach.”
