SiFive Raises $400M Series G for RISC-V AI Chips

SiFive Locks In $400 Million Series G — RISC-V Is Now the Architecture Wall Street Is Watching

The semiconductor world just got a major shake-up. SiFive, the California-based company that has long championed RISC-V processor architecture as a genuine alternative to Arm and x86, has officially closed a $400 million Series G funding round — and the investor list alone tells you how serious the market is taking this. With names like NVIDIA, Apollo Global Management, and Atreides Management on the cap table, this is not just another startup funding headline. This is AI funding news that signals a structural shift in how the world is thinking about chip design, data center compute, and the future of open-architecture AI infrastructure.

The round was oversubscribed, which in investor language means there was more demand than supply — a strong signal of confidence at a time when the semiconductor funding landscape is fiercely competitive. SiFive's valuation now stands at $3.65 billion, a meaningful jump from the $2.5 billion it commanded back in 2022. For a company that operates on an IP licensing model rather than manufacturing chips itself, this kind of valuation is a reflection of something deeper: the market believes that RISC-V's moment has genuinely arrived.

Who Backed SiFive and Why It Matters

The funding round was led by Atreides Management, with participation from Apollo Global Management, NVIDIA, Point72 Turion, T. Rowe Price Investment Management, Prosperity7 Ventures, and Sutter Hill Ventures. That is a deliberately diverse group of investors — spanning hedge funds, private equity giants, strategic tech players, and long-standing venture backers. The fact that both Apollo and NVIDIA are sitting at the same table in a semiconductor IP deal is itself a story worth examining.

NVIDIA's participation is especially significant when you consider the company's broader strategy. NVIDIA has been steadily building an ecosystem around its accelerators, and integrating RISC-V-based CPUs through SiFive's NVLink Fusion compatibility adds a flexible, open-architecture layer to that ecosystem. Rather than being locked into Arm or x86 for the host CPU role in data centers, NVIDIA's customers would gain the option to use energy-efficient, customizable RISC-V cores that communicate directly with NVIDIA GPUs at high bandwidth. This is not a passive investment — it is strategic positioning in one of the most consequential hardware transitions of the decade.

Apollo's involvement from the private equity side adds a different dimension. Apollo Global Management entering this round suggests that SiFive is being evaluated not just as a growth-stage startup but as a company with durable cash flows and institutional-grade fundamentals on the horizon. When PE money moves into a semiconductor IP company alongside strategic investors, the IPO conversation tends to follow closely — and in SiFive's case, that conversation is already public.

The RISC-V Advantage: Open Architecture in an AI-Driven World

To understand why this AI funding news is resonating across the industry, it helps to understand what makes RISC-V different from the architectures that have dominated computing for decades. The x86 instruction set, owned and controlled by Intel and AMD, has powered personal computers and servers for over 40 years. Arm, on the other hand, has built a licensing empire that spans smartphones, embedded devices, and increasingly, data centers. Both architectures come with constraints — licensing costs, design restrictions, and in Arm's case, a growing concern that the company is beginning to compete with its own customers by developing its own chips.

RISC-V was created as an open, royalty-free alternative. SiFive, which was founded in 2015 by computer architecture researchers Krste Asanović, Yunsup Lee, and Andrew Waterman out of UC Berkeley, became the first company to commercially productize RISC-V processor cores. Its model is straightforward: design high-performance RISC-V processor cores, then license them to semiconductor companies, cloud providers, and chipmakers who want to build their own chips without the constraints of proprietary architectures.

What makes SiFive's latest generation of technology particularly relevant to the AI conversation is its ability to unify scalar, vector, and matrix computing through a single coherent interface. These three types of computation map almost directly onto the needs of modern AI systems. Scalar computing handles regular processing tasks. Vector computing enables parallel operations across large datasets. Matrix computation, especially matrix multiplication, is the mathematical bedrock of transformer models — the architecture behind virtually every major large language model in use today. Getting all three to work efficiently on a single chip design, under a standardized interface, is exactly the kind of engineering challenge that the AI era demands.

NVIDIA NVLink Fusion and the Data Center Play

One of the most technically consequential elements of SiFive's current roadmap is its announced integration with NVIDIA's NVLink Fusion interconnect. NVLink Fusion is NVIDIA's proprietary high-bandwidth fabric that connects CPUs, GPUs, and other accelerators inside data center infrastructure. Traditionally, CPU-to-GPU communication has been handled via PCIe, which is fast but introduces latency and bandwidth ceilings that become increasingly painful as AI workloads scale.

By building NVLink Fusion support directly into its RISC-V IP, SiFive enables data center architects to build tightly coupled systems where RISC-V CPUs and NVIDIA GPUs communicate with the kind of low-latency, high-throughput bandwidth that modern AI training and inference demands. This is not a minor feature. In large-scale AI workloads, the speed at which a CPU can communicate with an accelerator is often as important as the raw performance of either component in isolation.

SiFive has explicitly positioned its RISC-V CPU cores as the preferred control-plane processor for agentic AI systems inside data centers. Agentic AI — systems that can plan, execute multi-step tasks, and interact dynamically with external tools and environments — requires a compute backbone that can orchestrate across multiple accelerators simultaneously while maintaining energy efficiency. SiFive's pitch is that a RISC-V core, purpose-built with this use case in mind, is better suited for this role than legacy CPU architectures that carry decades of backwards-compatibility overhead.

The $400 million in fresh capital will be deployed across three fronts: deepening research and development in scalar, vector, and matrix compute IP; accelerating the software ecosystem for data centers, including compatibility with CUDA, Red Hat Enterprise Linux, and Ubuntu; and expanding customer programs tied to the NVIDIA NVLink Fusion integration. This is a coherent spend plan that addresses the three layers — hardware design, software stack, and ecosystem integration — that any chip IP company must nail to achieve commercial scale.

The IPO Signal: SiFive vs. Arm in the Public Market

Perhaps the most consequential line in the entire funding announcement came from SiFive CEO Patrick Little, who described this Series G as the company's likely final private fundraising round before an IPO. That statement does more than just hint at a public offering — it reframes this entire round as a pre-IPO positioning move. When a company says publicly that its latest raise is its last before going public, it is communicating confidence in its growth trajectory, its financials, and its ability to generate the kind of sustained narrative that public market investors can hold onto.

The IPO, if it proceeds, would be a landmark moment not just for SiFive but for the broader RISC-V ecosystem. Currently, the processor IP market has one major publicly traded benchmark: Arm Holdings, which listed on NASDAQ in September 2023 in one of the most anticipated IPOs of that year. Arm's market capitalization quickly climbed into the tens of billions, reflecting investor appetite for semiconductor IP businesses with recurring licensing revenue and deep design-win pipelines.

SiFive's eventual listing would be the first time that a RISC-V IP company trades on a public exchange, creating a direct, market-driven comparison with Arm. This is significant because it would force institutional investors to make explicit choices between two fundamentally different architectural philosophies: Arm's established proprietary model with its vast installed base versus SiFive's open, customizable model with its growing appeal among cloud providers, hyperscalers, and sovereign chip programs looking to reduce dependency on single-vendor architectures.

What gives SiFive an edge in this narrative is the timing. Arm has recently moved into chip design itself — a strategic expansion that puts it in direct competition with many of the same customers who license its architecture. That conflict-of-interest dynamic is increasingly uncomfortable for companies that have built product roadmaps around Arm IP. SiFive, by contrast, operates purely as an IP licensor with no ambition to become a chip manufacturer or systems builder. For customers who want a high-performance processor IP partner with no competitive agenda, that positioning is genuinely differentiated.

What This Round Means for the Broader AI Funding Landscape

Zooming out, this round fits into a broader pattern of AI funding news that has defined the first half of 2026. Capital is flowing not just into software AI companies building models and applications, but increasingly into the hardware and infrastructure layer that makes AI possible at scale. Semiconductor IP, chip design tools, data center interconnects, and silicon for agentic AI are all attracting significant institutional attention.

SiFive's $400 million raise is one of the most significant hardware-layer AI funding events of the year because it represents a bet on architectural independence. The entire global AI compute stack is currently heavily concentrated — a handful of GPU architectures, a small number of CPU designs, and limited interconnect options. SiFive's success would add meaningful diversification to that stack, giving the industry more levers to pull as AI workloads grow in complexity, scale, and diversity.

For the AI World Organisation, this development is a reminder that the AI funding ecosystem is increasingly a hardware story as much as it is a software one. The companies building the physical compute infrastructure — the cores, the interconnects, the memory hierarchies — are receiving the kind of capital and strategic attention that was once reserved almost exclusively for model builders and application developers. SiFive's Series G round, at $3.65 billion valuation with NVIDIA and Apollo at the table and an IPO on the horizon, is one of the clearest signals yet that the semiconductor layer of the AI economy is entering a new phase of maturity and market validation.