$400 Million and a $3.65 Billion Vote of Confidence

SiFive, the chip design company founded in 2015 by the engineers who created RISC-V at UC Berkeley, has just closed a $400 million funding round. The round was oversubscribed — meaning investors wanted to put in more money than SiFive was raising. That kind of demand signals serious confidence in the company's direction.

The funding values SiFive at $3.65 billion, up from $2.33 billion in 2022. That is a significant jump in valuation, particularly in an environment where many tech companies have seen their valuations stagnate or decline.

Among the investors: Nvidia. And that detail changes everything about how you should read this story.

What Is RISC-V and Why Does It Matter?

To understand why SiFive is significant, you need to understand what RISC-V is — and why it is fundamentally different from the chips that power most of the world's devices today.

RISC-V is an open-source chip architecture. That means anyone can use it, build on it, and customise it without paying a licence fee to a central company. It was created at UC Berkeley as an academic project, designed to be simple, modular, and freely available.

Compare that to the alternatives. Intel's x86 architecture is proprietary — if you want to build an x86-compatible chip, you need Intel's permission and payment. Arm Holdings' architecture is also licenced — every chip in your phone that runs on Arm technology involves royalty payments to Arm.

RISC-V breaks this model entirely. There is no gatekeeper. There is no licence fee. There is no single company that can be pressured, sanctioned, or geopolitically leveraged to cut off access. It is genuinely open, in the same way that Linux is open.

SiFive does not manufacture chips. It designs them — high-quality, production-ready RISC-V chip designs that other companies can licence and build into their products. Think of it as the professional services layer on top of the open-source foundation.

Why Is Nvidia Investing in a Competitor's Architecture?

At first glance, Nvidia investing in SiFive seems strange. Nvidia makes GPUs. SiFive designs CPU architectures based on RISC-V. They are not direct competitors, but they operate in related spaces.

The answer lies in understanding what Nvidia is actually building — and it is not just chips.

Nvidia is building an ecosystem.

CUDA, Nvidia's software platform, is already the industry standard for AI computing. NVLink Fusion, Nvidia's interconnect technology for AI server infrastructure, ties together chips from multiple manufacturers into unified high-performance systems. Nvidia is positioning itself not as just a chip company, but as the platform on which AI gets built and run.

If RISC-V chips — potentially designed by SiFive — become the standard CPU in AI data centre, and those CPUs are running in servers connected by NVLink Fusion and programmed using CUDA, then Nvidia wins regardless of who made the CPU. The ecosystem is what matters, not any individual component.

This is the chip industry's version of a platform war. And Nvidia is playing to win the platform, not just the product.

From Microcontrollers to AI Data Centre

RISC-V started its commercial life in embedded systems — the small, low-power chips inside smart appliances, industrial sensors, and basic electronics. It was respected but niche.

That is changing rapidly.

SiFive has been steadily moving RISC-V into higher-performance computing. The company is now targeting AI CPUs — the processors that sit alongside GPUs in AI training and inference infrastructure. This is a dramatically more demanding application than running a smart thermostat, and it requires chip designs of significantly greater complexity and performance.

The fact that SiFive's funding round was oversubscribed suggests that serious institutional investors believe RISC-V is ready for this transition. The AI data centre boom — driven by demand for large language models, enterprise AI, and inference at scale — needs infrastructure. SiFive is positioning itself to supply part of that infrastructure on open, customizable terms.

The Business Model: Licences, Not Factories

SiFive's business model is similar to what Arm Holdings built in its early years — design the chip architecture, licence the designs to manufacturers, collect royalties. It is an asset-light model that does not require building or running expensive semiconductor fabrication plants.

The contrast with where Arm is heading now is instructive. Arm Holdings has recently been moving toward actually manufacturing chips — including a reported partnership with Meta on custom AI chips. That is a fundamentally different bet: moving from designing the blueprint to building the building.

SiFive is staying in the design layer for now. That keeps it lean, flexible, and accessible to a wide range of customers who want to build their own custom silicon without designing from scratch.

In an era where Google, Amazon, Microsoft, and Meta are all designing their own custom AI chips — Google's TPUs, Amazon's Trainium, Meta's AI accelerators — there is growing demand for companies that can help organisations build custom silicon efficiently without starting from zero.

Geopolitics: Why Open Chip Standards Are Strategically Important

There is a dimension to this story that goes beyond business strategy, and it involves the escalating tension between the United States and China over semiconductor technology.

The US has placed significant export restrictions on advanced chips sold to China — limiting access to Nvidia's most powerful AI chips and constraining China's ability to build cutting-edge AI infrastructure. In response, China has been aggressively investing in domestic chip development.

RISC-V sits in an interesting position here. Because it is open-source and not owned by any US company, it is arguably not subject to US export controls in the same way that proprietary architectures are. China has been investing heavily in RISC-V development as a result — building chips based on an architecture that no American government can sanction or restrict.

This makes RISC-V geopolitically neutral in a way that Intel and Arm are not. For countries and companies that want chip independence — freedom from both US and Chinese technology dependencies — RISC-V offers a credible path. SiFive, as one of the leading commercial players in the RISC-V ecosystem, sits directly in that opportunity.

The Future: Open Hardware Meets Proprietary Software

The most likely future for AI infrastructure is not a clean winner-take-all scenario. It is a hybrid.

Open hardware — such as RISC-V chips designed by companies like SiFive — will coexist with proprietary software platforms like CUDA. The hardware layer becomes commoditised and customizable. The software and ecosystem layer remains a source of competitive advantage.

This is exactly the model that made Android successful in mobile — open at the OS level, with proprietary services and ecosystems built on top. It is the model that made Linux dominant in servers. And it may be the model that defines AI infrastructure for the next decade.

Nvidia understands this. Its investment in SiFive is not a bet against its own chips. It is a bet on its own ecosystem — and a smart one.

What This Means for the AI Industry

The SiFive funding round is a signal worth paying attention to, even if chip architecture is not usually headline news.

It tells us that the AI hardware market is maturing and diversifying. It tells us that open standards are becoming commercially viable at the highest performance levels. It tells us that Nvidia's strategy extends far beyond building faster GPUs. And it tells us that the next phase of the AI boom will be defined not just by model capabilities, but by who controls the infrastructure those models run on.

Key Takeaway

The chip war is becoming a platform war. Intel and AMD are competing chip by chip. Nvidia is building the ecosystem that ties everything together — and its bet on SiFive suggests it plans to win at the infrastructure level, not just the GPU level.