'We're winning meaningful customers': FuriosaAI CEO on competing with NVIDIA in the AI era
“We are measured always against the NVIDIA product these days. To win the customer, you need to show your solution in a very comprehensive way,” says June Paik, Co-Founder and CEO of FuriosaAI.
FuriosaAI recently established a new Singapore subsidiary, which will serve as a central hub for business development, sales and technical support across the Asia-Pacific region as the company scales commercial deployment of its RNGD AI inference platform.
The South Korean AI semiconductor startup has already made headlines globally with its RNGD gaining popularity among customers for its capability. The company's flagship 5nm second-generation data center inference accelerator, RNGD is designed to run large language models and agentic AI efficiently within conventional air-cooled power envelopes.
The expansion into Singapore will see FuriosaAI expand customer and partner engagement across Southeast Asia and the broader APAC market, collaborating with enterprises, governments, cloud providers and infrastructure partners to accelerate enterprise-scale AI inference deployments.
CRN Asia caught up with June Paik, Co-Founder and CEO of FuriosaAI at Tech Week Singapore to understand more about the company’s growth plans, opportunities in the industry and also the competition in the market.
“It's quite an exciting time for us because we started our first batch of volume production in January. And we are successfully ramping up production now, even despite all the constant supply chain challenges. In some sense, we are proud of this volume production milestone. And at the same time, our products are right now getting more validated and deployed through the enterprise customers along the way,” Paik said.
Commercial deployments of RNGD globally have picked up following production. Paik explained that RNGD is currently deployed by enterprise customers including Samsung SDS and LG AI Research, with the company rapidly expanding customer and partner engagements across Asia, Europe, the Middle East and the United States.
“We are definitely expanding globally. That's an even more exciting thing. So, we set up our entity here in Singapore. We are building a strong team here on both the engineering side and design to engage with customers here. I see such a dynamic right now, especially with infrastructure activities going on in this region. And we are very identifying some good opportunities and partnerships. And also, Singapore is one of the leading countries where companies are integrating AI into their fabric right now. So, we are engaging with some of the customers here as well now,” he added.
The Singapore office will be led by Addison Chi, Vice President of APAC Sales, who will oversee regional team expansion across commercial operations and technical support. The new entity adds to Furiosa’s global footprint, which includes its Seoul headquarters, and offices in Silicon Valley and Lisbon.
Global expansion
Looking at the opportunities globally, Paik pointed out that one of the biggest drivers for their growth in the rising token consumption with organizations spending on AI inference is in an uncontrollable way.
“They all start to think about this inference deficiency. It's not just scaling their infra and spending. I mean, even the biggest company cannot control the cost right now. They will go bankrupt. So, I think we see quite a good amount of attraction from U.S. customers as well. Because we are providing the compute solution, which can make this token cheaper and have more efficient inference,” Paik explained.
“Every supply component is rising in cost, and we are sometimes using those components. But what we promise is that we will try to maintain the win-win margin here with the customer and us. We can sustain our business and also enable customers to sustain their business, rather than some company enjoying the margin of 80%.”
Paik also highlighted that energy efficiency is a fundamental breakthrough innovation.
“Energy also turns out to be a fundamental cost of AI. So, if you consume more energy, you've got to build your power capacity, supply, cooling, everything. It's not just the electricity cost. Everything associated with the AI data center relates to power. So, I think fundamentally, making it energy efficient will kind of lower the cost of AI as well,” he explained.
Competing with NVIDIA
When asked how FuriosaAI deals with the competition in the industry, especially with NVIDIA dominating the market, Paik acknowledged the NVIDIA’s solution is actually a great product as it's a complete solution, with a chip and a s software stack in the whole ecosystem.
“We are measured always against the NVIDIA product these days. To win the customer, you need to show your solution in a very comprehensive way. It's not just enough that you show some parts of the things are working. Because customers try to run their entire agentic workflow on your chip,” Paik said.
“It's why a lot of AI chip startup companies struggle. They cannot meet the entire requirements. But now we have some confidence. Our product is right now ready enough in all layers. We need to deliver to the customer, which is the key. If you look at the other AI chips, they are not winning any customers, but we are already winning some meaningful customers. And the design wins are even accelerating right now. We are not just serving one customer. We are serving dozens of customers right now,” he added.
For Paik, this also means that FuriosaAI’s software stack and solutions are much more ready and robust.
“So that's how we stand out - our product. The total product's readiness from the production to up to the software stack is where we are differentiating from other startup solutions, which still remains more in proof of concepts,” he added.
At the same time, Paik also mentioned that being an independent merchant silicon company is always a very challenging thing to do.
“Because it's a very volume business, it's very capital intensive. So, to run your business, you got to be able to build a volume business with your chips. A lot of companies on the way usually got acquired, consolidated and such. But we are on the right path.”
“The market is such a big opportunity. And I mean, NVIDIA is a great company. AMD is a great company. Google is a great company. But they cannot serve all the demands. All enterprises need tokens these days. So, I think we can grow our own kind of market and business. It's a very rare opportunity. We can do so because the market is there right now. So, I think we will have a greater impact by being independent, in my opinion,” Paik said.
“We would like to have a much greater impact as being a kind of company who can build a volume chip and who can provide this compute to the enterprise company,” he concluded.