Vsora는 추론에 중점을 두는 European AI 칩 스타트업으로서, 자신을 'AI 추론의 European Nvidia'로 자리매김한다.

Khaled Maalej, CEO of Vsora, attributes the French inference chip maker's success to premium memory, advanced packaging through close ties with Taiwan, and a distinctly European product free from American supply constraints.

In conversation with SDxCentral, Maalej claimed to be the only European alternative to Nvidia and AMD—and arguably "the one big player when it comes to inference."

Vsora's flagship Jotunn8 stands out as a rare European chip leveraging HBM3E (High Bandwidth Memory 3 Extended). "We are probably the only company in Europe having access to HBM3E today," Maalej said. The high-bandwidth memory is packaged alongside the AI accelerator die using chip-on-wafer-on-substrate (CoWoS), Taiwan Semiconductor Manufacturing Company Limited's (TSMC) advanced packaging technology that bundles multiple chips tightly together.

Vsora's close relationship with TSMC stems from Dibcom, a French digital-TV receiver chip specialist that Maalej co-founded. After selling Dibcom to Parrot SA in 2011, Maalej co-founded Vsora in 2015, maintaining Dibcom's legacy by focusing initially on digital signal processing IP for 5G baseband and wireless processing.

"The team was known to execute very well with Dibcom. This opened paths to our partners, convincing TSMC and the HBM vendors today to give you access to the technology. I think it's a very limited club today in the world having access to this," Maalej explained.

**From Dibcom to Vsora**

Vsora has since moved from IP to selling its own silicon. The Jotunn8 offers 288 gigabytes of memory and 3.2 petaflop processing power, designed to overcome the memory wall—the performance gap between fast processors and relatively slow memory that constrains AI inference efficiency. The goal is to enable AI to run at significantly lower infrastructure costs and energy consumption.

"Back then, we didn't even know that it was called memory wall," Maalej recalled. "It took us like two years to put the technology in place, and we started addressing the 5G market, which required the highest processing power, and we started as an IP company. And very quickly people understood a little bit of our architecture, told us we need to look after AI."

Vsora approached major German carmakers to evaluate its technology for autonomous driving and found their architecture delivered superior processing efficiency and lower latency than competitors. This sparked a multi-year level-4 autonomous driving collaboration, during which Vsora's design scaled from 100 teraflops to 1 petaflop by early 2020—roughly 60% above NVIDIA's A100 chip's performance.

This success prompted a shift to a full fabless semiconductor model, developing silicon in partnership with TSMC. By 2022, the team discovered that about 80% of AI token cost derives from the chip itself, meaning the only realistic way to lower cost per token is to increase silicon throughput.

"The market is realizing today that the big market is going to be inference. Today 50% of the processing power in the world of data centers is probably still used for training, 50% for inference. We see that rate going to 90% inference," Maalej said. "But still, this growth will not happen if we don't actually reduce the cost per token."

**Vsora and Sovereignty**

Headquartered in Meudon-La-Forêt, Vsora now employs around 70 staff across Asia and the Middle East. The firm plans to launch a next-generation chip later this year and is collaborating with partners on servers and racks to accelerate time-to-market and address system-integration challenges.

"We remain a silicon vendor. We are not willing to sell systems," Maalej stressed.

Following a recent, undisclosed fundraising round, Vsora aims to expand into the U.S. and has hired Sandra Rivera as its new chair; she previously led Intel's AI business.

Maalej emphasized Vsora's sovereign value proposition, noting growing global appetite for "non-U.S. silicon," with data center operators in Europe and elsewhere seeking at least 10–20% of their chips from non-American sources to reduce dependency and strengthen sovereignty.

"Europe will have to have its own silicon. I don't think that we can completely rely on silicon coming from the U.S. because you never know what can happen, and AI is becoming fundamental. I mean, you can see it as like having water or electricity or something equally fundamental," Maalej said. "Where we were standing six months ago is completely different from today. Europe is trying to put in place programs and these kinds of things. But we need to run faster."