This is a facet of Nvidia that we know less about. But little by little, the world leader in graphics chips for training large artificial intelligence models is accelerating in the automobile industry. Cameras, radars, GPS: with so many sensors, the autonomous cars of the future will generate a mass of data that will require computing power, which is precisely what Nvidia provides.
To capture this market, Nvidia is not content with being an industrial partner of manufacturers. The American has just invested for the first time in Wayve, a British start-up of 290 people, created in 2017, and which develops AI models for autonomous cars, as part of a record fundraising . The start-up, based in London and Silicon Valley, announced on Tuesday that it had raised $1.05 billion in Series C, an amount never seen before for an AI start-up in Europe.
This round of funding was led by SoftBank, which is taking a stake in Wayve and also joining the board of directors. Microsoft, which had already invested $200 million at the beginning of 2022 during the previous round, also put in the pot again. The fundraising was even welcomed by Rishi Sunak, the British Prime Minister, for whom Wayve “consolidates the United Kingdom’s status as an AI superpower”.
“Create a form of exclusivity”
Nvidia is not entering the automobile industry. In fiscal year 2024, the group achieved $1.1 billion in revenue (or 2% of the total) with this industry, by selling its Nvidia DRIVE computing platform to various manufacturers including Volvo, Mercedes Benz and Hyundai. “The automobile industry has enormous computing needs,” confirms Pedro Moreno Lahore, in charge of Nvidia’s automotive offering for France and Southern Europe. In autonomous cars, certain calculations cannot be done in the cloud due to their criticality. This is where Nvidia comes in, with a hardware brick in the cars and software layers on top.
Wayve was also a customer of Nvidia. “When we arrive at such a level of partnership, it makes sense to enter into the capital. Nvidia is also creating a form of exclusivity in the medium term. They prevent a competitor from getting their hands on key technologies,” explains Sébastien Amichi, partner at Kearney in charge of automobiles.
Because Intel and Qualcomm, the two other chip giants, have also positioned themselves in the autonomous car niche: the first by buying the Israeli Mobileye in 2017, specializing in intelligent sensors, and the second by taking over Veoneer, the specialist of driving assistance, in 2021. “Everything will come down to the three of them” predicts Sébastien Amichi.
“The OpenAI of the car”
For its part, Wayve will be able to recruit and purchase computing capacities. And above all transform its fundamental research into an industrial product for automobile manufacturers. The company positions itself as “the OpenAI of the car”: it has developed several large AI models that allow autonomous cars to learn and interact with their environments, particularly during a sudden event (for example when ‘a pedestrian suddenly crosses the street).
For Wayve, AI must be integrated into the physical world (in robots or cars for example) to feed on surrounding information. However today, large AI models, like those of OpenAI or Meta, are fed almost exclusively with text from the Web. Problem: text being too weak a resource, current large models cannot reason or make decisions. This is what hampers the development of 100% autonomous cars (level 5).
With its AI Driver solution, Wayve potentially allows vehicles to reach level 3 (high autonomy) and 4 (near full autonomy). “Automotive is an interesting use case for AI. This makes it possible to understand the man-machine, machine-environment and machine-network interface and to replicate the lessons elsewhere,” explains Sébastien Amichi, from Kearney. But Wayve has not yet sold its solution to customers. After this record fundraising, the challenge of monetization awaits.
This article is originally published on .lesechos.fr
