2024 started on the wrong foot for Israeli super-entrepreneur Prof. Amnon Shashua, who presides over an empire of companies built around artificial intelligence, the main one being Mobileye (Nasdaq: MBLY). Downgraded guidance, layoffs, and an image crisis have hit several of his companies, which, on the face of it, have nothing in common apart from Shashua himself.
The latest news came from OrCam, the smart eyeglasses company founded by Shashua and his former colleague at Mobileye Ziv Aviram. The company has developed eyeglasses that can read text aloud to their wearers, and that have other features to assist people with impaired vision. The company claimed in the past that it was profitable, and it was aiming at a flotation, but the most recent developments by Google and OpenAI, which perform many of the tasks performed by OrCam’s products using a smartphone and for free, rendered the business model of OrCam, which sold its products for over $4,000, obsolete.
A couple of weeks ago it was decided to shut down OrCam’s eyeglasses division, and after laying off over 100 employees, the company is left with a smaller team focusing on hearing products.
Strangling innovation
Shashua built his series of companies on the same principle: development of a black box product based on the last word in artificial intelligence, sold at a reasonable price, and offering high reliability. Mobileye and AI21 Labs, a developer of large language models for enterprise applications, were built on that principle. But efforts by the tech giants such as Nvidia, OpenAI, Qualcomm, and Google, are shaking Shashua’s empire.
"He saw where things were going a long time before the technology giants saw it. While all the major vehicle manufacturers were closing down their artificial intelligence products, he invested in the field and became the biggest," Lucid Capital CEO Shahar Cohen told "Globes". "He’s an experienced professor with a high profile, with the biggest exit and flotation in Israel behind him. But, having spotted the potential, the technology giants are throwing themselves into the game, and their behaviour is a classic example of the way in which Big Tech strangles innovation."
Mobileye’s share price is down 63% this year, after the company cut its guidance for a second time, because of what it described as a lack of visibility in the market and among its customers. In the explanations that Mobileye provides to investors, it does not identify the Big Tech companies as a threat. On the contrary, it perceives itself as the leader in the ADAS (advanced driver assistance systems) market, with huge customers like China’s Zeekr and Germany’s Volkswagen, and it is due to launch many vehicle models with them in the coming years. To its investors, Mobileye describes the competing activity of US semiconductor giants Nvidia and Qualcomm as not providing customers with the required level of reliability.
There’s truth in that. The acquisition of Mobileye by Intel helped it to add radar and LIDAR sensors to its product portfolio, and to supply vehicle manufacturers with a safe driving system. Mobileye’s EyeQ chip is the best-selling brand in the world in the autonomous vehicle boxes market, endowing vehicles with the ability to avoid obstacles, prevent accidents, and carry out some driving tasks autonomously.
Competition with Nvidia
Since Mobileye cut its growth forecast last week, its share price has fallen by about 28%, and its current market cap is around $12 billion, less than the valuation at which it was sold to Intel, and less than the valuation at which it was floated in October 2022, which was $16.7 billion. The company explains the decline by surplus stocks at customers, but it has apparently not absorbed the dimensions of the crisis. In the first quarter it forecast a 9% decline in revenue, and now it forecasts a decline of 15%.
Mobileye has expanded its sales in China to the point where it has created dangerous dependence on that country - 36% of sales in the first quarter. There’s nothing wrong with sales in China. 18% of Apple’s revenue is from there, but the vehicle market there is especially competitive, characterized by government subsidies, and subject to the volatile economic situation in which China finds itself. Mobileye apparently does not sell advanced chips that are liable to be included in the products that the US bans from sale in China, but when China is strengthening its alliance with the anti-American axis and threatening to invade Taiwan, there is a geopolitical risk. It’s not surprising that the decline in Mobileye’s sales is partially explained by the raising of customs duties in the West on Chinese electric vehicles and a delay in the launch in the West of a model of a Chinese manufacturer. Mobileye, however, sees these things as temporary trends.
On the face of it, there is no reason that the current situation shouldn’t be temporary. The Chinese vehicle market is the most innovative in the world today, and the country has become a target market for other chip companies that are sniping at Mobileye. One of them is Nvidia, which has entered the automotive field, partly on the basis of its development center in Yokne’am, and its revenue from that field amounted to almost $1 billion last year, with an orders backlog of $14 billion until 2028.
To tempt the vehicle manufacturers to work with it, Mvidia is offering a series of products designed to provide an end-to-end solution, from a server farm that will support the training of autonomous vehicle models to a safe driving system, cameras and sensors, and a software system that manages all computing aspects of a vehicle. As a highly profitable company, Nvidia can offer very low prices to bring in new customers.
Nvidia’s automotive activity now has many customers, among them customers of Mobileye, such as Zeekr and Polestar in China. Nvidia also has agreements with Mercedes-Benz, Jaguar Land Rover, Hyundai-Kia, and Volvo. Volvo and Jaguar Land Rover will launch models with embedded Nvidia systems from 2026. According to Chinese research institute Gaogong, two companies, Nvidia and China’s own Horizon Robotics, together account for 80% of the luxury vehicles market in China, while the rest is divided between Mobileye, Texas Instruments, and Huawei.
Nvidia is not the only chip company that has entered Mobileye’s market. US company Qualcomm, which became a giant thanks to being the provider of smart telephone chips to Samsung, has won a strategic agreement with BMW. According to its second quarter financials, Qualcomm’s revenue from vehicle chips exceeded $800 million for the first time, almost double that of Mobileye. And while Mobileye’s revenue fell 3% year-on-year, Qualcomm raised its revenue from the segment by 87%.
Nvidia and Qualcomm are benefiting from the current trend among vehicle manufacturers of tailoring hardware and software systems to their needs, something that does not suit Mobileye, which comes along with a ready-made black box. The companies buy chips from the giants, and add off-the-shelf building blocks to them, or else develop some of the software themselves. Many vehicle manufacturers see an advantage in this: they completely control the value chain of the software and hardware components in the vehicle, but that could turn out to be a double-edged sword, since defective programming and misplaced code could lead to breakdowns and recalls. In fact, some of the companies are going back to Mobileye and similar companies that are considered to be reliable brands.
All the same, it’s hard not to wonder about Shashua’s fateful decision to sell the company to Intel, a decision that did bring in a great deal of cash to Mobileye and to his own bank account and those of many of the employees, but is not contributing anything to its competitive positioning against the likes of Nvidia and Qualcomm. Intel plays no part in artificial intelligence, it’s a loss-making company, and it has run out of cash. In fact, Intel sees Mobileye mainly as a financial asset. Mobileye may not be profitable, but it has a lot of cash, and, at least until recently, it has been traded at a much higher multiple than Intel. Intel is now mulling the sale of more shares in Mobileye, and it faces as a dilemma, since that could depress the share price even more.
The adventure that Mobileye has foregone is the operation of a fleet of autonomous taxis, the Robotaxi, and the company now seeks to supply autonomous vehicles to other companies, such as Volkswagen. This is a very expensive and competitive race. No fewer than seven Chinese companies are currently conducting trials of autonomous taxi fleets. Meanwhile, this is a distant dream, and the investment in it magnifies Mobileye’s losses. For the time being, Mobileye is happy to let others - such as Waymo and Tesla - contend with the municipal authorities, the regulators, and users’ complaints, and to learn from them for the future.
Public transport app Moovit, which Mobileye bought shortly after it was acquired by Intel, was meant to become part of the Robotaxi product, and to encourage users of public transport to choose Mobileye’s solution. Now that that dream has been shelved, the company has been forced to recalculate. Moovit co-founder and longstanding CEO Nir Erez has left, and the company recently laid off 25 people. It could be that Moovit is for sale again.
Shashua faces two big challenges that have become the "disease" of the artificial intelligence era. One is the need to invest increasing amounts of capital in developing and producing new chips. Mobileye’s chips currently work with five nanometer technology, and the launch of a three nanometer chip is liable to cost hundreds of millions of dollars. Mobileye believes it can do it at an investment of tens of millions only, since the chip is designed specially for its algorithms.
Another problem is to do with the ability to cope with the competition in artificial intelligence. Mobileye has to use data in at least the same quantities as Nvidia and Qualcomm in order to train its vehicle. Nvidia will carry out complex training based more on artificial intelligence simulations and on data that it can buy, while Mobileye has the data on journeys by millions of vehicles.
Away from global competition, digital bank One Zero, of which Shashua is a shareholder, has encountered difficulties, despite the promise of becoming a bank that makes unique use of artificial intelligence. In the first quarter, the bank revealed losses amounting to nearly NIS 803 million since it was set up in 2019. In addition, a large capital raising round by the bank was cancelled, apparently because of the war, and it had to lay off employees in January as part of a restructuring that will make the Israeli company a subsidiary of an Italian holding company partly owned by Banca Generali Private.
The new model
AI21 Labs, which develops language models for chat bots based on generative artificial intelligence, is also exposed to threats from Big Tech, as a direct competitor of OpenAI, Google, and Meta. The company became known for the language model it developed, Jurassic, based on the main technology used in the development of other models, such as GPT. Recently, however, it launched a new model, developed using a method that has become more popular for context analysis, called Jamba, which is supposed to reduce the rate of errors in language models.
The new model is not rated highly. It does not appear in Hugging Face’s table of the 100 leading chat bot models. That is probably because it’s an unofficial beta version, and the company is confident that the official version that will be launched shortly will find its way into the ranking, but until that happens the company will lose precious momentum.
Like all its competitors, AI21 Labs is not profitable. It competes with the technology giants in the supply of AI-based training models in large companies, generally banks, insurance companies, and infrastructure companies, and it constantly has to raise new capital to finance development. In its latest round, however, it raised only $200 million, which compares with $650 million raised by French company Mistral AI, and $1.3 billion raised by US-based Anthropic. The Israeli company will have to raise hundreds of millions of dollars more in order to keep pace, otherwise it is liable to be swallowed up by a larger company.
"The lifespan of AI models is like that of a dairy product in August," says Shahar Cohen. "A company spends hundreds of millions of dollars to train a model, and after four months along comes a better model that does the same thing for half the price. The Big Tech companies will be the big survivors in this field, because of the huge expenditure on equipment, the ability to secure land and power sources, a user pool that provides feedback to the model, being a talent magnet. The medium-size companies, such as Mistral, Perplexity, and AI21, will be sold as they are, or as a management team, or else their graphics processors will be sold in a liquidation."
Published by Globes, Israel business news - en.globes.co.il - on August 13, 2024.
© Copyright of Globes Publisher Itonut (1983) Ltd., 2024.