Horizon CEO: Ecological win-win is a clear and effective path to success in intelligence!

At the 13th China Automotive Blue Book Forum 2021 hosted by Automotive Business Review, Yu Kai, founder and CEO of Horizon Robotics, a leader in edge AI chips, delivered a speech entitled "Empowering the new generation of automotive intelligent chips and jointly building an open software ecosystem for intelligent vehicles". He shared many wonderful views on topics such as the development trend and strategic path of the intelligent automotive industry, solutions to chip shortages, and Horizon Robotics' response strategies.

The theme of this Automotive Blue Book Forum is "First Move". Whoever can more deeply understand the essence of the new automotive revolution is considered to be more likely to take the first move and achieve success. As for what is "new", Yu Kai believes that the first is a new concept, that is, what is regarded as a strategic goal and what is regarded as a tactical goal. This is also the key to distinguishing between new and old car-making forces. In the new automotive era, user value is regarded as the core keyword of innovation and competition.

The "chip shortage" in the global automotive industry has intensified since the end of last year due to limited production capacity, insufficient supply and skyrocketing prices.

The global chip shortage has caused many popular car models to be "hard to find". Although the global chip industry's production capacity has tilted towards automobiles, the chip shortage has caused a cumulative global suspension of automobile production of 2.99 million vehicles, resulting in losses of more than $110 billion. In order to minimize losses, in addition to giving priority to high-end and hot-selling models to ensure profits, and cutting some non-core functions to save chips, automakers' senior executives have "squatted in chip factories" and "scanned globally" to start a chip grab war. However, despite the frequent use of strange tricks, automakers still cannot fundamentally solve the current situation of chip shortages in the short term.

"The ultimate solution to the global chip shortage is to advance the automotive electronic architecture from a distributed ECU architecture to a domain architecture, and then to a central computing architecture," said Yu Kai.

Three Paths to Intelligent Strategy

In 2021, the automotive industry will enter the era of car manufacturing 2.0. The trend of cross-border smart car manufacturing in various industries such as mobile phones, real estate, and home appliances is becoming stronger. Intelligence has become the second half or even the main field of automotive technological innovation, but this road lacks mature experience to learn from. Every player is thinking about how to move towards the future. Yu Kai believes that China's independent brands face three paths in their strategic transformation to intelligence, including relying on full-stack intelligent giants, ecological win-win, and vertical self-research.

Which of these three paths is better? Yu Kai believes that relying on full-stack intelligent giants can help car companies save time, effort and money at the beginning, but this kind of cooperation with closed technology solution suppliers is a buyer-seller relationship, which will make car companies lose the dominant position in technology updates and user experience improvements, and over time they will become "helpless mediocrity"; while the vertical self-research model represented by Tesla and Apple is exciting but full of risks. This model is "unreplicable excellence."

Yu Kai believes that for most car companies, ecological win-win is the "ruled path to success". "In the past two years, the architecture of neural network algorithms has changed every year. If you don't embrace an open ecosystem, it is easy for you to make a sudden mistake in a certain link and it will be irreversible." China has advanced into an arena for world-class leading automotive smart chips. As an open computing platform service provider, Horizon will work with partners to build an open and win-win industrial ecosystem.

The ultimate solution to chip shortages

Since the end of 2020, the global automotive industry's "chip shortage" has intensified. According to the latest data from AutoForecast Solutions, the cumulative number of global automobile production stoppages has reached 2.99 million, and the losses have exceeded US$110 billion. How to solve the chip shortage has become a topic of great concern to the world. Yu Kai believes that moving towards a central computer architecture will be the ultimate solution to the shortage of automotive chips, because it can greatly simplify supply chain management and improve efficiency.

As the automotive electronic and electrical architecture advances from distributed ECU architecture to domain architecture and finally to central computer architecture, cars will also enter the era of full-scenario vehicle intelligence. The intelligent vehicle computing architecture will also move from the 1.0 era dominated by traditional programming, logic, and processes to the software 2.0 era represented by neural network computing and data-driven. In the future, the whole vehicle will have an AI middle platform, where all data such as environmental perception, intelligent human-computer interaction, and intelligent driving will be calculated, and a more extreme vehicle intelligent driving experience will be created with people as the center.

Breaking the myth of computing power and pursuing true AI performance

The central computer architecture is bound to put forward higher requirements for the computing performance of the whole vehicle. At present, smart car players have set off a computing power arms race and announced the launch of models with 1000T computing power. But does computing power represent the true performance of automotive smart chips? "Top chip companies must not tell this story superficially and simply with how many T. What does 1000T mean? It is not your utility, performance, or user value. It is the cost to the car company." Yu Kai said.

Yu Kai believes that adding physical computing power does not really create user value. The peak computing efficiency of physical computing power is equal to the number of transistors multiplied by the main screen, which is determined by the chip manufacturer, which is the cost of the car company. The real value should lie in how fast the most advanced neural network can run on the chip, that is, the real AI performance.

Journey 5, more than just super AI performance

Based on the forward-looking concept of combining software and hardware, Horizon Robotics was able to quickly achieve iterative upgrades and mass production of products and solutions. It took less than a year for Horizon Robotics' Journey 2 and Journey 3 chips to go from release to mass production, and the key factor lies in the powerful real AI performance. In May of this year, Horizon Robotics' third-generation automotive-grade product, the high-computing Journey 5 series chips for L4 high-level autonomous driving, also successfully taped out at one time and will be officially released within the year.

The AI ​​computing power of a single chip in the Journey 5 series can reach up to 128 TOPS, combining the industry's highest FPS (frames per second) performance with the lowest power consumption. "I think this chip is the most complex and challenging automotive-grade chip in China's entire semiconductor field so far." Yu Kai introduced that the Journey 5 chip is designed for central computing platforms, integrating intelligent human-computer interaction and intelligent driving, and reflects the strategic trend judgment of Horizon AI mid-end.

The power of the Journey 5 chip is not limited to its AI performance, but also reflected in its security mechanism and its support for flexible expansion of computing power of the on-board computing platform.

Horizon is a key contributor to the world's three major safety standards (information security, functional safety and expected functional safety), and the Journey 5 chip provides a complete solution at the security level.

Regarding information security, the Journey 5 chip provides data security protection, a built-in encryption engine, supports more than 9 cryptographic algorithms, and provides security far beyond software encryption, while providing physically isolated storage protection for key/model parameters. In addition, the Journey 5 provides program security protection, an isolated program security operating environment, and a secure software upgrade protection mechanism.

Regarding functional safety, Journey 5 is China's first automotive-grade smart chip based on the product development process system of the highest functional safety level ASIL-D. At the same time, the Journey 5 chip itself meets the ASIL-B level functional safety, with an ASIL-B level BPU architecture, independent safety island & lock-step CPU, complete hardware random failure detection mechanism and full-stack software that complies with functional safety.

Based on the Journey 5 series chips, Horizon Robotics will also launch a series of intelligent driving central computing platforms, Matrix 5.0, with AI computing power of up to 200 to 1000TOPS. The platform integrates autonomous driving and intelligent human-computer interaction functions to create a full-scenario vehicle central computing platform. At the same time, in response to the challenge that car companies cannot accurately estimate the required computing power, a major highlight of Horizon Robotics Matrix 5.0 is that it can provide a variety of specifications to help customers achieve flexible computing power configuration combinations.

From China's first automotive-grade AI chip, Journey 2, which was first installed in Changan UNI-T in June 2020, to Journey 3, which was installed in the 2021 Ideal ONE in May this year, Horizon, as a startup in the field of artificial intelligence chips, has successfully launched each generation of chips, achieved mass production, and all of them are hit models. "We maintain the speed of the past, that is, releasing a generation of chips a year, and then mass-producing the whole vehicle a year later." Yu Kai said.