The Attack of Artificial Intelligence How Nvidia detonated the next trillion dollar market

Mondo Technology Updated on 2024-03-05

However, due to Nvidia's performance is too appalling, the valuation of ** is getting cheaper and cheaper.

Graphics card from 2016"Door-to-door delivery"Although the incident shows NVIDIA's transformation to AI, the importance it attaches to AI is debatable.

Despite artificial intelligence pioneer Hinton's plea for sponsorship, Nvidia turned it down. It can be seen that at that time, Nvidia did not attach much importance to artificial intelligence.

Musk signed DXG-1 as chairman of OpenAI, 2016.

Before sending graphics cards to OpenAI, there were only two new tracks that Mr. Huang dreamed of: mobile chips and automotive chips.

In 2013, Nvidia founder Jensen Huang praised Xiaomi in broken Chinese. The Xiaomi Mi 3 is equipped with Nvidia's Tegra chip, marking Nvidia's entry into the mobile phone market and opening up a second battlefield in addition to graphics cards.

Xiaomi Mi 3 launch, 2013.

Nvidia exited the smartphone chip market and turned its attention to other areas. Despite having the Tegra series, the business has been surpassed by Qualcomm and MediaTek. Huang attributed this to Nvidia's strategic realignment to focus on areas where it has more strengths.

Nvidia's automotive chip business has been unknown for a long time, but the company cares for it, led by veteran Gary Hicok. Nvidia CEO Jensen Huang plans to increase the share of automotive revenue to 30 percent, demonstrating the importance and confidence in the business.

The past cannot be admonished, the coming can still be chased, and the track of automotive chips can also create another NVIDIA.

The new energy vehicle industry is undergoing the reshaping of the industrial chain from electrification to intelligence. The wave of electrification in the first half has reshaped the composition of components, and the second half of intelligence will further promote the transfer of power in the industrial chain.

Power batteries replace engines and become the core components of electric vehicles, accounting for more than 30% of the cost. Giants such as LG Energy and CATL have risen in the capital market with their battery business.

With the popularization of high-computing power chips in autonomous driving, the traditional distributed electronic architecture will face challenges. Automobiles will be transformed into mobile devices with powerful computing power, driving the evolution of automotive electronic architectures towards centralization.

Musk's architectural innovation has opened the door to Nvidia's third curve.

Tesla has revolutionized the intelligence of automobiles and replaced the traditional microcontroller architecture. Now, the advanced computer system handles all functions centrally, including smart wipers, automatic locking. This centralized design improves vehicle performance, reliability, and safety.

Tesla has revolutionized its in-vehicle systems, replacing its scattered MCUs with more powerful** chips, and concentrating a variety of functions on autonomous driving and intelligent cockpit chips. This reduces the number of chips, simplifies the system, and improves efficiency.

First, car companies can modify and control the iteration of software functions by themselves.

Tesla's centralized electronic architecture revolutionizes the way cars are upgraded, and is no longer constrained by hardware functions that are fixed at the factory. By directly rewriting the software, Tesla made it happen"Hardware embedded, software upgraded", first accumulate computing power, and then gradually unlock the potential of hardware through OTA, breaking the limitation of traditional car functions that cannot be modified.

In 2019, the well-known car **Top Gear said that the acceleration of the Porsche Taycan is better than that of the Tesla Model S. Tesla CEO Elon Musk was quick to respond, accusing him of favoritism and modifying the motor software algorithm to increase the Model S's horsepower by 50 hp, ultimately beating the Taycan.

The second is to drive high-speed pilot, automatic parking and even higher-level autonomous driving functions.

Tesla continues to make breakthroughs in self-driving technology. In 2020, the Transformer architecture fuses 3D perspective and timing information to achieve 4D spatial awareness. In 2022, the introduction of the occupancy network further improved the general obstacle recognition capability.

With the large-scale modelization of autonomous driving algorithms, the demand for computing power has surged, and the super chips required to drive intelligent functions and autonomous driving algorithms have huge computing power"Super Brain"。As Huang said, TOPS (Computing Power Measure) has become the new yardstick for measuring chip performance.

Tesla reconstructed the image in 4D from a large model.

Although the power of family cars such as Leiling and Sylphy is far less than that of F1 cars, the computing power requirements of cars equipped with advanced autonomous driving systems will far exceed those of traditional vehicles, up to 100 or even 1,000 times.

With the vigorous development of self-driving cars, the global automotive chip market has reached $45 billion in 2022, which is comparable to the size of the mobile phone chip market.

IDC** sees a surge in shipments of L3 and above autonomous vehicles over the next two years, with an annual growth rate of more than 100%.

is undergoing a revolutionary restructuring"Superchips"Replacing a large number of MCUs, leading to the concentration of vehicle value. Tesla's HW3A case in point is the 0 system, where two FSD chips account for 61% of the cost, compared to only 5% of the total cost of a traditional MCU. This shift is indicative of the soaring computing power in automotive technology.

Veteran automotive chipmakers, such as Renesas and NXP, excel at producing reliable MCUs. However, in the competition for computing power, Nvidia and Qualcomm lead the industry with their powerful processing power.

Musk understands the critical importance of core components to the new energy industry. As a result, Tesla has developed and controlled its key technologies independently for several years, including self-driving chips (FSDs), cloud computing chips (DOJOs), and Linux-based operating systems.

There are two main categories of on-board chips:

Smart cockpit chips: support navigation, entertainment and other functions, such as the 8295 chip and AMD Ryzen series.

Autonomous driving technology has a high demand for computing power, which has spawned many players to join the market. In addition to Mobileye and Horizon, which were early entrants, consumer electronics giants such as Qualcomm and Huawei are also actively grabbing market share.

Tesla had relied on NVIDIA's Drive PX 2 system, but the high cost (1$50,000) led to a parting of the ways. In 2019, Tesla turned to self-developed FSD chips, announcing the end of its cooperation with NVIDIA.

Self-developed chips can indeed achieve the optimal adaptation of software and hardware, but the problem is that Tesla's gameplay is too difficult.

Under the wave of new energy vehicles, traditional automotive chip giants are difficult to transform due to inherent thinking, and choose the ARM ecosystem instead. Although the new car-making forces advertise their own development, they actually rely on third-party manufacturers. In the field of chips, the automotive industry is at a watershed, with traditional giants and new forces parting ways, each pursuing a different direction.

Where old players have more than enough energy to spare, it is the huge market of "soul ** business".

Since entering the automotive market in 2015, NVIDIA's automotive chips have launched six generations of products:

In 2015, NVIDIA launched the Tegra chip with 1 tops of computing power to provide intelligent cockpit computing power for early smart cars. Models such as the Model S X and the NIO ES6 8 have been equipped with this chip.

The performance of NVIDIA's Tegra Parker processor has increased rapidly in recent years, from 3TOPS in 2016 to 30TOPS in 2020, and is now widely used in automotive applications, such as the Xpeng P7.

Equipped with the cutting-edge 250TOPS computing power, the NVIDIA ORIN processor empowers new energy vehicles such as NIO ET7 and Xpeng P7, opening a new era of intelligent driving.

Nvidia released the THOR processor with an astonishing 2000 Tops computing power, surpassing the originally planned 1000 Tops ATLAN product line and demonstrating NVIDIA's strength in the AI field.

Thor integrates the functions of the cockpit and autonomous driving chips to achieve the integration of soul chips. Geely's Zeekr has been booked for the first time, indicating that the technology is about to land.

In 2022, Nvidia released the Thor chip.

Nvidia's chips are known for their versatility, covering areas such as gaming, AI, and autonomous driving. This versatility is not a weakness, but a business advantage. Nvidia's competitors typically focus on application-specific chips, while Nvidia has succeeded in multiple markets with its general-purpose chips.

The biggest feature of NVIDIA's product lines is the "common architecture", which updates the architecture every two years to cover all products.

X**ier's Volta architecture, which is also used in the high-performance GPU V100;

Orin uses the Ampere architecture, which is the underlying technology for Nvidia's consumer graphics cards, the RTX 30 series, and the high-performance GPU A100. This versatile platform allows a single architecture to meet the performance needs of three different domains.

The competition of chips is not only about computing power, but also about reducing the cost of a single chip. Just as pharmaceutical companies need to share huge sales volume in R&D investment, the competitiveness of chips essentially depends on their large-scale production capacity.

Breaking through the R&D cost barrier, NVIDIA has significantly reduced the cost of Thor's automotive chip development with a huge shipment of games and data center businesses. The Hopper architecture has been amortized by the H100 to make the automotive chip extremely cost-effective.

Under the wave of automotive intelligence, consumer electronics giants such as Qualcomm, Huawei, and Nvidia have become strong supporters behind car companies to build intelligent cockpits and autonomous driving systems by virtue of their leading technological advantages.

Compared to the first two rivals, Nvidia's car dream may need the last piece of the puzzle.

In 2016, a fatal Tesla accident prompted Elon Musk to break up with Mobileye. Musk has instead partnered with Nvidia to power self-driving solutions.

Model S crash scene.

As a giant in the field of autonomous driving chips, Mobileye once occupied a near-monopoly market share. However, there are signs of a rift in Tesla's partnership with Mobileye, and the crash could accelerate the process.

Mobileye adopts the strategy of binding algorithms and chips to provide a "black box" solution that cannot be modified. This low-cost model is suitable for car companies, but for Tesla, which is looking for algorithmic autonomy, it cannot meet its needs.

In 2015, Huang got his hands on a Model X P90D ahead of schedule

NVIDIA has built an open automotive platform that provides high-performance chips and a full set of software tools to enable automakers to independently develop software and algorithms to accelerate intelligent automotive innovation.

The core competitiveness of autonomous driving, algorithms, prompts car companies to develop their own. In the early days, the new forces mostly adopted the Mobileye solution, but as the development turned to NVIDIA, highlighting the advantages of the latter.

With the powerful toolchains of Drive OS, DriveWorks, Drive**, and Drive ix, NVIDIA empowers automotive companies. Whether strong or fledgling, car companies can benefit from the R&D of the underlying system to the R&D of the upper-level application. NVIDIA provides a complete solution to help car companies build excellent software and achieve smart mobility.

NVIDIA is known for its knife skills and clever use of diversified product portfolios to meet different customer needs. By building a high-end, high-end, and low-end product line, NVIDIA provides tailored solutions for a variety of application scenarios to meet differentiated needs from entry-level to high-end.

And the puzzle that Nvidia is currently lacking is "adaptation". That is, the delivery team that connects with the car company, and the corresponding delivery capability.

In June 2020, NVIDIA and Mercedes-Benz reached a strategic cooperation, under which NVIDIA provided Mercedes-Benz with an AI software architecture covering autonomous driving. It marks the upgrade of NVIDIA's cooperation in the automotive field from Tesla Model X to Mercedes-Benz S-Class, demonstrating NVIDIA's leading position in the field of autonomous driving.

Jensen Huang with Mercedes-Benz CEO Ola Kaellenius

Nvidia's partnership with Mercedes-Benz ties Nvidia's revenue to sales of Mercedes-Benz products, aiming to generate a 30% revenue share for its automotive business. However, Nvidia has repeatedly failed in cooperation, triggering rumors that Mercedes-Benz is considering introducing new ** merchants.

Nvidia attaches great importance to adaptability and poached many experts such as Wu Xinzhou, vice president of Xpeng, and Luo Qi, team leader. Wu Xinzhou became the highest-ranking Chinese executive in a global technology company, highlighting NVIDIA's determination to make up for the shortcomings of adaptation.

Nvidia dominates the field of graphics computing with its chip design and software development. However, its lack of manufacturing capacity has led to a competitor with Huawei, which has a strong manufacturing base. As a result, Nvidia sees Huawei as its most important competitor to balance its strengths and meet market challenges.

For the first time, Huawei has joined the ranks of NVIDIA's competitors, standing out in the fields of GPU-accelerated chips, cloud-developed chips, ARM-based CPUs, and network products. With its advantages in the field of communications, Huawei surpasses AMD and competes head-to-head with NVIDIA in four major areas.

The automotive industry is facing a new challenge: the high cost of autonomous driving chip manufacturers. Similar to the NVIDIA graphics cards in PC production, the **label on the self-driving chip is driving the car***

- What are your thoughts on this? -

-Welcome atLeave a message in the comment area** and share. -

Related Pages