The mainstream mobile phone market has become saturated, with major players now shifting their focus to the next big opportunity: self-driving cars. Drawing from the lessons of the internet era, we know that chips are always at the top of the industry chain and serve as the foundation for technological leadership. Core chips often define the infrastructure of a computing age.
As a new computing era emerges, it presents a golden opportunity for startups to overtake established players. While GPUs and FPGAs have been widely used, they are not specifically designed for artificial intelligence. This means there are inherent limitations. AI chips, on the other hand, offer a level playing field where both tech giants and startups can compete from the same starting line.
Currently, deep learning and other AI applications rely heavily on general-purpose chips like GPUs, FPGAs, and custom ASICs to accelerate parallel computing tasks. However, as AI evolves, the need for more specialized and efficient hardware becomes increasingly important.
In this key area, Intel, NVIDIA, and Qualcomm stand out as the three major chip giants. Their competition has driven much of the drama in the ongoing chip war.
NVIDIA: Unmatched Performance
When it comes to AI chips, NVIDIA is unquestionably a global leader. This dominance is reflected not only in its technology but also in its stock price, which has risen tenfold over the past two years.
NVIDIA’s GPGPU (General-Purpose GPU) is widely recognized in the deep learning space. Most deep learning training today is powered by NVIDIA's GPUs. This is due to their superior computational power compared to CPUs and the ease of use offered by NVIDIA's development tools, which significantly lower the barrier for developers.
Before autonomous driving became a reality, NVIDIA entered the automotive market through its Tegra processors, mainly for in-vehicle entertainment. But now, the company is making a stronger push into self-driving technology.
For example, the new Audi A8 with L3 autonomy uses NVIDIA’s Tegra K1 chip for surround vision. The Tegra K1 delivers 350 GFLOPS of single-precision floating-point performance.
NVIDIA has since launched two generations of powerful self-driving computers. The most advanced version, Drive PX 2, uses two Tegra Parker SoCs, offering up to 8 TFLOPS of computing power. However, this high performance comes at a cost—consuming up to 250W and requiring water cooling, with a price tag of $15,000. These factors make it less ideal for mass-produced autonomous vehicles.
To address this, NVIDIA introduced the Xavier chip, based on its Volta GPU architecture. It offers 20 TFLOPS of single-precision performance while consuming only 30W. Xavier is powerful enough to support L4-L5 autonomous driving and is compact and energy-efficient, making it suitable for vehicle integration.
Xavier is expected to start production early next year, with large-scale shipments scheduled for the fourth quarter.
Intel: Strategic Acquisitions
Over the past two years, Intel has been playing catch-up in the AI chip race. Instead of focusing on GPUs, Intel made a bold move by acquiring Altera, the world’s second-largest FPGA company, for $16.7 billion in 2016. This acquisition allowed Intel to integrate FPGA chips into its autonomous driving platform, Intel Go, which is used in systems like Audi’s new A8.
In March of this year, Intel acquired Mobileye for $15.3 billion. Mobileye holds over 70% of the global ADAS market and has developed the EyeQ series of chips specifically for advanced driver assistance systems. This acquisition enabled Intel to build a comprehensive vehicle computing solution: Intel Atom + Mobileye EyeQ + Altera FPGA.
While Intel has strong presence in the server market, allowing it to provide cloud-based computing power, the real-time data transmission between vehicles and the cloud is still limited. Therefore, onboard computing remains critical for autonomous vehicles.
However, building the necessary standards and infrastructure for a smart city brain model will take time and significant investment.
Qualcomm: NXP Acquisition Uncertain
Qualcomm has been active in the automotive space since 2014 with its Snapdragon 602A chip, aimed at in-car entertainment and audio processing. In 2016, it released the Snapdragon 820A, featuring hardware acceleration for deep neural networks, targeting the ADAS market.
While Snapdragon offers better energy efficiency than some competitors, its peak performance is limited by power constraints. Additionally, it was not specifically designed for ADAS, making it harder to gain traction in the market.
To strengthen its position, Qualcomm announced the acquisition of NXP, the world’s largest automotive semiconductor manufacturer, for $47 billion. NXP holds 14.6% of the global automotive semiconductor market and has strong relationships with car manufacturers and suppliers.
However, the deal is still under regulatory review, with the EU recently suspending the process due to insufficient disclosure. For Qualcomm, finalizing the acquisition is crucial for its long-term strategy in the automotive sector.
As AI continues to evolve, major tech companies like NVIDIA, Microsoft, Qualcomm, and Google are investing heavily in AI chips to control the core architecture and gain a competitive edge in the coming computing era.
In the next 3–5 years, as AI-specific chips advance, industries will fully embrace artificial intelligence. The intelligent market is expected to be dozens of times larger than today’s mobile internet market, opening up new opportunities for innovation and growth.
Flex Power Supply,Flex Power Supply 80Plus,700W Adapter Server Power Supply,Flex 700W Power Supply
Boluo Xurong Electronics Co., Ltd. , https://www.greenleaf-pc.com