news

is tesla’s computing power equal to the sum of chinese car companies? "transforming" ai company's computing power and algorithms into a decisive high ground for the future

2024-10-05

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

① in addition to xpeng motors, which is not willing to "change its name", nio, a new force in car manufacturing, and chery, a representative of traditional car companies, are all focusing on ai. ② zhang yongwei, vice chairman and secretary-general of the china electric vehicles association of 100, pointed out that after entering the ai ​​era, what the automotive industry lacks most is intelligent computing infrastructure, not production capacity.

financial associated press, october 5 (reporter xu hao)with the application of end-to-end and generative large models, automobile companies that have entered the second half of smart electrification are accelerating their transformation into ai companies. in this transformation, computing power and algorithms will become key elements in determining the future destiny of enterprises.

"xpeng motors will be an explorer of future travel and become a global ai car company." as tesla fsd is about to enter the chinese market, he xiaopeng, chairman of xpeng motors, went to the united states to experience tesla's ai capabilities in autonomous driving. after applying it to driving, it was announced at the end of august that the company would be upgraded to "xpeng ai automobile company". according to the plan, xpeng motors’ ai system includes ai cars, ai robots and flying cars, as well as ai chips.

tesla is the initiator of applying end-to-end technology to the automotive industry.since 2014, tesla has launched the first version of autopilot, its self-developed assisted driving system, and then released an enhanced autopilot with richer functions in 2017. by november 2022, tesla will further open the option of fsd beta smart driving system to car owners in north america. the blessing of ai has also become an important driver of tesla's high valuation.

in addition to xpeng motors, which is not willing to "change its name", nio, a new force in car manufacturing, and chery, a representative of traditional car companies, are all focusing on ai. in addition to smart cockpits and smart driving, ai has entered many fields and links such as "research, production, supply, and sales."

"now the entire company is using ai code tools to generate some codes, and the production efficiency has been improved by 30%. the best application scenario of artificial intelligence is actually the smart electric vehicle industry." li bin, chairman of nio, said recently that a successful a smart electric vehicle company must have ai r&d capabilities and operational capabilities; a successful smart electric vehicle company must also be a successful ai company.

it is reported that chery automobile is also currently building lion technology, dazhuo intelligence and next ai research institutes, focusing on the three major areas of smart cabin large models, end-to-end smart driving large models and ai intelligent management, empowering new products through ai. realize intelligent commercial applications. "as cars upgrade from transportation to a new generation of mobile intelligent terminals, ai not only redefines cars, but also reshapes the management model and business boundaries of cars." wang lang, deputy general manager of chery automobile, said that ai is the new driving force of the times. , is an important tool for automotive products to achieve uniqueness and differentiation. “every era has its own main contradictions and new driving forces of its own era. just like chery firmly chose to independently develop engines 20 years ago, ai must be a major opportunity that chery must seize in the next 20 or 40 years. "

with the application of ai, cars need to process a large amount of perception data, decision-making and control tasks, so the requirements for computing power have increased dramatically. according to industry insiders, entering the era of ai competition, car companies must form core competitiveness in computing power, chips, and algorithms.

"to complete the research and development and training of end-to-end intelligent driving, the starting computing power is 1 eflops. huge investments must be made in computing power, and investment must be continued to form a large-scale team around data, computing power, and algorithms." china. zhang yongwei, vice chairman and secretary-general of the electric vehicles committee of 100, pointed out that after entering the ai ​​era, what the automotive industry lacks most is intelligent computing infrastructure, not production capacity. “without a computing power cluster with thousands or tens of thousands of cards, and without an algorithm team with thousands or tens of thousands of cards, it will be difficult for companies to become competitive on the new track.”

in zhang yongwei's view, in addition to building local computing power, we must also speed up the solution to the problems of immature domestic computing power, supporting tool chains, and weak ecological capabilities, and reduce the risk of computing power being "stuck" in hardware.

data shows that the current computing power of domestic car companies is less than 10 eflops.by the end of 2024, the planned computing power of china mobile, china telecom, and china unicom are 17 eflops, 21 eflops, and 15 eflops respectively, and the three major operators total 53 eflops. however, based on the "end-to-end" large model, the ideal computing power required by an enterprise the power is as high as 100 eflops. in comparison, tesla has a computing power level of 100 eflops, which is the total computing power of all car companies.

according to zhang yongwei’s vision, in the future, we should strengthen the co-construction and sharing of computing power to improve resource utilization efficiency from four aspects. first, the concentration of computing power to improve model iteration efficiency; second, basic model sharing to reduce resource reuse; third, software and hardware collaboration, improve software and hardware adaptation; fourth, ecological co-construction and improve the software development ecosystem.

developing computing power is a long-term systematic project that requires the entire industry to build together. li ming, director of jiangxi automobile group holding co., ltd. believes that with the growing demand for large computing power in smart cars, it is recommended that anhui province further strengthen the construction of computing power centers and cloud platforms, making them part of public service facilities and empowering enterprise development.