Pular para o conteúdo
Menu
Mix Vale

Elon Musk confirms resumption of Dojo3 supercomputer by Tesla after advance in AI5 chip

Bruna
Por 6 min de leitura
Compartilhe
Tesla
Foto: Tesla - Tobias Arhelger/ Shutterstock.com

Elon Musk announced the resumption of work on Tesla’s Dojo3 supercomputer. The decision comes after the company achieved stability in the design of the AI5 chip, dedicated to artificial intelligence embedded in vehicles. The Dojo project had been paused last year to focus resources on developing chips optimized for inference and training neural networks. The supercomputer is essential for processing large volumes of data collected by the fleet of Tesla vehicles. Esses data feeds the training of the Full Self-Driving system, responsible for autonomous driving. The return of Dojo3 represents a strategic shift in the company’s approach to high-performance computing. Musk highlighted that the previous focus on a single chip path now allows for splitting efforts again. The initiative reinforces Tesla’s commitment to proprietary AI hardware solutions.

Publicidade

The team responsible for Dojo had been temporarily demobilized. The priority went to the chips that directly equip the brand’s cars.

The announcement was made directly by Musk on his social network.

Publicidade

History of the project Dojo at Tesla

The Dojo project was launched by Tesla with the main objective of creating a customized supercomputer for training artificial intelligence models. Ele uses real data captured by cameras and sensors from vehicles in circulation to improve Full Self-Driving. Diferente from generic solutions, the Dojo is designed to specifically handle large-scale video processing. The first version, known as D1, used 7 nanometer technology and had billions of transistors organized in integrated tiles. Essa architecture allowed high efficiency in training neural networks aimed at computer vision.

However, in 2025, Tesla chose to suspend significant advances on Dojo. The company directed all efforts towards the development of AI5 and AI6 chips, considered more versatile for embedded applications. Musk justified at the time that dividing resources between two different chip architectures was not viable. Vehicle chips needed to support both real-time inference and basic training capabilities. Essa pause allowed us to accelerate the production schedule of the components that equip the current fleet.

Recent announcement and technical context

Musk reported that the AI5 design has reached a stage considered solid. Essa achievement frees up engineers and resources to retake Dojo3 in parallel. The executive emphasized that AI5 and later chips will be excellent for in-vehicle inference and reasonably efficient for training. Ainda thus, Dojo3 represents a platform dedicated exclusively to training on a massive scale. The third generation of the supercomputer will be the first completely developed internally, without dependence on external components such as those of the Nvidia.

The communication included a call for hiring talent. Tesla seeks professionals experienced in complex, high-volume hardware technical issues.

  • Send a resume with three points highlighting the most difficult challenges overcome.
  • Focus on chip production on a global scale.
  • Opportunity to work on AI components with the highest manufacturing volume in the world.
Elon Musk
Elon Musk – Foto: X.com

Differences between training and inference in AI

The chips embedded in Tesla vehicles prioritize inference, that is, the rapid execution of models already trained to make decisions in real time. Eles operate with low latency and reduced energy consumption inside cars. Já training requires much higher computational power to process terabytes of data and adjust billions of parameters in neural networks. The Dojo was designed exactly for this intensive step, enabling faster learning cycles from the real fleet.

Publicidade

With the resumption, Tesla gains flexibility to combine both approaches. AI5 chips handle inference in vehicles and robots, while Dojo3 accelerates training of new versions of Full Self-Driving. Essa division optimizes costs and performance at different phases of autonomy development.

Advances in Tesla proprietary chips

The AI5 chip represents the fifth generation of artificial intelligence hardware developed in-house by Tesla. Seu Advanced design incorporates significant improvements in energy efficiency and performance per dollar. The company plans to produce AI6 in partnership with factories in Texas, ensuring control over the supply chain. Esses components will be manufactured using cutting-edge lithographic processes, rivaling market-leading offerings.

Tesla has already demonstrated the ability to scale production of customized chips. Milhões units equip current vehicles, processing sensor data in real time. The AI5 will raise the level of safety and precision of the Full Self-Driving to levels above human levels in various conditions. Future Versões, such as AI6 and AI7, will expand applications for robotics and other devices.

Practical applications of Dojo3

The supercomputer will process videos and telemetry collected daily by the global Tesla fleet. Esses raw data is anonymized and used to refine perception and route planning algorithms. With Dojo3 operational, Full Self-Driving update cycles may occur at shorter intervals. Isso accelerates the evolution of autonomous driving software without relying exclusively on external providers.

The platform will also support the development of complementary technologies. Robôs humanoids like Optimus require intensive training for complex tasks in varied environments.

  • Recognition of moving objects.
  • Precise physical action planning.
  • Adaptation to unpredictable scenarios in the real world.

Integration with hardware ecosystem

The resumption of Dojo3 occurs at a time of maturation of the Tesla chip ecosystem. The company already operates training clusters with thousands of Nvidia units, but seeks total independence. The custom supercomputer reduces long-term operational costs by optimizing specific architectures for video workloads. Testes initials of previous generations demonstrated efficiency gains compared to generic solutions.

Engineers at Tesla are now working on the integration between the new chips and the Dojo infrastructure. Essa combination will enable unprecedented scalability in training computer vision models.

Production and supply chain

Tesla has entered into manufacturing agreements for the AI ​​chips in state-of-the-art facilities at Estados Unidos. Esses contracts guarantee high volumes from 2027 onwards, with projected performance up to 40 times higher than current generations. Dojo3’s in-house production eliminates external bottlenecks and accelerates hardware iterations.

The company actively recruits high-volume chip design experts. The focus is on solving problems of scale that few organizations face globally.

Dojo3 positions Tesla as a leader in computing dedicated to vehicle artificial intelligence. The platform will process massive amounts of real-world data to continually evolve Full Self-Driving. Combinado to advances in AI5, the project reinforces the strategy of verticalization in hardware and software. The company maintains a fast pace of innovation in autonomy and robotics, using its fleet as an inexhaustible source of practical information.

Publicidade

Bruna Santos é colaboradora da redação do Mix Vale, portal de notícias com cobertura de Brasil e mundo. Produz conteúdo nas editorias de economia e benefícios, com foco em informação de serviço para o leitor brasileiro. Escreve e colabora para o portal desde 2024. Formação: Professora.

Notícias relacionadas

Mais lidas

  1. Ciência

    Meteoro no Nordeste: bólido cruza o céu de cinco estados às 21h34 e mobiliza buscas por fragmentos na Paraíba

  2. Caixa Econômica Federal

    Pé-de-Meia paga parcela de R$ 1.000,00 a aprovados no ensino médio de 2026 entre fevereiro e março de 2027

  3. Política

    Com obstrução parcial no intestino, Jair Bolsonaro é internado no hospital DF Star em Brasília

  4. INSS

    Recuperação da senha Gov.br pelo celular exige reconhecimento facial ou código por SMS para o Meu INSS

  5. EUA

    Terremoto em Los Angeles atinge magnitude 4,2 e causa tremores leves em Commerce na madrugada

  6. Matéria Publicitária

    MineMV é o novo jogo grátis de construir com blocos; veja como jogar e liberado para todas as idades

  7. Política

    Regulamentação da escala 5×2 pelo governo encarece cálculo da hora extra a partir de 16 de dezembro

  8. Santa Catarina

    Acidente de trânsito na SC-108 envolve carro e moto e deixa motociclista morto em São Ludgero

  9. Mundo

    Jornalista italiano Corrado Augias morre aos 91 anos; presidente da Itália lamenta a perda

  10. Loterias

    Rendimento de R$ 40 milhões da Mega-Sena 3088 chega a R$ 337 mil por mês no Tesouro Direto