AI News AI资讯 2h ago Updated 1h ago 更新于 1小时前 48

NVIDIA and Doosan Group Expand Cooperation, Covering Physical AI, Robotics, and AI Factory Infrastructure 英伟达与斗山集团扩大合作,覆盖物理AI、机器人及AI工厂基础设施

NVIDIA and Doosan Group have signed a cooperation agreement covering robotics, construction machinery, energy, and even circuit board materials. The press release is impeccably worded, with sophisticated language and a grand vision. But beneath the haze of PR rhetoric, what I see looks more like a tacit “land grab,” with one party being a thriving computational empire and the other a traditional industrial giant anxious about the future. 英伟达和斗山集团签了个合作协议,覆盖了机器人工厂、建筑机械、能源甚至电路板材料。新闻稿写得滴水不漏,词汇高级,愿景宏大。但拨开这些公关辞令的迷雾,我看到的更像是一场心照不宣的“圈地运动”,一方是如日中天的算力帝国,另一方则是焦虑于未来的传统工业巨头。

70
Hot 热度
65
Quality 质量
70
Impact 影响力

Analysis 深度分析

Let’s start with robotics. Doosan Robotics is set to integrate NVIDIA’s Isaac Sim, Cosmos, and Jetson Thor. Behind all the technical jargon lies a simple ambition: Doosan wants to transform from a seller of “iron arms” into a provider of “AI souls.” The ideal sounds sexy, but reality is harsh. The gap between building a robotic arm that can reliably grasp materials and creating a robot that can autonomously navigate complex environments or even mimic human-like dual-arm collaboration is vast—far from bridged by simply purchasing a few development kits. NVIDIA, of course, welcomes this. Its “Isaac” ecosystem is like fertile digital soil; players like Doosan come to cultivate and experiment, ultimately making the land more valuable. As for humanoid robots? Frankly, most so-called humanoid robots on the market today are more like meticulously orchestrated puppet shows, still a long way from being true “intelligent agents.” Doosan’s entry at this stage is more about staking a claim in the narrative of a future story—a story whose ending, I dare say, even the author hasn’t figured out.

Next, consider Doosan Bobcat, aiming to embed NVIDIA’s physical AI technology into excavators and forklifts. The push to “drive the establishment of an industry-standard ecosystem for autonomous equipment”—does that sound familiar? Every company that aspires to be the leader loves to say this at the outset of a partnership. NVIDIA’s Omniverse and Isaac technologies are indeed powerful, capable of simulation and training. But construction sites and farmlands are the least standardized, most chaotic real-world environments. Sensor failures, sudden obstacles, extreme weather—any variable can cause a carefully trained AI model to lose its mind in an instant. This isn’t about fiddling with blocks in a clean lab; it’s about battling in real mud and dust. The so-called “standard” will likely end up as a binding framework for NVIDIA’s chips and software, while Doosan’s equipment becomes mere “physical vessels” carrying this expensive system.

The most interesting part comes with energy. Doosan Heavy Industries & Construction will provide power for NVIDIA’s AI factories, with solutions including gas turbines, steam turbines, small modular reactors, and hydrogen fuel cells. This is exquisite irony. Training a large model on NVIDIA’s GPUs consumes as much electricity as a small city, making it a carbon-emission behemoth. Now, it needs another traditional energy company to fuel its digital future with more diverse (and possibly more conventional) methods. This perfectly outlines the absurdity of today’s tech industry: the most cutting-edge AI dreams are deeply rooted in the most ancient energy sector. Doosan’s role here resembles a pit crew constantly refueling a speeding race car—the faster the car, the greater the pressure on the crew. Is this technology empowering energy, or is energy dragging technology’s heels? Perhaps it’s a bit of both.

As for using copper-clad laminates to support NVIDIA’s server circuit boards, this part of the collaboration is the most “hardcore” and closest to the essence. It reveals another truth: no matter how dazzling the upper-layer applications, they ultimately land on tangible physical materials. NVIDIA’s MGX ecosystem—from GPU chips to network switches to the circuit boards that carry them—forms a tightly sealed hardware empire. The inclusion of Doosan Electronic Materials means that even the “bricks and mortar” of this empire’s supply chain are starting to bear the imprint of NVIDIA’s collaborative innovation.

So, don’t view this partnership as a simple “alliance of the strong.” It represents NVIDIA’s strategic deepening from being a “seller of picks and shovels” to becoming a “seller of soil.” It’s no longer content with just selling shovels (GPUs); it wants to define the rules for gold mining (Isaac Sim), delineate the boundaries of the mine (ecosystem), and even attempt to influence the mine’s power supply (energy collaboration). Meanwhile, Doosan, once a leader in heavy industry and machinery manufacturing, is now, with unprecedented openness, connecting its core business segments one by one to NVIDIA’s “future matrix.” This is a high-stakes gamble: if it pays off, it might achieve a daring leap from “hard” to “smart”; if it fails, it could end up as just another faceless hardware supplier in a vast digital empire.

There’s no right or wrong in this partnership—just mutual necessity. NVIDIA seeks ubiquitous application scenarios and ecosystem stickiness, while Doosan seeks to board the last high-speed train before being left behind. Just remember, when the massive body of traditional industry begins to dance to AI’s rhythm, don’t forget to ask: who ultimately holds the copyright to this dance?

英伟达和斗山集团签了个合作协议,覆盖了机器人工厂、建筑机械、能源甚至电路板材料。新闻稿写得滴水不漏,词汇高级,愿景宏大。但拨开这些公关辞令的迷雾,我看到的更像是一场心照不宣的“圈地运动”,一方是如日中天的算力帝国,另一方则是焦虑于未来的传统工业巨头。

先说机器人。斗山机器人要整合英伟达的Isaac Sim、Cosmos和Jetson Thor。技术名词一串,翻译过来就是:斗山想从卖铁胳膊的,变成卖“AI灵魂”的。理想很性感,但现实骨感得硌牙。造一个能稳稳抓取物料的机械臂,与造一个能在复杂环境里自主决策、甚至模仿人类双臂协作的机器人,这中间的技术鸿沟,绝不是买几个开发套件就能填平的。英伟达当然乐见其成,它的“Isaac”生态就像一块肥沃的数字土地,斗山这样的玩家来耕种、来试错,最终能让这片土地更值钱。至于人形机器人?恕我直言,目前市面上绝大多数所谓的人形机器人,更像是精心编排的提线木偶,离真正的“智能体”还有十万八千里。斗山此时入场,更像是在抢占一个未来故事的叙事位,至于故事本身,恐怕连作者自己都没想好结局。

再看斗山山猫,要把英伟达的物理AI技术塞进挖掘机和叉车里。推动“自主装备行业标准生态建设”,这话听着耳熟吗?所有想当老大的公司,在合作初期都爱这么讲。英伟达的Omniverse和Isaac技术确实强悍,能做仿真,能训练。但建筑工地和农田,是最非标准化、最混沌的现实环境。传感器失灵、突发障碍、极端天气,任何一个变量都足以让精心训练的AI模型瞬间失智。这不是在干净的实验室里摆弄几个方块,这是在真实的泥泞和尘土中搏命。所谓的“标准”,最后很可能变成英伟达芯片和软件的绑定标准,而斗山设备则成了承载这套昂贵系统的一个个“肉身”。

最有意思的来了——能源。斗山重工要为英伟达的AI工厂提供电力,方案包括燃气轮机、蒸汽轮机、小型模块化反应堆和氢燃料电池。这简直是绝妙的讽刺。英伟达的GPU训练一次大模型,耗电量堪比一个小城市,是碳排放的巨兽。而现在,它需要另一家传统能源企业,用更多样(也可能更传统)的方式,来为它的数字未来“喂电”。这完美勾勒出当前科技产业的怪诞图景:最前沿的AI梦想,其地基深深扎根于最古老的能源产业。斗山在这里的角色,像极了给狂奔的赛车不断加油的后勤队,赛车越快,后勤的压力就越大。这到底是科技赋能能源,还是能源在拖着科技的后腿?也许兼而有之。

至于用覆铜板支持英伟达的服务器电路板,这部分合作最为“硬核”,也最接近本质。它揭示了另一个事实:无论上层应用多么炫酷,最终都要落地在实实在在的物理材料上。英伟达的MGX生态,从GPU芯片到网络交换,再到承载它们的电路板,构成一个密不透风的硬件帝国。斗山电子材料的加入,意味着连这个帝国的“砖瓦”供应链,都开始打上英伟达协同创新的烙印。

所以,别把这次合作看成简单的“强强联手”。它是英伟达从“卖水人”到“卖土壤”的战略深化。它不再满足于只卖铲子(GPU),它要定义挖金矿的规则(Isaac Sim)、划定矿场的边界(生态系统),甚至试图影响金矿的电力供应(能源合作)。而斗山,这位曾经的重工和机械制造翘楚,正以前所未有的开放姿态,将自己最核心的业务板块,一一接入英伟达的“未来矩阵”。这是一次豪赌,赌对了,或许能实现从“硬”到“智”的惊险一跃;赌错了,则可能沦为庞大数字帝国里,一个面目模糊的硬件供应商。

这场合作没有谁对谁错,只有各取所需。英伟达要的是无处不在的落地场景和生态粘性,斗山要的是在被淘汰前搭上最后一班高速列车。只是,当传统工业的庞大身躯开始随AI的节拍起舞时,别忘了问问,这首舞曲的版权,最终握在谁的手里。

Disclaimer: The above content is generated by AI and is for reference only. 免责声明:以上内容由 AI 生成,仅供参考。

机器人 机器人 芯片 芯片 GPU GPU
Share: 分享到: