NVIDIA Starts Installing Chips on Roads | Rewire Evening News Update

marsbit發佈於 2026-03-17更新於 2026-03-17

文章摘要

NVIDIA CEO Jensen Huang announced at GTC that the company's data center orders for Blackwell and Vera Rubin platforms are projected to exceed $1 trillion by 2027, doubling last year's estimates. He emphasized that computing demand will far surpass this figure. Beyond data centers, NVIDIA is expanding its autonomous driving ecosystem, adding BYD, Geely, Nissan, and Isuzu to its Drive Hyperion platform. A partnership with Uber aims to deploy robotaxis in Los Angeles and San Francisco by early 2027, expanding to 28 markets by 2028—a moment Huang calls "the ChatGPT moment for autonomous driving." In related news, Uber co-founder Travis Kalanick revealed his stealth robotics startup, Atoms, after eight years of operation. The company focuses on automating physical infrastructure, mining, and robotic platforms. Kalanick is reportedly acquiring autonomous driving firm Pronto, with Uber's support, signaling a strategic re-entry into automation. Meanwhile, Murata Manufacturing, the world's largest MLCC supplier with over 40% market share, raised prices for AI server and automotive-grade components by 15-35%, effective April 1. This marks its first major price hike in three years and highlights hidden cost pressures in AI infrastructure supply chains. The SEC is also considering allowing public companies to switch from quarterly to semi-annual financial reporting, reducing compliance costs and potentially benefiting tech firms making long-term AI investments. Additional updates inc...

After labeling the trillion-dollar orders, Jensen Huang has brought BYD and Geely to the autonomous driving table.

1| Jensen Huang's Trillion Dollars Doesn't Even Include the Roads

This morning, at GTC, Jensen Huang announced that procurement orders for Blackwell and Vera Rubin will exceed $1 trillion by 2027. This number is double last year's estimate. He added, "We are supply-constrained, and I am sure the demand for computing will far exceed this number."

But the $1 trillion is only the data center's account. On the same day, NVIDIA announced that BYD, Geely, Nissan, and Isuzu have joined the Drive Hyperion platform to develop Level 4 autonomous driving. Robotaxis developed in collaboration with Uber will be deployed in Los Angeles and San Francisco in the first half of 2027, covering 28 markets by 2028. Jensen Huang called this the "ChatGPT moment for autonomous driving."

(Source: CNBC / TechCrunch / The Verge)

2| Kalanick, After Lying Low for Eight Years, Is Betting That the Physical World Has Not Yet Been Automated

Uber co-founder Travis Kalanick has revealed his secretly operated robotics company, Atoms, which has been running for eight years. Formerly known as CloudKitchens and City Storage Systems, the company has thousands of employees, none of whom were allowed to list the company name on LinkedIn. Its three business lines focus on food infrastructure, mining automation, and robotic chassis. In a public letter, he wrote a defining statement: "Software has automated language and mathematics, but the full automation of the physical world remains an untapped frontier."

According to Fortune, Kalanick is about to acquire the autonomous driving company Pronto, founded by former Uber colleague Anthony Levandowski, with support from Uber. Eight years ago, when he was ousted from Uber, autonomous driving was still Waymo's solo show. Eight years later, he has returned with robotics, and NVIDIA just announced the "ChatGPT moment for autonomous driving" in the same week. The timing is not a coincidence—it's a signal.

(Source: All-In Podcast / Fortune / TechCrunch)

3| Murata Raises Prices by 35%, the Hidden Bill for AI Infrastructure Has Arrived

According to 36Kr, citing a Shanghai Securities News report, Murata Manufacturing, the world's largest MLCC supplier, has initiated a comprehensive price increase for AI servers and high-end automotive-grade products, with a range of 15% to 35%, effective April 1. Murata holds over 40% of the global MLCC market share and 70% of the AI server MLCC market share. This is the first large-scale price adjustment in three years.

Everyone is counting the price of GPUs. No one is counting the price of that capacitor on the circuit board. Murata's monopoly in passive components for AI servers is no less than NVIDIA's monopoly in GPUs—it just doesn't host a GTC. When the invisible layer of the supply chain starts pricing scarcity, the $1 trillion infrastructure bill will only go up.

(Source: 36Kr / Shanghai Securities News)

4| SEC Wants Public Companies to Submit Reports Only Twice a Year

According to TechCrunch, citing a WSJ report, SEC Chairman Paul Atkins is discussing with exchanges the possibility of allowing public companies to switch from quarterly to semi-annual reporting. The reason is that the compliance cost of quarterly financial reports is too high, discouraging companies from going public. This is the biggest potential change in the U.S. public company disclosure system in over 50 years.

For tech companies betting heavily on AI infrastructure, submitting two fewer reports means having to explain "where the money went" two fewer times. The AI capital expenditure that Meta burns through in a quarter can now be hidden in the larger numbers of a semi-annual report. The beneficiaries of deregulation are, first and foremost, those companies that need time to prove the correctness of their long-term investments.

(Source: TechCrunch / WSJ)

Also Worth Knowing ↓

Alibaba has established the Alibaba Token Hub business group and is providing all employees with AI tool Token allowances. Employees can use paid tools such as Wukong and Qoder series for free, and purchases of Bailian Coding Plan or external AI development tools can be reimbursed. From "encouraging trial use" to "standard equipment for all," Token is becoming the second means of production for employees of large companies. (Source: 36Kr)

The FDIC is preparing to halt pass-through insurance for stablecoin deposits. If passed, stablecoin holders will no longer be able to obtain FDIC deposit protection through issuers. The "deposit-like" narrative of stablecoins has just had its last safety rope pulled away by regulators. (Source: Payments Dive)

Netanyahu deepfake conspiracy theories are spreading on social media, forcing the Israeli government to respond. The so-called "evidence" includes extra fingers in videos and coffee cups defying gravity. During wartime, deepfake has evolved from a technical issue to a national security issue. (Source: The Verge)

Picsart has launched an AI Agent market, allowing creators to "hire" AI assistants on demand. The first four Agents are now available, with weekly additions. Creative tool platforms are shifting from selling features to selling labor. (Source: TechCrunch)

相關問答

QWhat major announcement did Jensen Huang make at GTC regarding Blackwell and Vera Rubin orders?

AJensen Huang announced that the GTC orders for Blackwell and Vera Rubin are projected to exceed $1 trillion by 2027, which is double last year's estimate. He also stated that demand for computing would far surpass this number.

QWhich automotive companies did Nvidia announce have joined the Drive Hyperion platform for L4 autonomous driving development?

ANvidia announced that BYD, Geely, Nissan, and Isuzu have joined the Drive Hyperion platform to develop L4 autonomous driving technology.

QWhat is the significance of Murata Manufacturing's price increase for MLCCs in the context of AI infrastructure?

AMurata Manufacturing, the world's largest MLCC supplier with over 40% global market share and 70% in AI server MLCCs, raised prices by 15-35% for AI servers and high-end automotive products. This highlights hidden costs in AI infrastructure and indicates that the $1 trillion infrastructure bill could increase due to supply chain constraints.

QWhat change is the SEC considering for corporate reporting, and how might it benefit tech companies investing heavily in AI?

AThe SEC is discussing allowing public companies to switch from quarterly to semi-annual reporting to reduce compliance costs. This would benefit tech companies heavily investing in AI by allowing them to explain large capital expenditures less frequently, masking short-term costs in broader semi-annual reports.

QWhat did Travis Kalanick reveal about his secret robotics company Atoms, and what is its focus?

ATravis Kalanick revealed Atoms, a secret robotics company operational for eight years with thousands of employees. It focuses on automating the physical world through three business lines: food infrastructure, mining automation, and robotic chassis. He stated that while software has automated language and math, the physical world remains largely unautomated.

你可能也喜歡

年薪百万抢电工,Meta急到自己办技校

AI竞赛正面临新的瓶颈:工地。美国正面临严重的电工、建筑工等技术工人短缺,这已成为微软、Meta、OpenAI等公司快速建设超大规模AI数据中心(如OpenAI耗资160亿美元的“星际之门”项目)的“头号障碍”。 尽管AI公司愿支付高薪(如电工年薪可达24-28万美元),远超传统行业,但熟练技工仍供不应求。麦肯锡预测,美国在2023-2030年间需额外培养13万名电工和24万名建筑工,而劳工统计局预计每年仍有8万个电工岗位空缺。这种短缺导致项目延迟,每月可能造成数百万美元的收入损失。 AI数据中心建设复杂,需应对巨大功耗(一座设施耗电堪比数十万户家庭)、复杂的配电系统以及高密度散热(需液冷技术)等挑战,因此亟需大量技术娴熟的工人。 为此,科技巨头开始亲自下场培养人才。例如,Meta投入1.15亿美元建立建筑工人培训学校,提供免费培训及生活补贴,以快速输送工人上岗。OpenAI则与建筑工会合作,提前锁定熟练劳动力。同时,企业也将招聘目光投向高中生,鼓励年轻人投身技工行业。这些举措已见成效,Z世代对技工职业的兴趣显著上升。 然而,更深层的挑战在于电力。AI数据中心用电量正以惊人速度增长,已推高部分地区的电价。此外,数据中心建设是项目制的,建设期需要成千上万的工人,但建成后仅需少量常驻人员。这意味着未来可能面临熟练工人短期过剩、并流向其他行业压低薪资的风险。如何实现劳动力与电力资源的长期平衡,仍是悬而未决的问题。

marsbit30 分鐘前

年薪百万抢电工,Meta急到自己办技校

marsbit30 分鐘前

OpenAI 不靠最贵的模型卖钱了

OpenAI近期调整了API定价策略,旗舰模型GPT-5.6 Luna降价80%,Terra降价20%,而顶级模型Sol未降价但新增了更快的“Fast”模式。此次调整的核心并非单纯的价格战,而是标志着OpenAI首次引导用户根据任务需求选择不同模型,而非一味追求最强模型。官方建议复杂任务可由Sol进行规划,再由Luna等成本更优的模型执行。 此举与Anthropic近期推出半价替代旗舰的Claude Opus 5策略相似,显示硅谷领先公司正形成默契:旗舰模型(如Sol、Fable)的角色正从直接盈利转向树立技术品牌和探索上限,而商业化走量和利润则主要由中端及性价比模型(如Luna、Opus)承担。这类似于汽车等行业“旗舰立形象,走量车型赚利润”的模式。 更深层的变化在于成本下降的驱动力。OpenAI透露,此次降价部分得益于由Sol模型自身参与优化底层生产内核,实现了效率提升与成本降低,开启了“模型优化模型”的自我加速循环。 行业竞争焦点正从“谁最聪明”的单项性能比拼,转向为企业提供基于任务重要性、出错成本、规模等因素的“模型组合”最优解,追求整体ROI。这类似于云计算按数据冷热使用不同存储的策略。 OpenAI的长期战略日益清晰:其真正目标并非依靠单一模型的高毛利,而是通过极低的调用成本构建庞大的开发者生态与调用量,形成类似操作系统或云平台的渗透率与用户粘性。当API调用成本低至可被忽略,AI将如水电般无缝嵌入所有业务流程,其发展也将从追求智能巅峰的“产品时代”,进入驱动全面自动化与效率的“基础设施时代”。

marsbit30 分鐘前

OpenAI 不靠最贵的模型卖钱了

marsbit30 分鐘前

交易

現貨
活动图片