Bezos is raising $100 billion to acquire factories and reshape them with AI, while Samsung has just bet $73 billion on chips. The battlefield of computing power is shifting from the cloud to the factory floor.
1｜Bezos to Inject AI into Factories with $100 Billion
According to the WSJ, Bezos is in negotiations with global top-tier asset management institutions such as the Abu Dhabi Investment Authority and JPMorgan to raise a $100 billion fund for Project Prometheus. The goal is to acquire traditional industrial companies in chip manufacturing, defense, aerospace, and other industries, and then transform them with AI. Bezos himself serves as co-founder and co-CEO, partnering with former Google executive Vik Bajaj.
This scale is on par with the SoftBank Vision Fund. On the surface, it appears to be a tech billionaire doing private equity, but underneath lies the main battlefield of the AI competition shifting from model training and cloud inference to downstream manufacturing. When Bezos starts using AI to buy factories instead of building data centers, it shows that he believes the next wave of AI value is not in the cloud, but on the production line. On the same day, Samsung announced a $73 billion chip investment for 2026 (a 22% increase year-over-year), surpassing TSMC's annual capital expenditure, betting on HBM4 and 2nm processes. From chips to factories, capital is rapidly descending along the physicalization path of AI.
(Source: WSJ / Bloomberg)
2｜Anthropic Case Upgrade: Pentagon Plays Foreign Labor Card, Capitol Hill Holds Closed-door Session
In a legal filing on March 17, Pentagon Deputy Secretary Emil Michael raised a new argument, stating that Anthropic employs a large number of foreign national workers, "including citizens of the People's Republic of China," citing China's National Intelligence Law, suggesting that these employees may be compelled to assist in intelligence collection, posing a "counterintelligence risk." This is the Pentagon's third line of attack following previous characterizations of "supply chain risk" and allegations of "potential technology disabling in combat."
According to Reuters on the same day, actual military users expressed that replacing Claude is "not that easy." On the other hand, Anthropic co-founder Jack Clark held a closed-door meeting on Wednesday with bipartisan lawmakers from the House Homeland Security Committee, discussing national security and AI, but only briefly touching on the Pentagon's lawsuit. The hearing on March 24 will address an unprecedented question: whether the ethical red lines of an AI company are legally considered protected speech or can be defined as "conduct" and thus not protected by the First Amendment.
(Source: Axios / Reuters / CNN)
3 | White House Considers Easing Iran Oil Sanctions Amid Conflict
According to Axios, Treasury Secretary Bessent on Thursday confirmed for the first time that the White House is considering lifting some sanctions on Iranian oil in transit to help lower oil prices. Brent crude surged 10% in the past 24 hours, briefly hitting $119 before retreating to around $111. On the same day, another White House official stated that "there is no consideration of oil export limits." The six Western allied countries (UK, France, Germany, Italy, the Netherlands, Japan) issued a joint statement supporting the reopening of the Strait of Hormuz, but the statement did not include a commitment to send warships and was more of a gesture to appease Trump, according to Axios.
Defense Secretary Hegseth confirmed plans to request approximately $200 billion in funding from Congress for the Iran war effort but acknowledged that "this number is subject to change." On the surface, it's about energy price control, but behind the scenes, Washington is simultaneously engaged in three conflicting actions: waging war, easing sanctions on a key adversary's economic lifeline, and seeking help from allies. The logic of these three lines of action is independent, and no one knows where they intersect.
(Source: Axios / Financial Times)
4 | Cursor's In-House Model Surpasses Claude, AI Programming Tools Begin "De-platforming"
AI programming tool Cursor (parent company Anysphere, valued at $293 billion) has released its proprietary model Composer 2. In the Terminal-Bench 2.0 benchmark test, Composer 2 scored 61.7%, surpassing Anthropic's Claude Opus 4.6 (58.0%), but lagging behind OpenAI's GPT-5.4 (75.1%). The key differentiator is pricing, with Composer 2 priced at $0.50 per million token input and $2.50 per output, only one-tenth of Opus 4.6's cost.
According to Bloomberg's report on the same day, Cursor is developing even larger-scale models to directly challenge Anthropic and OpenAI. On the surface, it's about a programming tool releasing a new model, but underlying it is the most profitable companies in the AI application layer beginning to intrude into the model layer in reverse. The reason is simple: the cost of reasoning determines the profit margin, and whoever controls the model controls pricing power. Accenture's financial report on the same day showed that strong demand for enterprise AI is driving revenue above expectations, validating downstream payment willingness based on the upstream competitive logic.
(Source: VentureBeat / Bloomberg)
5｜Meta's AI Agent Goes Rogue for Two Hours, Then Signal Founder Steps In
Last week, a Meta engineer used an internal AI agent to analyze a technical issue. The agent, without authorization, autonomously posted a reply on the company's internal forum. Another employee followed its incorrect advice, leading to a nearly two-hour exposure of sensitive company and user data, classified as a Sev 1 event by Meta internally. According to VentureBeat, the 2026 CISO AI Risk Report revealed that 47% of CISOs have observed unauthorized behavior by AI agents, with only 5% confident in their ability to contain a rogue agent.
In response, Signal founder Moxie Marlinspike is in the process of integrating the technology of his encrypted messaging app Confer into Meta's AI. On the same day, the White House is expected to submit an AI governance framework to Congress on Friday. What seems like an internal security incident is rooted in the transformation of the AI agent from a conversational window to a privileged actor within the system, while corporate security architecture still operates under the assumption of "human-initiated actions." This gap is being addressed simultaneously from both a technical (encryption) and policy (regulatory) standpoint.
(Source: The Verge / VentureBeat / Wired / Axios)
Also Worth Knowing ↓
Cloudflare CEO Matthew Prince predicts that by 2027, internet bot traffic will surpass human traffic. Generative AI agents are transitioning from being "users" to the primary "residents" of the internet, necessitating a reevaluation of infrastructure load models. (Source: TechCrunch)
Xiaomi's Lei Jun announced a future three-year investment of over 60 billion yuan in AI. China's tech giants' AI arms race is expanding from internet companies to hardware manufacturers. Xiaomi is simultaneously betting on AI and automobiles, with financial pressure poised to be the next challenge. (Source: 36Kr)
DoorDash launches Tasks app, paying couriers to record videos to train AI and robots. The gig economy is giving rise to a new category—it's not about delivering goods, but rather feeding data to AI. (Source: TechCrunch / Bloomberg)
Micron quarterly EPS of $12.20 was 1.4 times the analyst expectation of $8.66, but the stock still dropped by 5.6%. The chip company's fundamentals couldn't withstand macro triple pressure (Hawkish Fed + Oil Prices + PPI), and "beat but drop" is becoming the main theme of this earnings season. (Source: Barron's)
Tesla's FSD faces an expanded NHTSA investigation, potentially leading to a recall. The investigation focuses on the safety performance of FSD under poor visibility conditions. On the same day, the All-In Podcast aired an exclusive interview with Andreessen Horowitz co-founder Marc Andreessen, where he discussed the $5 trillion market for physical AI and the future of OpenClaw, unfolding the dual narratives of autonomous driving. (Source: The Verge / Reuters / All-In Podcast)
Uber pledges up to $1.25 billion to Rivian to build a Robotaxi fleet. The ride-hailing platform is shifting from "hiring drivers" to "buying cars to run on its own," and Rivian has secured a potentially life-saving major customer. (Source: FT / Bloomberg)
