美股投资young

美股科技投资者 ai狂热者分享每日快讯. all in ai

303Following

759followers

Feed

Livestream

Feed

The more seriously Google and Amazon develop their own chips, the more willing I am to study TSMC. Everyone is competing over which type of chip will win, but no matter how the design route changes, the product must ultimately be manufactured stably. Participating in AI investment doesn't necessarily mean you have to bet on the architecture that wins in the end every time. What attracts me to $TSM lies in the manufacturing link. Customers can design chips, but between that and large-scale delivery lie yield, capacity, packaging, and production experience. Customers aren't just buying the name of an advanced process; they need to calculate the actual cost of qualified chips and know when they can get the products. No matter how good the technical specs are, if production is unstable and delivery can't keep up, the subsequent computing power deployment will be delayed. This is why I value TSMC's manufacturing integration capabilities more. AI products increasingly rely on the connection between computing chips and high-bandwidth memory. Advanced packaging technologies like CoWoS are also part of whether the product can be delivered smoothly. When process and packaging work well together, customers can turn design performance into actual products. This kind of accumulated experience is hard to replicate immediately just because competitors announce capacity expansions. Google makes TPUs, Amazon makes its own chips, and NVIDIA continues to push GPUs. They compete over who is more efficient for different tasks. For core manufacturing companies, customers changing design routes doesn't automatically mean demand disappears. As long as performance, power consumption, and mass production requirements continue to rise, there is still a basis for competing for business in manufacturing. Of course, not all self-developed projects can be directly counted as TSMC orders; which process to use and who produces it still depends on the specific product. The construction disputes brought by the Democratic Party can't completely isolate TSMC either. Delays in data center deployment may cause customers to adjust chip procurement, which further affects wafer and packaging demand. But compared to betting on whether a single computing power operator can complete several projects, I prefer a manufacturing leader that can serve different customers and participate in different product routes. My long-term judgment doesn't need to be based on one cloud provider completely defeating all others. So TSMC is a company I am willing to study independently, for reasons more concrete than "AI will continue to grow": the more customers want to make more efficient chips, the more they need reliable manufacturing and delivery. Geopolitical risks, overseas expansion costs, and purchase valuation all need to be considered. I won't ignore price just because of strong manufacturing status, but I also won't equate a single data center approval dispute directly with the loss of value in its years of accumulated manufacturing capabilities.

Even if the Democratic Party loosens restrictions on data center construction later, it doesn't mean all semiconductors can rebound to their original levels. During the waiting period for project approvals and power supply, chip products continue to update, and customers will recalculate their costs. The generation of equipment originally planned for purchase—when the data center finally becomes usable—will it still be worth buying as planned? This will determine who just earns money a bit later and whose business might be taken by others. What gives me more confidence in $NVDA is that when customers buy it, it's hard to just compare the price of a single GPU. Software, networking, development tools, and already adapted applications are all part of the complete system. Switching suppliers takes time, and re-tuning costs money. For customers eager to launch their services, these are reasons to continue choosing NVIDIA. But project delays might also make customers reconsider: continue buying mature systems or wait for the next generation? Long-term demand hasn't disappeared, but the timing of revenue realization might change. Therefore, I am optimistic about NVIDIA in the long term but won't treat every long-term procurement plan as confirmed revenue. If the company continues to iterate and customers ultimately choose to upgrade, the business still has a chance to stay; if customers keep waiting, old product purchases slow down first, and new products can't be deployed temporarily, the intermediate quarters might still be tough. This is the scenario I consider when studying the impact of delays, not to be directly interpreted as current order changes. $AMD should not be seen merely as a stock that follows NVIDIA's ups and downs. Customers reevaluating purchases might create opportunities for other suppliers, but whether those opportunities turn into business depends on actual task performance, software adaptation, and whether the entire system can run stably. A few more months of comparison time might help competitors or might make customers more certain to continue using familiar solutions. We cannot automatically infer that AMD benefits just because NVIDIA faces policy resistance. What I want to see more is, after political tensions ease, what customers actually resume executing. Are projects restarted, are procurement plans updated, and have delivery and revenue expectations improved? If only the stock price rises first but operations haven't progressed, I won't rush to count all the good news for the coming quarters. I remain optimistic about AI's long-term direction, but after political impacts lessen, product competition will not pause. I tend to favor leaders with a customer base and the ability to continuously deliver new products because they are better positioned to retain business delayed by postponements. As for companies that rely solely on industry-wide expansion to claim growth, they will need to be scrutinized more carefully than before.

Jensen Huang has admitted to marital infidelity, but "cheating" and "leaking secrets to China" are not equivalent. This matter has gone beyond ordinary gossip; NVIDIA's board has launched an independent investigation, and Congress has also requested a briefing. $NVDA To put it bluntly, with the current situation, big brother A company, are you still going to IPO? How much blood can your IPO suck when there’s hardly anyone left? Honestly, maybe postpone it to save me from being drained. The issue first surfaced from a chip export investigation. In 2025, The New York Times reported that Singapore data center company Megaspeed was under export control investigation, and its executives had contact with Jensen Huang. The question then was whether advanced chips bypassed restrictions to flow to China. Business gatherings and photos can show contact between parties but do not prove extramarital affairs, let alone espionage. By October 30, 2026, The New York Times, citing anonymous sources familiar with surveillance, reported that U.S. officials believed this executive was passing information to Beijing and maintained close contact with Jensen Huang. The so-called "spy" accusation points to this businessperson; Jensen Huang was described in the report as a target approached by foreign intelligence operations, which cannot be directly rewritten as "Jensen Huang is a spy." The report also acknowledged that publicly it remains unclear what information was passed and its actual value. The Chinese side denied the allegations, and NVIDIA responded that customers only had access to ordinary commercial information. On November 5, chairmen of three U.S. House committees sent a letter requesting Jensen Huang to attend a closed-door briefing to explain the contacts and NVIDIA’s counter-espionage measures. It’s important to clarify: the letter was sent by Republican lawmakers, so don’t attribute this matter to the Democrats and semiconductors just because they have been discussed recently. Congress requesting clarification is different from having already determined illegality. On November 8, there was a significant development confirmed by Jensen Huang himself. NVIDIA’s official website published a letter from him to employees, in which he admitted to being unfaithful to his wife and apologized, stating he would cooperate with the investigation. However, the letter did not name the other party nor admit to leaking secrets. On the same day, the board announced that an independent board committee and external independent lawyers would investigate; the board currently sees no reason to believe national security or company secrets have been compromised and continues to support his leadership. This statement carries weight, but the investigation has just begun and should not be taken as concluded. From an investment perspective, I would not infer that GPU demand suddenly disappears because of this apology letter. Customers’ purchasing decisions depend on performance, software, cost, and delivery, none of which automatically change due to exposure of private matters. But saying "it’s just private life" is not enough to move on. If private relationships involve customers, whether there is a conflict of interest, whether information access crossed boundaries, and whether internal oversight failed are all issues shareholders should care about. Jensen Huang is also a core figure in NVIDIA’s product strategy, customer relations, and external communications; if this affects his duties or management arrangements, the market will naturally reassess. My long-term AI judgment does not rely on defending Jensen Huang personally. At present, unconfirmed leaking accusations should not be treated as facts, nor should potential export restrictions or customer reviews be considered already implemented. What will truly change investment judgments is what the investigation finds, what measures the board takes, and whether regulators take actual action. Based on current materials alone, it cannot be concluded that orders have been damaged; likewise, since the investigation is not yet clear, it is not appropriate to call any possible decline a golden opportunity.

Google is still working on how to build the data centers, which is one of the reasons I haven't turned long-term bearish on AI due to political controversies. Companies are not just waiting for policy relaxations; they are also adjusting power supply, cooling, and local infrastructure. The resistance to construction is far from the issue of customers no longer needing computing power. In the Wyoming project announced by $GOOGL on 11/5, it committed to covering all its own electricity costs and the infrastructure demands directly caused by growth. It plans to use air cooling to reduce local water resource pressure and is preparing to form a community committee with partners. What has been announced so far is a plan, and there is still a process before actual operation, but Google has already started answering residents' most direct questions: who pays the electricity bill, how water is used, and what the project will bring to the local area. Its willingness to increase investment to solve these issues at least shows it still wants to continue construction. Another important reason I am optimistic about Google is that it has existing business to absorb computing power. Search, cloud services, models, and self-developed chips can be arranged together. After construction is completed, it can serve its own products and also provide services to cloud customers. The company can compare different architectures and adjust deployment order based on business needs, without relying solely on tenants that may appear in the next few years to prove the project's value. With higher construction requirements, this existing business and adjustment capability will give it more options than a single project developer. Of course, Google bearing more costs itself is not only beneficial to shareholders. My view is that if these investments can secure reliable power supply, usable facilities, and ultimately support services that customers are willing to pay for, the increased costs are acceptable; if costs keep rising but utilization and revenue lag behind, then the returns need to be recalculated. This project strengthens my confidence in Google's willingness to build and ability to solve problems, but it cannot be used directly to prove the stock price is already cheap. $NVDA is also related to the progress of these projects. Power supply, site, and construction conditions need to be in place before subsequent hardware deployment can be more grounded. Compared to customers with only huge long-term plans, I prefer to look at major customers who are already solving actual deployment problems. But Google has its own TPU, so the entire campus investment amount certainly cannot be counted entirely as Nvidia orders. For Nvidia, it also depends on which tasks customers choose GPUs for, and whether CUDA, interconnect, and system efficiency can help it maintain procurement. These specific choices will ultimately determine whose revenue the industry investment turns into. Therefore, in this political market, I still lean toward leaders who have customers, operational foundations, and the ability to complete projects. The approvals, power, and community situations in Wyoming and Texas differ; an investment announcement in one place does not mean the national issues are resolved, and a blockage in one place does not mean all locations stop. Companies that can change plans, adjust sites, and bear reasonable costs still have ways to move construction forward. AI is a long-term competition, but actual projects have to be completed one by one. I do not expect all controversies to end quickly, nor do I need to rely on the Democratic Party making concessions to maintain a bullish view. As long as these companies continue to advance power supply, facilities, and service deployment, progress should be factored into judgments, and a few construction halt news items should not be used to conclude the entire industry. The leaders who can deliver actual results later are the ones I am more willing to spend time researching.