Trump Q1 Holdings Reveal Shift Toward AI Infrastructure

The market's top "signal kings," one with the surname Chuan and one with the surname Huang, are both increasing their bets on semiconductors and the next tech theme.

Written by Mike, Frank, MSX Maetong

Since 2025, two men's trade signals have been the most effective in the market.

One is Jensen Huang; whenever he takes the stage at a launch event to speak about GPUs, Blackwell, or data centers, the market reimagines the ceiling of AI. The other is Trump, whose public statements and policy actions—beyond simply naming specific stocks—have a profound impact on the expectations of entire industries.

Interestingly, recently Trump also legally filed his personal financial disclosures with the Office of Government Ethics, including the stocks, funds, transaction records, and value ranges he holds. Although the disclosure documents do not prove that each transaction was personally decided by Trump, nor can they be simply interpreted as clear buy or sell recommendations, they at least provide a window for observation:

When a person with significant policy influence begins to make clear directional adjustments to their related accounts, the market naturally wonders what industry insights this may reflect.

However, after a thorough analysis, MSX found that the most noteworthy aspect of this Q1 disclosure is the increased trading activity in Trump-related accounts, with a clear shift toward AI infrastructure—particularly through significant reductions in older platform technologies and defensive assets, while increasing investments in AI infrastructure supply.

Undoubtedly, as the ultimate decision-maker on U.S. policy, his portfolio structure reflects, to some extent, his judgment on future industry directions, offering ordinary investors a window into what the world’s most powerful "smart money" is thinking.

If you look at the most straightforward data first, you'll find it's a prime example of "active trading."

According to disclosed filings, Trump-related accounts executed 3,711 securities trades in Q1, which roughly equates to dozens of trades per trading day; based on the lower end of the reported range, the total trading volume exceeded $220 million—clearly indicating an active account, approaching the quarterly trading volume of a small to mid-sized hedge fund.

More interestingly, this differs significantly from his investment style during his first term (2017–2021), when disclosures showed he held around 100 individual stocks across sectors such as finance, healthcare, and industry, resembling more of a diversified blue-chip portfolio. After entering the White House, he transferred his assets to family members and related institutions; his individual stock holdings significantly decreased, and his active trading behavior was not as pronounced as it is now.

Notably, previously Obama invested in Treasury securities and diversified mutual funds, while Biden did not trade stocks at all during his term. Past presidents generally chose to divest assets or establish blind trusts to avoid conflicts of interest, but Trump’s approach in his second term completely broke this tradition.

Looking more closely, you can see a highly thematic set of adjustments.

First, look at where the funds are leaving from.

In the first quarter, the largest sales from Trump-related accounts were concentrated in Microsoft, Amazon, and Meta. According to the disclosed ranges, these transactions all reached the highest tier of $5 million to $25 million. These three companies are undoubtedly still core assets in U.S. tech stocks, but they also share a common trait—they represent the super winners of the previous era of consumer internet, advertising platforms, e-commerce, and cloud services.

Microsoft has software and cloud, Amazon has e-commerce and AWS, Meta has social networks and advertising systems—they are not without AI stories, and in fact are all major players in AI investment. However, from a portfolio perspective, these companies have already reaped substantial valuation gains over the past few years, so significant share reductions do not necessarily indicate bearish sentiment; rather, they more accurately reflect a strategic reduction in exposure to legacy platform technologies.

Particularly note that the disclosure documents do not completely eliminate these companies; some still show small purchase records. This “large sell, small buy” structure appears more like an active reduction of exposure rather than a complete exit.

Also appearing on the large sell order list are dividend-focused ETFs such as the Vanguard Dividend Appreciation ETF. This indicates that capital outflows are not only coming from established tech giants but also include a portion of more defensive, stable assets.

This is crucial. Selling only Microsoft, Amazon, and Meta, then buying another set of tech stocks would merely represent an internal rotation within the tech sector. But if defensive ETFs are also being reduced, it suggests the portfolio’s overall risk appetite may be increasing, with capital shifting from stable, older-platform assets toward more aggressive industry sectors.

So, where did the money go?

The answer is also clear—semiconductors, AI hardware, enterprise software, consumer electronics, broad-market indices, and certain bonds and preferred stocks.

Buying only NVIDIA would simply be betting on the leader in AI computing power, but what’s more notable in this disclosure is that Trump-related accounts purchased not a single asset, but an entire AI infrastructure chain.

At the first level are semiconductors, including NVIDIA, Broadcom, Texas Instruments, Intel, AMD, Micron, and Marvell—all appearing on buy or accumulate lists. This encompasses GPUs, CPUs, analog chips, memory, and interconnects, covering both the most commercialized AI computing leaders and U.S.-based manufacturers with stronger policy attributes, providing comprehensive chain-wide coverage.

NVIDIA and Broadcom need no further explanation: the former is a core play on AI computing power, while the latter benefits from trends in custom chips, networking chips, and in-house chip development by major cloud providers. AMD represents the narrative of GPU and data center computing alternatives, Micron corresponds to storage demand, and Marvell relates to interconnects, custom chips, and high-speed data transmission.

More interestingly, Synopsys and Cadence are also on the buy list; these companies produce EDA tools—software used for chip design—that average investors might not immediately think of. However, within the semiconductor supply chain, they represent a highly upstream “selling shovels” segment. Almost every complex chip, from design to fabrication, relies on these tools, further demonstrating that this portfolio adjustment is not merely chasing the hottest AI leaders, but rather extending upstream along the semiconductor supply chain toward foundational tools.

The second layer consists of AI hardware and servers, with Dell being one of the most sensitive and widely discussed targets. Disclosure documents show that Trump-related accounts established a position in DELL ranging from $1 million to $5 million on February 10. Several months later, Trump publicly endorsed Dell’s hardware products, after which Dell secured significant government contracts and its stock price strengthened noticeably.

This timeline is sensitive precisely because it follows the sequence of account purchases first, followed by public endorsements, then government procurement, and finally a rise in stock price. From a rigorous standpoint, disclosure documents alone cannot prove a causal relationship between the trades, public statements, and subsequent contracts. However, from a market observation perspective, such transactions naturally attract attention because they align with three highly sensitive milestones: AI hardware, government procurement, and public statements by the president.

Intel represents a different kind of sensitivity; unlike Dell, Intel’s core significance lies not just in business logic but in policy logic. The U.S. government has already decided to make a major equity investment in Intel, and Intel has long been a central target in U.S. efforts to bolster domestic semiconductor manufacturing, supply chain security, and industrial policy (see further reading: Intel’s Moment of Truth: How Chen Liwu Settled His Legacy and Fought for Survival at the ICU Door?). Against this backdrop, Trump-related accounts purchasing INTC multiple times in the first quarter naturally led the market to interpret these moves as highly significant.

NVIDIA represents the commercial winner in AI computing power, while Intel represents the domestic manufacturing foundation that the U.S. government aims to support. Their logic differs, but both point in the same direction: AI infrastructure is no longer just a market theme—it is increasingly being driven by industrial policy and fiscal resources together.

The third layer consists of enterprise software companies such as Oracle, ServiceNow, Adobe, and Workday, all of which also appear on the buy list. Unlike NVIDIA, Dell, and Intel, which provide computing power and hardware, these companies embed AI directly into enterprise workflows. Oracle corresponds to databases and cloud infrastructure, ServiceNow to enterprise process automation, Adobe to creative and marketing productivity, and Workday to human resources and financial management systems.

The logic behind this trend is clear: AI ultimately cannot remain confined to models and chatbots—it must integrate into real enterprise budgets and permeate daily workflows in operations, customer service, marketing, finance, HR, development, and data analysis. Ultimately, the greatest advantage of enterprise software companies is that they are already embedded in their customers’ workflows. Once AI capabilities become default features within these software platforms, the impact won’t just be a new narrative—it could transform retention rates, pricing power, module upgrades, and customer loyalty. (Further reading: Software Stock “Repair” Myth: After the Rally, Is AI Agent a Killer or a Savior?)

So, what’s truly noteworthy in this disclosure isn’t just which AI hardware companies were purchased, but also that enterprise software AI adoption is becoming another important clue.

The fourth layer is consumer electronics—for example, Apple has received significant增持 and multiple follow-up purchases. Compared to pure AI chips and enterprise software, Apple is more representative of an AI terminal entry point. Whether it can truly ride the AI device cycle remains debated in the market, but in a portfolio covering both AI infrastructure and end applications, Apple is an unavoidable super entry point.

Additionally, fifth-layer broad-market indices such as the S&P 500 ETF, Russell 1000 ETF, and QQQ also appear on the large-buy list, indicating that this portfolio is not entirely detached from the broader market or solely betting on a single thematic line, but rather maintains overall exposure to the U.S. equity market while actively increasing positions in AI infrastructure and key industrial chains.

The disclosure documents also include numerous bond transactions, such as municipal bonds, corporate bonds, high-yield bond ETFs, and bank preferred shares, with municipal bonds covering multiple states and corporate bonds including those of Netflix, Occidental, and CoreWeave.

Therefore, from a portfolio perspective, we can create a clear investment self-portrait: maintaining a foundation and liquidity with broad-market indices, bonds, and preferred stocks on one side, while enhancing aggressiveness with semiconductor, server, enterprise software, and AI infrastructure assets on the other.

Upon seeing this disclosure, many people’s first reaction might be whether they should buy accordingly.

But simply copying homework doesn't really make sense, and the reason is simple:

• First, OGE disclosures have a time lag—by the time retail investors see the filings, the trades have already occurred;
• Second, the disclosed amounts are ranges, not exact figures—for example, $1 million to $5 million, or $5 million to $25 million—with significant gaps in between, making it difficult to determine the actual position weights based on this information;
• Third, the relevant accounts may be independently managed by third-party institutions, and it is not publicly known whether each transaction is the result of active judgment, portfolio rebalancing, or model-based allocation;

Therefore, this disclosure should not be used as a signal for short-term trading.

Where it truly holds value is in revealing a larger directional shift: the most astute "smart money" is moving away from legacy platform technologies and certain defensive assets toward AI infrastructure supply, specifically shifting from the previous generation of internet core assets—such as advertising, e-commerce, and traditional cloud services—to chips, servers, storage, interconnects, local manufacturing, and enterprise software AIization.

This direction also overlaps to some extent with the current U.S. policy priorities.

After all, domestic semiconductor manufacturing, supply chain security, AI infrastructure, government procurement, and corporate digitization are not merely market stories—they are directions driven by policy, fiscal support, industry initiatives, and capital, especially for targets like Intel, whose significance goes beyond earnings elasticity to represent America’s desire to regain control over advanced manufacturing and chip supply chains.

This is also the most noteworthy aspect of Trump-related accounts increasing their holdings in Intel: it doesn’t necessarily mean Intel is the best chip stock, but it indicates that, within the AI infrastructure theme, the market currently favors those positioned at the center of policy resources. Similarly, Dell’s case illustrates that AI infrastructure is not limited to the GPU level—servers, hardware, government procurement, and enterprise deployment will all become part of how AI capital expenditures translate into real-world applications.

Therefore, what ordinary investors should truly take away from this disclosure is not any single stock, but three structural insights.

• AI trading is shifting from models and applications to infrastructure: Previously, the market bought AI based on the imagination and computational power expectations of large models; now, capital is increasingly focusing on who can provide chips, servers, storage, networks, packaging, design tools, and enterprise software.
• Semiconductors are no longer just about NVIDIA: NVIDIA remains the core focus, but this disclosure shows that capital is also flowing into supply chain nodes such as Broadcom, AMD, Micron, Marvell, Intel, Synopsys, and Cadence; the deeper you go into AI infrastructure, the less it’s a story about a single leader and more about the repricing of the entire supply chain;
• The AI integration of enterprise software may be the most underestimated aspect: hardware builds the computing power, while enterprise software puts AI into practice. The value of companies like Oracle, ServiceNow, Adobe, and Workday lies not in their ability to tell a brand-new AI story, but in their capacity to embed AI into existing workflows and turn customer retention and product upgrades into revenue.

Regarding the significant sell-offs by Microsoft, Amazon, and Meta, don't simply interpret this as "these companies are going to decline." More accurately, it's a signal of capital reallocation—after all, when established platform giants have already risen substantially, capital naturally begins seeking assets closer to the next wave of capital spending, policy support, and infrastructure development.

Regardless, the era's advantages of consumer internet have not disappeared, but AI infrastructure, semiconductor localization, and the AI integration of enterprise software are indeed accelerating as the next phase of focus for capital.

This is also the most noteworthy aspect of the Q1 portfolio rebalancing disclosures from the world's most powerful individuals.