Authors: Victor, Mr. Z, 168X

"AI is the largest infrastructure project in human history. Currently, hundreds of billions of dollars have been invested globally, but this is just the beginning; trillions of dollars worth of infrastructure remain to be built." — Jensen Huang, 2026 Davos Forum

For the first time in forty years, semiconductors are directly contributing to production, leading to a complete shortage across the supply chain. TSMC's orders are already booked until after 2030. Wall Street is still using outdated valuation models to call it a bubble—this is our cognitive bias.

The "Three-Horse Chariot" is a cross-market allocation framework proposed by 168X: US stocks, Taiwan stocks, and cryptocurrencies . It constructs a nested barbell structure around three leading stocks ( NVIDIA, TSMC, and BTC ) to systematically invest in AI and cutting-edge technologies. The core discipline is singular: all trades are for accumulating more of these leading stocks.

If you, like us, firmly believe in the "second phase" of semiconductors, and are not content with just market returns but seek high-growth, high-explosive excess returns, then the three pillars are the exclusive investment framework designed just for you in 2026.

I. Semiconductor Phase Two Preparations to Commence

Quoting Herman Jin, former Goldman Sachs FICC executive and founder of Zen Family Office, in an interview on 168X ( related reading: Interview with former Goldman Sachs FICC executive: Semiconductor shortages benefit those who catch up! Buy as many optical modules as you can ):

"For the past four decades, semiconductors have driven PCs, mobile phones, and the cloud, but they have never directly generated revenue. Humans develop software on semiconductors, and the users of that software generate productivity. But starting with this generation of AI models, semiconductors are directly creating productivity."

This trend first emerged in early 2024 and was then confirmed at TSMC's Q1 earnings call in 2026 with a CapEx of $52-56 billion. As Jensen Huang stated, the infrastructure layer is organizing tens of thousands of processors into a token factory, directly "manufacturing intelligence." Computing power equals production capacity, and production capacity equals revenue.

After our discussions with Herman, we are even more convinced that semiconductors have only just completed the first phase, and the key now is how to bet on the second phase.

The first stage was "discovery." The market realized there was real demand for AI, and NVIDIA's stock price rose from $12 to $150, while the valuations of model makers and private equity firms soared. TSMC's measured expansion over the first three years suppressed the bubble, resulting in high profit margins for all manufacturers and a surge across the entire semiconductor supply chain.
The second phase is "repricing," which is now. This second phase of the three-horse race's bet is built on three core concepts:

I. During the AI infrastructure construction cycle, the semiconductor supply chain is in dire need of everything.

According to Jensen Huang's "AI Five-Layer Cake Theory": energy, chips, infrastructure, models, and applications, every successful AI application will pull down demand to the next five layers. We can see that the semiconductor shortage has spread from chips and advanced packaging to various components and assembly plants, causing significant price increases throughout the supply chain. TSMC's orders are already booked until 2030, and the shortage is expected to continue for a long time.

II. Perceived interest rate spread: Wall Street has not yet grasped the importance of CapEx.

In 2026, the combined CapEx of the four major cloud giants exceeded $700 billion, a 77% annual increase, but Wall Street has not yet caught up. For the past few decades, Wall Street's valuation logic has been "asset-light, high-margin, low CapEx," and a decline in cash flow is instinctively seen as a bubble. But AI has turned companies into token factories; without CapEx, there is no production capacity, and without production capacity, there is no revenue. When valuation models catch up with the industry's actual understanding, companies that are currently investing in production regardless of cost will be significantly revalued—this is the second stage of alpha.

Third, Anthropic and OpenAI are not yet available on the market.

Anthropic and OpenAI's private valuations are approaching one trillion dollars, and their model revenue continues to rise. Anthropic's annualized revenue per unit (ARR) has increased tenfold in the past 12 months, reaching $19 billion, while OpenAI's ARR surpassed $25 billion in February of this year. The IPOs of these giants will pose a significant liquidity drain risk to the market, but their strong revenues mean they may postpone their listings. In other words, the risk of the entire industry being undervalued is relatively low before these giants go public.

II. Why traditional investment methods will be insufficient in 2026

As we navigate the second phase of AI and semiconductors, how should we invest to capture more excess returns in the Beta phase?

There are several mainstream investment methodologies in the market:

1. Index investing : Long-term holding, regular fixed-amount investment in SPY or QQQ, etc. For general investors who are unwilling or do not have time to actively research individual stocks, investing in the overall market remains the best option.

However, the cost of index investing is that you only receive beta returns. This year, any stock in the semiconductor supply chain could significantly outperform the broader market. By 2026, Taiwan's stock market, with its high semiconductor sector concentration, will surpass the UK and Canada, becoming the sixth largest in the world by market capitalization.

2- High-growth investing : All bets are placed on disruptive innovation targets with high valuations and high volatility. The most iconic example is Cathie Wood's ARKK, which returned 153% in 2020 during the pandemic, but then fell by more than 60% in 2022, showing extreme volatility.

Since the end of 2019, ARKK has seen a cumulative increase of 56.3%, far lower than QQQ's 233% increase during the same period. Amid the recent AI boom, ARKK's holdings are diversified across software, biotechnology, fintech, and consumer goods, resulting in a significantly lagging performance.

3. Traditional barbell strategy : 90% allocated to ultra-low-risk assets (government bonds, cash), and 10% allocated to high-risk speculative positions. Theoretically, it can resist black swan events and capture asymmetric opportunities. However, its proposer, Taleb, also admits that this method cannot "make you rich," but only "prevent you from becoming poor."

Our simple trial calculation:

Let's say you have 1 million in principal. You invest 900,000 in government bonds with an annualized return of 4%, and 100,000 in higher-risk investments such as small-cap stocks or cryptocurrencies. The government bonds will give you 40,000 a year, and the high-risk investments could double to 100,000, or even go to zero. The overall return is 140,000, an annualized return of 14%. It looks good. But if you had directly bought QQQ, which tracks the broader market, during the same period, and QQQ's stock price increased by 20.77% in 2025, your 1 million would have become 1.2 million.

Traditional barbells underperform the index in a bull market, and unless you have a very large amount of capital, the absolute return of traditional barbells is simply not enough.

Is there a way to capture the second phase of AI and semiconductors that offers both the asymmetric upside potential of a barbell but doesn't sacrifice returns due to excessive conservatism on the low-risk side?

III. The Three Driving Forces: US Stocks, Taiwan Stocks, and Bitcoin Allocation Strategy

2026 is the "Bingwu Horse Year" with extremely strong fire element, traditionally regarded as a period of transformation full of changes and great upheavals.

Looking ahead to the second half of 2026, 168X proposes the "Three-Horse Chariot" investment strategy, allocating across three markets—US stocks, Taiwan stocks, and Bitcoin—to help investors weather the changes and turbulence in the global situation this year and capture excess returns above beta.

The "three-horse carriage" strategy can be seen as an aggressive version of the barbell strategy, replacing low-risk bonds with leading stocks in the three major markets.

First batch of stocks: US stocks (45%), the world's strongest AI ecosystem and cutting-edge technology

Configuration: faucet $NVDA + 3-5 high-growth stocks

NVIDIA is a core holding in the US stock market for the three major AI companies. NVIDIA doesn't just sell chips; its strength lies in its entire ecosystem, and every pitfall it has encountered has become a moat. Buying NVIDIA is like buying the right to collect rent from the entire AI ecosystem.

Besides NVIDIA, the US stock market boasts the most comprehensive portfolio of AI and cutting-edge technology stocks globally, including AI cloud infrastructure, storage, and optical communications, which we are closely watching, as well as space technology and robotics, which are likely to be the next big sectors. These are the frontier battlegrounds for AI's expansion.

Second batch of stocks: Taiwan stocks (45%), the world's strongest semiconductor supply chain.

Configuration: faucet $TSM + 3-5 high-explosive stocks

TSMC (US stock code $TSM , Taiwan stock code 2330) is the world's leading semiconductor foundry. Regardless of which model manufacturer or chip provider wins in the AI field, they will still have to rely on TSMC for production. TSMC's supply shortage has also driven the expansion of the entire supply chain, with shortages in advanced packaging, IC design, passive components, test interfaces, heat dissipation, and more.

The downstream segments of the semiconductor supply chain are dominated by small and medium-sized Taiwanese manufacturers, many of which are unique to the Taiwanese stock market. These local companies are usually not favored by international investors, but during periods of shortages, their revenue and stock prices have surged beyond expectations.

We remain bullish on Taiwan. Until the end of 2026, Taiwan will be in a state of triple resonance of industrial advantages, favorable policies, and capital momentum.

The third batch of stocks: Crypto (10%), a catcher of macro liquidity spillover.

Configuration: faucet $BTC

Bitcoin is a hard currency with scarcity and censorship resistance, and there is genuine demand for it in the gray market. On the other hand, Bitcoin is extremely sensitive to macro liquidity. When global liquidity is abundant, the excess funds often flow into Bitcoin, and once it starts to rise, its explosive potential is extremely strong, but it needs to be held for years.

This year, liquidity is concentrated in AI and semiconductors, and the overall crypto market is in a bear market cycle. Therefore, we will only allocate to BTC and will not allocate to any other coins until we see clear bull market signals.

IV. The structure of the three-horse carriage: a large barbell nested within a smaller barbell.

The underlying architecture of the three-horse carriage is a nested barbell structure .

Big Barbell: Taiwan and US Stocks (90%) vs. Bitcoin (10%).

On one end are stocks of companies with solid fundamentals and verifiable financial data. On the other end are cryptocurrencies with explosive growth potential, driven by liquidity.

Small barbell: Leading stocks vs. satellite stocks within each market.

Allocate approximately three to five satellite stocks around the leading stock, selecting targets with high growth potential and distinct themes:

Conservative (Levels 3-4): Stocks in sectors with medium- to long-term supply chain logic, which can be held for several months .
Explosive type (levels 1-2): Betting on short-term catalyst events, high volatility, held for only a few days to a week , with extremely rapid rotation.

Taking US stocks as an example, the core holding is in NVIDIA. In the Taiwan stock market, TSMC is the core asset, and satellite stocks cover those benefiting from shortages in important sectors such as advanced packaging, heat dissipation, and optical communication, as well as thematic stocks such as low-Earth orbit satellites and robotics.

As the only asset in crypto, BTC does not require an internal barbell; it is itself the high-volatility end of a large barbell.

Large barbells are used to manage cross-market risk, while small barbells are used to capture the rotation rhythm within the market.

V. The core discipline of the three driving forces: hoarding leading stocks.

The core operating principle of the three driving forces is: all transactions are for the purpose of accumulating more leading stocks.

Even if the leading company has a very large market capitalization, its explosive power can still be very strong when it starts to rise.

Taking the US stock market over the past six months as an example, when the leading stock NVDA was fluctuating at high levels, funds rotated to other sectors, with memory stocks surging and optical communications stocks being ignited. After the entire semiconductor sector had risen, NVDA finally broke through to a new high with a massive market capitalization of 5 trillion dollars during the Trump-Xi meeting in May.

This logic has been repeatedly validated in multiple Crypto cycles: "The purpose of all altcoins is to exchange them for more BTC." Bitcoin's market capitalization is as high as 1.5 trillion, but between 2023 and 2025, only a handful of altcoins outperformed Bitcoin.

The three-pronged approach executes the core logic of "accumulating leading stocks" across three markets. When short-term profits are realized in satellite stocks, we will reinvest those profits into the leading stocks at appropriate prices for the long term. The leading stocks are the cash reserves of the three-pronged approach.

Let's illustrate this logic with a simple trial and error:

Let's say you allocated $1 million to US stocks six months ago, with an initial allocation of $700,000 in NVDA and $150,000 each in satellite stocks A and B. After one round of trading, stock A rose by 80%, so you took profits and got $270,000 back. Stock B rose by 40%, so you got $210,000 back. All of your principal and profits were then poured into NVDA.

After one round, your NVDA holdings increased from 700,000 to 1,180,000. Then, NVDA rose by 11% in May, and your total assets became 1,309,800, achieving a performance of over 30% in just six months.

Even if NVDA itself is just consolidating sideways during this period, after one or two rounds, the position of the leading stock will be so large that your overall portfolio will be highly correlated with the trend of the leading stock, which is exactly the result we want.

The big get bigger, and the strong get stronger.

VI. Timing Window and Suitable Investors

Every investment framework has an expiration date; there is no one-size-fits-all investment method. We assess that the three pillars of investment strategy will be applicable until at least the end of 2026.

We believe the three-horse chariot has three failure conditions:

1. Hyperscaler CapEx significantly reduced. Any major cloud provider that significantly reduces AI-related capital expenditures or postpones construction plans in its financial report indicates a crack in demand.
Second, the supply chain shifts from shortages to inventory accumulation. When key links such as HBM and CoWoS experience inventory buildup or shortened delivery times, supply and demand reverse, and the valuation logic shifts from "shortage premium" back to "inventory cycle".
Third, the commercialization of AI applications lags significantly behind infrastructure investment. The logic of the five-layer pie is top-down pull. If the application layer cannot generate sufficient revenue to justify the investment in the underlying layers, the market will question the return on investment. Special attention needs to be paid to the IPO timing of Anthropic and OpenAI; the high expectations after their listings will certainly face scrutiny.

When any of these three events occur, a signal will be triggered indicating that the strategy may no longer be applicable.

The three pillars represent a highly concentrated, high-growth portfolio, with 90% of the holdings invested in the AI ecosystem and semiconductor supply chain. The remaining 10% in BTC represents an upside opportunity due to liquidity spillover. Holding periods for satellite stocks can range from several months to as short as a few days, requiring active management and disciplined execution.

This is not a buy-and-forget strategy, nor is it suitable for completely passive investors. The three-pronged approach is suitable for active investors with relatively small amounts of capital who firmly believe in the second phase of AI and semiconductors and seek high growth and high returns.

VII. The Timing for Starting the Three-Horse Chariot in the Year of Bingwu (1946)

2026, the Year of the Horse (Bingwu year), is a year of great upheaval and transformation. However, the "three driving forces" (referring to three key sectors) are not a strategy to rush in with all your funds today.

As of this writing (May 19th), short-term sentiment in the semiconductor sector is overheated, with excessive crowding of shares. The market is optimistic, and themes such as optical communications are experiencing irrational surges. "Sell in May and go away" is very likely happening, and excessive consensus is often a precursor to a short-term correction.

We believe the market is likely to experience a reset within one to two months. This reset will be the true starting point for the second phase of the semiconductor sector and the most important opportunity to build positions in the three growth drivers.

The "three-horse carriage" strategy is more than just an investment theory. The 168X team will continue to share our practical experience in the coming period, including specific targets, sector analysis, entry and exit timing, and rotation discipline.

All the transactions are aimed at accumulating more leading stocks—NVIDIA, TSMC, and Bitcoin.

In the Year of the Horse (Bingwu year), three powerful forces will bring continuous prosperity.

Galaxy Digital's linked wallet purchased 158,100 HYPE tokens in the past two hours, worth approximately $8.8 million.

Popular Articles