We’re Not Wasting AI’s Big Selloff

Last month we called the AI fatigue and this week that fatigue turned to panic as key semi stocks sold off globally. Our portfolio remained unfazed through it all, as we began to deploy our cash towards our highest-conviction AI winners who are likely to lead the pack for the remainder of the year.

Join us! We help hundreds of investors navigate the rapidly evolving AI innovation landscape, identify rock-solid businesses with strong growth trajectories & operational grit, and deliver proven alpha-generating returns.

Since March 1, the TPO Portfolio delivered returns of 28% 💪, compared to 14% for QQQ and 17% for AIQ.

The month of June officially sprung a trap on the AI hype cycle.

The arrival of China’s latest 2.8T-parameter Kimi K3 model from Moonshot AI intensified the debate over the AI spending cliff, while enterprises are already grappling with skyrocketing token budgets. This has further elevated skepticism surrounding the durability of the American AI premium, raising questions about the sustainability of the capex that has been central to the AI infrastructure buildout.

Chipmakers took the brunt. VanEck’s Semiconductor ETF SMH 0.00%↑ dropped close to 10% last week and now sits almost 20% down from its June high. Even robust, beat-and-raise ERs from TSMC TSM 0.00%↑ and ASML ASML 0.00%↑ failed to reignite sector momentum, with capex skepticism and the potential AI spending cliff continuing to dominate the investment narrative.

The final blow came from ballooning leverage that broke South Korea’s KOSPI, dragging the world’s best-performing market of 2026 straight into a bear market. And just like that, chip stocks sleepwalked into a trap, sending shockwaves through global markets last week.

But while the market panicked, our portfolio strategy delivered.

At The Pragmatic Optimist, our portfolio remained exceptionally resilient through it all, declining just (3)% for the month of July, compared to (7)% and (14)% drawdowns for QQQ QQQ 0.00%↑ and AIQ AIQ 0.00%↑.

This is because, amidst the euphoria last month, we warned investors about the impending volatility and alerted subscribers about trimming our multi-baggers in companies such as Marvell MRVL 0.00%↑, AMD AMD 0.00%↑, Micron MU 0.00%↑, Astera Labs ALAB 0.00%↑, etc. As the euphoria turned into peak exuberance, we amped up our warnings about an impending volatile correction in expectations, especially after updates from semiconductor’s newest bellwethers failed to excite investors further.

In last week’s tactical note to our subscribers, we mapped out the exact distress-signal levels where we planned to redeploy our dry powder to scoop up our favorite, high-conviction AI names at a steep discount.

In this post, we start with summarizing 3 key fears plaguing the AI trade so far. We compare these fears with structural signals that keep us encouraged about the foundational long-term bullishness in the AI trade.

We will also provide a summary of the stocks that we have already invested in this month and outline our game plan forward, with key technical levels we are watching across our highest conviction AI names.

📌Since March 1, The TPO Portfolio has delivered returns of 28% 💪, compared to 14% for QQQ QQQ 0.22%↑ and 17% for AIQ AIQ 1.72%↑. On a YTD basis, our portfolio is also outperforming the above benchmarks by 500 and 300 bps, respectively.

You can track our entire portfolio and all our live trades in the AI Stock Tracker 2.0 tool using the link below. 👇

Unlock Access to The TPO Portfolio

(Paid Members can access the AI Stock Tracker 2.0 directly from here)

At the moment, the AI trade appears to be in the midst of an intensely volatile period of heightened structural pressure characterized by a sharp succession of 3 broad-based headwinds. We briefly touched upon this in a separate broader macro article on Seeking Alpha. But we will expand this in more detail here.

First, enterprise AI budgets are getting rationalized as global organizations scrutinize the ROI vs. rapidly escalating AI expenses. In our H2 Market Outlook post from last week, we provided anecdotes from companies that have cut back on the ‘tokenmaxxing’ trend in a drastic attempt to rationalize AI budgets for the entire organization. We supported our anecdotal views by referencing Bloomberg’s LLM Token Expenditure index, a benchmark used for pricing model inference tokens (expressed in USD per million tokens). The drawdown in the LLM Token Exp index suggested enterprises were spending relatively less on AI model usage, corroborating our views from last month as to how enterprises were moving to cut inference costs.

We attached an updated chart below that tracks the LLM Token Exp index and compares it to UBS’s Hyperscaler Basket index, illustrating how dependent hyperscalers (and the broader AI infrastructure ecosystem) are on higher AI model spending.

The rationalizing of AI budgets may probably also be diverting a growing portion of AI budgets towards cheaper models, such as the Chinese LLMs like Moonshot’s Kimi K3 and Zhipu’s GLM 5.2. The growing fear is that US LLM companies are losing market share to Chinese LLM companies. Investors are also concerned that cheaper Chinese models would weaken the ROI on AI capex for hyperscalers, which we disagree with and discuss in detail in the next section.

Second, a systemic deleveraging is simultaneously unfolding, in parallel to the pullback in enterprise AI spending, where global portfolios are actively unwinding highly leveraged positions, impacting semiconductors, neoclouds, etc, AI’s largest beneficiaries.

South Korea’s KOSPI index is an example of how violently optimism can turn south in a matter of weeks. As the index reached exuberant levels, bolstered by record volumes of retail debt, the smart money had already begun to quietly hedge portfolios, also at record levels. We took these signs at face value and alerted our subscribers we would be trimming our Micron position to protect our triple-digit gains from volatility.

What we are now left with is the KOSPI index in deep bear market territory while margin calls have hit new records, triggering forced liquidations for hundreds of thousands of leveraged investors.

Upgrade To Premium For Just $13.33/Month

Third, we also noticed that institutional portfolios were actively going through a period of distribution, after locking in one of the best H1 gains ever, likely exacerbating the systematic pressure on the crowded AI trade. The distribution in hedge fund portfolios became more elevated over the past 4-6 weeks, which put pressure on semiconductor stocks, AI’s biggest beneficiary cohort.

So far, we strongly suspect these fears still hold a dominating presence in commanding the short-term narrative that is gripping the AI trade.

Market’s reactions to Taiwan Semi and ASML Holdings’ ERs are a prime example of how these fears continue to linger. Despite both semi-supply chain behemoths giving optimistic capex outlooks supported by strong forward growth (this is usually seen as a bullish sign by markets over the past years), markets are questioning the sustainability of the expansionary capex budgets, doubting whether nominal growth is being driven by real demand or cost inflation.

So if the above fears continue to exist in markets, there better be a solid reason why we’re just about turning bullish on stocks.

There are two things we will outline for our premium subscribers over the next two sections.

In this section, we update investors with the structurally positive signs that are emerging amidst the heightened concerns that now surround the AI trade as noted above. Based on these signs, we have begun to deploy our cash selectively towards some of our highest-conviction semiconductor stocks already. Remember, we had raised cash in June as we trimmed our multi-baggers building up our cash position all the way up to nearly 38% of the portfolio.

In the following section, we update subscribers with our tactical cash deployment plans towards our favorite AI beneficiaries that will likely emerge as winners in the next leg of the AI trade.