According to the Pragmatic Optimist newsletter, 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 dropped close to 10% last month and now sits almost 20% down from its June high. Even robust, beat-and-raise Ers from TSMC $TSM and ASML $ASML 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, the Pragmatic Optimist newsletter portfolio strategy delivered.

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

This is because, amidst the euphoria of June 2026, they warned investors about the impending volatility and alerted subscribers about trimming their multi-baggers in companies such as Marvell MRVL , AMD AMD , Micron MU , Astera Labs ALAB , etc. As the euphoria turned into peak exuberance, they amped up their warnings about an impending volatile correction in expectations, especially after updates from semiconductor’s newest bellwethers failed to excite investors further.

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

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.

First, enterprise AI budgets are getting rationalized as global organizations scrutinize the ROI vs. Rapidly escalating AI expenses. The H2 Market Outlook of June 2026 provided anecdotes from companies that have cut back on the ‘tokenmaxxing’ trend in a drastic attempt to rationalize AI budgets for the entire organization. The support of this anecdotal views is 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 the views from June 2026 as to how enterprises were moving to cut inference costs.

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.

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. The Pragmatic Optimist newsletter took these signs at face value and alerted their subscribers they would be trimming their Micron position to protect their triple-digit gains from volatility.

Third, it was 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, the Pragmatic Optimist newsletter 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.

To conclude, if the above fears continue to exist in markets, there better be a solid reason why the Pragmatic Optimist newsletter is just about turning bullish on stocks.