8 billion capex for 2026, an increase of 35% from previous forecasts, specifically targeting expanded HBM fabrication lines.
Critical foundry capacity remains a choke point. Taiwan Semiconductor Manufacturing Company (TSMC) is forecast to allocate approximately $42 billion in capital expenditure for 2026, primarily for advanced node expansion (3nm and 2nm), essential for next-generation AI processors from Nvidia and AMD. ASML Holding's (ASML) order backlog has swelled to over €40 billion, with lead times for its cutting-edge EUV lithography tools extending to 18-24 months. These prolonged lead times and significant capex requirements across the industry mean that the supply-demand imbalance for high-performance AI components is unlikely to ease materially before late 2027. Global semiconductor industry capex is projected to exceed 80 billion in 2026, a +15% increase from 2025, but it still struggles to keep pace with demand.
Macro valuation metrics reflect this scarcity and strategic importance. Institutional capital flows into AI-related technology funds totaled
4.2 billion in net inflows during Q3 2026, representing a +25% increase from the prior quarter. This sustained influx of capital has pushed the average Enterprise Value to EBITDA (EV/EBITDA) multiple for pure-play AI infrastructure providers to 32x, compared to a broader tech sector average of 18x. While elevated, these valuations are underpinned by strong projected revenue growth rates exceeding +60% annually for the next two years for market leaders. Investors are clearly willing to pay a premium for exposure to companies at the forefront of the AI revolution, recognizing the significant long-term secular tailwinds provided by innovations like autonomous agentic networks in finance.
Quantitative Order Flow & Volatility Metrics
Quantitative analysis of options order flow reveals a distinct bullish bias towards key AI infrastructure plays. Nvidia's (NVDA) average daily options volume soared to 3.5 million contracts over the past week, with a notable 65% of that volume consisting of call options. The 25-delta risk reversal, a measure of call skew relative to put skew, for NVDA stood at an elevated +12.5% for 3-month options, indicating that out-of-the-money calls are significantly more expensive than equivalent puts. Super Micro Computer (SMCI) exhibited similar patterns, with its 25-delta risk reversal at +10.8%, fueled by expectations of robust server demand for AI clusters.
This strong options market sentiment is mirrored in broader market indices. The SOX Semiconductor Index (SOX) climbed +3.5% yesterday, reaching new all-time highs, largely on the back of positive earnings and guidance from its constituents. On the institutional side, block trade data indicated net buying of $8.5 billion across NVDA, SMCI, and ASML over the last five trading sessions, underscoring conviction from large asset managers. Concurrently, there has been significant put selling activity, particularly in Taiwan Semiconductor Manufacturing Company (TSM), with approximately $750 million in notional value of out-of-the-money puts sold for December 2026 expiry, suggesting investors are comfortable taking on downside risk in exchange for premium, implying a belief that significant downside is limited.
Implied volatility for NVDA 3-month options hovered around 45%, while the broader tech volatility index (VXN) maintained a level of 18.5, indicating that while NVDA is perceived as a higher-beta play, the market is not pricing in extreme downside scenarios for the AI leaders. The concentrated nature of call buying and put selling reinforces the market's belief in continued upside for companies powering agentic AI, with particular focus on the computational backbone rather than pure software plays, reflecting the foundational demand generated by these new execution paradigms in financial markets. The KOSPI index's strong +2.1% performance was directly influenced by institutional foreign buying in large-cap tech, specifically Samsung Electronics and SK Hynix, as global capital chases the HBM narrative.
Quantitative Outlook
The confluence of increasing demand for agentic AI compute, persistent supply chain constraints for advanced semiconductors, and robust institutional capital inflows suggests a sustained period of elevated revenue growth and margin expansion for key hardware and cloud providers. The present market dynamics, as evidenced by Nvidia's +7.8% surge and SK Hynix's +9.1% gain, are rooted in fundamental shifts in enterprise technology adoption. The substantial investments, exemplified by the
8.5 billion from quant firms and the $42 billion capex from TSMC, underscore a structural demand exceeding current supply capabilities.
The current valuation premiums, with NVDA's forward P/E at 55x and AI infrastructure EV/EBITDA multiples at 32x, reflect these dynamics, indicating that the market has largely priced in strong forward expectations for these companies. The options market data, including the +12.5% call skew for NVDA, further reinforces this bullish sentiment, suggesting conviction among sophisticated investors that the growth trajectory for AI enablers remains firmly intact. Future shifts in this picture would likely stem from either a significant and unexpected slowdown in enterprise AI adoption, a rapid and unanticipated oversupply of high-bandwidth memory or advanced GPUs that eases pricing power, or a substantial deceleration in global macroeconomic growth that impacts corporate IT spending more broadly than currently anticipated.
Tags: Memory Chips, SK Hynix, Semiconductors, Wall Street, AI Trading, Nvidia, Cloud Computing, Quantitative Finance
