AI & Semiconductors

Last updated: July 18, 2026

The semiconductor sector transitioned from extreme volatility to a sharp correction this week as investors reassessed the sustainability of AI valuations. While TSMC reported record second-quarter net income of NT$706.6 billion and raised its 2026 CAPEX forecast to $60–$64 billion, its shares fell after failing to meet aggressive investor expectations. Similarly, SK Hynix suffered a severe correction, with shares plunging up to 15% in Seoul following a Q2 operating profit of KRW 60.4 trillion, which missed consensus estimates by 8%. These losses were compounded by a broader regional tech rout and a global sell-off triggered by the release of China's Moonshot Kimi K3 model.

Strategic divergence is widening between hyperscalers and legacy providers. Apple has detached from the broader sector sell-off, hitting new all-time highs and surpassing Nvidia to become the world's most valuable company with a valuation nearing $4.9 trillion. Conversely, IBM shares crashed 25% after warning that AI infrastructure spending is diverting funds away from traditional enterprise software. This shift is further evidenced by Meta's pivot toward cloud infrastructure, including a potential $10 billion compute lease deal with Anthropic and the hiring of a senior AWS executive to lead its cloud push.

Investment implications are shifting as the market moves toward a "flight to quality" in Big Tech and a fractured landscape for AI unicorns. While SpaceX shares collapsed below their $135 IPO price following a Starship mission failure, Chinese firms like DeepSeek are aggressively scaling toward public listings with valuations implied at $52 billion. High-end hardware production remains robust, with ASML beating Q2 estimates and Nvidia confirming the Vera Rubin accelerator is on track for delivery. However, the sector faces mounting headwinds from tightened US export controls, as Nvidia has halved its Asia buyer list to crack down on sales to China.

Key variable to watch: Whether the diversion of budgets from traditional software to AI infrastructure, as signaled by IBM's crash, triggers a broader correction in enterprise software valuations.

Last updated: July 11, 2026

The semiconductor sector transitioned from a risk-off phase to a risk-on trajectory this week, driven by record-breaking liquidity events and robust memory demand. SK Hynix completed the largest-ever US listing by a foreign company, raising $26.5 billion through oversubscribed ADRs. This momentum was supported by global memory sales hitting a record $74.6 billion, a 31.7% month-on-month increase, and Samsung reporting a 19-fold jump in second-quarter operating profit. Despite these gains, Nvidia's valuation hit its lowest level since the AI boom began, trading under a 20 price-to-earnings ratio, though the company maintains that a global memory shortage will persist for several years.

Strategic shifts toward vertical integration and domestic autonomy are accelerating among key actors. Meta is moving toward hardware independence with its "Iris" AI chip scheduled for September production to support a 14-gigawatt capacity goal. In China, breakthroughs continue as a new neuromorphic chip reportedly outperformed Nvidia by up to 478x in brain-surface reconstruction, while YMTC unveiled 294-layer NAND. Furthermore, the US government has potentially eased some frictions, with reports that China may be allowed to purchase limited quantities of Nvidia H200 chips for public data training.

Investment implications are shifting toward long-term infrastructure and high-conviction memory plays. Massive capital commitments continue, evidenced by Amazon seeking $25 billion in bonds and Micron investing $3 billion into the US ecosystem. The market is also rewarding strategic partnerships, such as the multi-year, $30 billion agreement between Apple and Broadcom for US-based chip manufacturing. However, the emergence of Chinese firms like CXMT attempting to build full supply chains to compete with the "Big Three" memory makers introduces long-term structural risk to Western margins.

Key variable to watch: Whether Meta's shift to internal silicon and the potential for Chinese-made memory in Taiwanese PC brands signal a broader transition away from Nvidia and Western memory dominance.

Last updated: July 04, 2026

The semiconductor sector entered a sharp risk-off phase this week as valuations decoupled from fundamentals, triggering a "bear market" for high-beta momentum stocks. Memory chips led the decline, with Micron falling 10.3%, SK Hynix dropping 8.2%, and the Roundhill Memory ETF sliding 10.7%. Institutional skepticism is mounting, highlighted by Michael Burry's short positions in Nvidia and Micron, while Bank of America's Bubble Risk Indicator for the sector hit 0.91, the highest level since the dotcom era.

Despite the price correction, fundamental demand and state-led investment remain aggressive. South Korea announced a massive investment drive with Samsung and SK Group expected to deploy up to $1.3 trillion over the next decade, including 800 trillion won for four new fabs. Technical milestones continue, with Samsung achieving a 70% yield on HBM4E and initiating development of a 1.4nm foundry process. Furthermore, UBS projects significant Q3 price increases for DRAM (32%) and NAND (30%), while Samsung plans a 20% DRAM price hike.

Geopolitical volatility has introduced new operational risks, specifically regarding export controls. Taiwan authorities raided Super Micro Computer offices and detained employees over allegations of smuggling Nvidia chips into China, causing the stock to drop 10%. Conversely, the US government eased restrictions on Anthropic's Fable 5 and Mythos 5 models. Strategically, Nvidia is pivoting toward a revenue-sharing "AI Factory" model with cloud providers to maintain ecosystem dominance as overall AI usage costs drift lower.

Investment implications have shifted from broad momentum to a divergence between infrastructure plays and Big Tech. While semiconductor stocks reached a record 19.7% of the S&P 500's market cap, the "Magnificent Seven" saw combined market caps drop by $2.3 trillion in June. Investors are rotating out of TMT stocks at the fastest pace in over a decade, reacting to a "chip-crunch" that has forced price hikes for Apple and Microsoft hardware and created compute resource shortages for Google.

Key variable to watch: Whether the projected Q3 DRAM and NAND price hikes can offset the current valuation correction and stabilize investor sentiment.

Last updated: June 29, 2026

The semiconductor sector experienced extreme volatility this week, characterized by a massive rally in memory stocks followed by the steepest intraday plunge since October 2023. Micron drove the initial surge, reporting quarterly revenue of $41.5 billion and an adjusted EPS of $25.11, which propelled its market capitalization from $60 billion to $1.3 trillion and made it the most-traded U.S. stock. However, the sector eventually faced a sharp correction, with Micron and Sandisk both falling 10% from their all-time highs and Nvidia slipping nearly 2% from its year-to-date gains.

Diversification of the AI hardware stack accelerated as new competitors entered the data-center space. Qualcomm unveiled its Dragonfly C1000 CPU with Meta as the launch customer, projecting $15 billion in annual data-center chip revenue by 2029. Simultaneously, OpenAI and Broadcom introduced the "Jalapeño" chip, while Google partnered with MediaTek to develop an enhanced TPU V9. In the memory segment, Samsung's HBM4 revenue exceeded $1 billion within four months of mass production, though SK Hynix has moderated its HBM4 ramp to prioritize commodity DRAM profits.

Geopolitical and capital flows are shifting toward non-U.S. assets and alternative suppliers. Retail investors have rotated $22.5 billion into semiconductor ETFs since April, diverting funds from gold and Bitcoin. Notably, Apple is seeking U.S. government clearance to purchase memory chips from the blacklisted Chinese firm CXMT, signaling a potential rotation toward Chinese suppliers. While the EU joined the US-led Pax Silica Alliance to secure supply chains, Chinese hedge funds have warned that the AI hardware bubble is nearing a burst.

Key variable to watch: Whether the reported decline in production forecasts for Nvidia’s "Rubin" chip slows the industry's transition to HBM4 adoption.

Last updated: June 28, 2026

NVIDIA reported another beat-and-raise quarter this week, with data centre revenue up 122% YoY and H200 GPU order backlogs extending into Q2 2027. The result confirmed that hyperscaler AI capex is not slowing — Microsoft, Google, and Amazon all reaffirmed $60B+ annual AI infrastructure spend targets in their own recent commentary. The AI infrastructure trade remains intact, though the valuation multiple compression risk is rising as growth rates inevitably decelerate from triple-digit levels.

The US-China semiconductor conflict intensified with new export controls on advanced AI chip architectures. NVIDIA's China revenue will decline meaningfully in H2 2026, but management has guided that domestic demand from US hyperscalers more than offsets the lost China business. AMD is emerging as a credible H100 alternative for cost-sensitive workloads, gaining share in inference. TSMC's Arizona fab expansion remains ahead of schedule — a strategic win for supply chain resilience, though unit economics remain inferior to Taiwan fabs for now.

The key battleground this week is memory: SK Hynix's HBM3e is supply-constrained, creating a bottleneck for next-generation AI accelerators that is independent of the logic chip supply chain. Samsung is 1–2 quarters behind on HBM3e yield, and Micron is ramping but not yet at scale. Whoever cracks HBM supply first holds leverage over the entire AI compute stack. Key variable to watch: TSMC's July capacity announcement, which will set H2 allocation expectations across the entire chain.