Tariffs, Inflation, and Trade Deficits

The Potential Short and Long-Term Impact on Semiconductors and Artificial Intelligence

The global technology landscape is being disrupted by US tariff policies, persistent inflationary pressures, and a US trade deficit. We believe current market dynamics present both challenges and opportunities throughout the technology sector, specifically across semiconductors and artificial intelligence (AI). As a firm, we are directionally bullish on this technology’s growth potential and believe much of the volatility we are seeing surrounding these sectors stems from public sentiment, uncertainty, and speculation rather than a deterioration in underlying intrinsic value. Consequently, we view broader multiple compression across the public markets as a potential catalyst to attractive entry points for private equity and technology growth asset classes.

Our thesis is predicated on a conviction that the long-term value of these technologies far outweighs short-term volatility posed by tariffs. We believe the fundamental growth drivers in each sector remain exceptionally strong, and see the current environment not as a deterrent, but as an opportunity to identify and back resilient companies poised to lead the next wave of growth.

2025’s Turbulent Macroeconomic Backdrop

Broad baseline tariffs and "reciprocal" tariffs aimed at various trading partners have driven the average US tariff rate to century highs, injecting significant levels of uncertainty across global supply chains. While headline inflation shows signs of easing, underlying pressures fueled by tariffs and structural factors may prove persistent, complicating policies, corporate costs, and consumer demand. For the semiconductor industry, these dynamics translate to direct cost increases via tariffs on chips and equipment, supply chain adjustments driven by geopolitical risk and resilience mandates like the CHIPS Act, and execution hurdles created by talent shortages and manufacturing cost differentials between Asia and the US.

Inflation's Persistence, Pricing Pressures, and Economic Impact

Inflation remains a central macroeconomic concern in 2025, shaping operational costs, consumer behavior, and monetary policy. Recent data indicates some easing from peak levels. The annual Consumer Price Index for All Urban Consumers (CPI-U) slowed to 2.4% year-over-year in March 2025, down from 2.8% in February and below consensus forecasts. Core CPI (excluding food and energy) also decelerated to 2.8% year-over-year, its lowest reading since March 2021. On a monthly basis, the CPI even registered a slight decline of 0.1% in March, the first decrease since May 2020.

While the easing in headline numbers might suggest pandemic-related supply shocks and demand surges are fading, the imposition of broad tariffs introduces a strong cost-push element. This dynamic creates a challenge for the Federal Reserve: yes tariffs may be inflationary, arguing against rate cuts, but the economic drag caused by these same tariffs typically calls for monetary easing. This policy conflict heightens uncertainty about the future path of interest rates, directly impacting M&A activity and discount rates used in exit valuations. For these reasons, we believe there is a need for portfolio companies to implement strategies for managing input cost volatility and navigating price adjustments, while short term valuation models may increasingly incorporate the possibility of higher-for-longer interest rates and potential impacts of inflation on demand within their target markets.

The US Trade Deficit

The US trade deficit serves as a driving force behind the administration's tariff policies. The scale of this deficit reached a record $130.7 billion for goods and services in January 2025, fueled partly by importers rushing orders ahead of anticipated tariffs, before narrowing slightly to $122.7 billion in February. The goods deficit alone stood at $147.0 billion in February. The full year 2024 good deficit reached a staggering $1.2 trillion, cited by the administration as evidence of fundamentally flawed trade relationships.  

Technology is a significant input to this imbalance. The US consistently runs a deficit in Advanced Technology Products (ATP), amounting to $31.9 billion in February 2025 alone. Within ATP, major deficits include information and communications technology (-$21.0 billion) and biotech (-$13.2 billion). While the US exports substantial capital goods like computer accessories and aircraft, these are outweighed by imports of consumer electronics (cell phones), computers, and critical industrial inputs like finished metal shapes. China, the EU, Mexico, Vietnam, and Taiwan heavily contribute to the overall deficit. While tariffs and policies like the CHIPS Act aim to re-shore manufacturing and enhance domestic capabilities, tariffs increase the cost of imported technology components for US firms in the short term. This cost burden presents direct risks for portfolio companies reliant on tech imports but also creates potential opportunities for companies facilitating domestic supply chains or benefiting from re-shoring initiatives.

The Semiconductor Sector

The semiconductor industry finds itself at the epicenter of current macroeconomic and geopolitical turbulence. Tariffs, supply chain vulnerabilities, investment initiatives, talent shortages, and bifurcated market demand have created complicated dynamics for chipmakers, upstream suppliers, downstream consumers, and investors alike.

US export controls, specifically those targeting advanced AI chips destined for China, function as non-tariff barriers, impacting revenue streams for leading semiconductor firms like Nvidia. On April 15th, Nvidia noted it would take $5.5 billion in charges related to inventory, reserves, and purchase commitments after the US limited exports of its H20 AI chip to China. Chinese businesses such as ByteDance (parent of TikTok)  and Alibaba have been increasing demand for the H20, especially following the release of China’s frontier model DeepSeek. Nvidia began designing these chips to stay within US export control limits, and while these chips may not be as performant compared to their US counterparts for training AI models, they have demonstrated near performance parity for inference workloads requiring high-memory bandwidth connectivity. In response, the industry is undergoing significant supply chain adjustments. Companies are actively diversifying manufacturing and sourcing away from China, pursuing regionalization strategies. Government initiatives like the US CHIPS Act and the EU Chips Act are explicitly designed to encourage re-shoring and "friend-shoring" initiatives. However, Taiwan remains dominant in leading-edge logic manufacturing, China controls key rare earth materials, and Europe (specifically ASML in the Netherlands) holds a near-monopoly on essential extreme ultraviolet (EUV) lithography equipment.

In the US, the CHIPS Act has acted as a powerful catalyst stimulating over $450+ billion in private sector investments across more than 80 projects since its enactment. Major funding awards have been announced for leading companies including Intel, Taiwan Semiconductor Manufacturing Company (TSMC), Samsung, and Micron, targeting the establishment of manufacturing facilities on US soil. Most recently, Nvidia announced plans to invest $500 billion into US AI server production over the next 4 years, while TSMC announced it aims to expand its initial $25 billion investment in US manufacturing to $65 billion, committing to add a third factory in Arizona by 2030.

Similar initiatives, like the European Chips Act, are underway in other regions, intensifying the global race for semiconductor manufacturing capabilities. In our purview, one of the most significant bottlenecks in the race is an acute shortage of skilled labor. The US semiconductor industry is projected to require nearly 115,000 additional workers by 2030, yet estimates suggest that roughly 67,000 of these positions, specifically for technicians, computer scientists, and engineers of all degree levels risk going unfilled based on current graduation and workforce entry rates.

AI R&D: Venture Funding and the Battle for Talent

The United States maintained a commanding lead in private AI investments in 2024, attracting $109.1 billion, outpacing China ($9.3 billion) and the UK ($4.5 billion). In 2024, global private investments in AI increased by 44.5% YoY, representing almost a third of total global venture funding.

The competition for AI talent is intense and global. Governments are actively implementing policies to attract and retain skilled AI professionals, viewing talent as a strategic asset. China's Thousand Talents Plan is an example aimed at repatriation, while frameworks like the USMCA are being considered as vehicles to streamline cross-border talent mobility within North America. While international bodies like the WEF, OECD, and the UN emphasize the need for global cooperation on AI governance and standards, reality is shaped by national interests and strategic competition.

High-Cost AI Infrastructure Amid Tariffs and Export Controls

Large language models (LLMs) and generative AI are compute-intensive, relying heavily on Graphics Processing Units (GPUs) for training and inference. Cloud hyperscalers have announced plans to spend $320 billion+ in 2025, up significantly from $230 billion in 2024.

We view export controls like the AI Diffusion Rule as a potential factor that creates artificial scarcity and tiered global systems for accessing state of the art semiconductors. This structure could inadvertently concentrate power among the large US cloud providers designated as authorized exporters (UVEUs/NVEUs). For these reasons, we anticipate the cost for state of the art compute will to continue to rise as new architectures are released, while legacy architectures (Ampere, Hopper etc.) steadily decline. We anticipate cloud cost optimization and efficiency will increasingly be central to enterprise cost cutting strategies, contributing to an uptick in the exploration of efficiency techniques such as low-rank adaptation (LoRA), distillation, pruning, and quantization (do we have to use state-of-the-art GPUs or can we reduce the model weights to a lower precision format and fit it on a smaller chip while minimizing performance degradation?). Finally, we anticipate hyperscalers will expand R&D efforts in search of alternative computing architectures designed for high-performance training and inference workloads (e.g., neuromorphic computing, photonic computing, quantum computing).

---

Important Information and Disclosures

The views expressed in this presentation are attributable solely to the author, Nick Alonso, in his individual capacity. They do not necessarily represent the collective view, official stance, or investment strategies of Songster Capital LLC or its affiliated entities ("Songster Capital").

This content been prepared solely for informational purposes as of August 29, 2025. It should not be interpreted as constituting investment research or a formal recommendation. Furthermore, nothing herein should be construed as investment, legal, tax, or other professional advice, nor is it an offer to sell or a solicitation of an offer to buy any security, financial instrument, or investment product. Readers are encouraged to use this information as merely one component of their own independent assessment and should not rely solely upon it for making any investment or financial decisions. The views presented reflect the author's perspective at the time of writing and are subject to evolution based on changing market conditions or other factors. There is no obligation, express or implied, on the part of the author or Songster Capital to update this information or notify readers of any changes to the views expressed. Any use of terms like "we" or "our" relates to the author and/or Songster Capital as the specific context requires. Please be advised that the commentary provided does not pertain specifically to any Songster Capital investment product or management strategy. Songster Capital and its personnel, potentially including the author, may engage in investment activities, advise clients, hold financial interests (long or short), or formulate recommendations that diverge significantly from, or are contrary to, the perspectives shared in this document. Assumptions made or techniques described by the author may not be utilized in managing client accounts or firm capital. While information presented is gathered from sources believed to be reliable, its accuracy, completeness, and timeliness cannot be explicitly guaranteed by Songster Capital or the author. Any statements anticipating future events, market movements, or performance are inherently speculative, based on current assumptions, and involve significant risks and uncertainties; actual outcomes may vary substantially from any projections or expectations discussed. All investments carry inherent risks, including the potential loss of capital. Past performance is not a reliable indicator or guarantee of future results. The suitability of any investment theme or strategy discussed herein depends heavily on individual investor circumstances, objectives, and risk tolerance.

Neither Songster Capital nor the author accepts any responsibility or liability for any direct or consequential loss or damage arising from the use of, or reliance upon, this information, or for any errors or omissions contained within it. By accepting this presentation in its entirety, the recipient acknowledges its understanding and acceptance of the foregoing statement.