In-Depth Research Analysis:
Macroeconomic Background:
Economic growth is expected to slow in 2025, with pandemic-related excess savings largely depleted. The main growth driver will shift from consumption to investment, with AI-related investments (data centers and power infrastructure) being a highlight. Trump's tariff policy remains an uncertainty, but its actual impact is likely overestimated, with intermittent effects on inflation.
Due to recent strong economic data and tight labor market conditions, the Federal Reserve will pause rate cuts to observe market reactions. There will likely be 1-2 rate cuts in 2025, with rates expected to end the year around 3.75%.
Another noteworthy point is the upcoming U.S.-Russia negotiations over Ukraine. Recently, the Dollar Index has retreated from its high of 110 to around 106, while the Euro and other currencies have surged. The potential end of the Russia-Ukraine war would help boost market risk appetite.
Overall, U.S. stocks still present opportunities this year, with the S&P 500 expected to rise between 6%-10%. However, the market will likely focus more on structural opportunities rather than broad-based gains.
Asset Class Analysis:
Equities/Stocks:
Recently, a Chinese company called DeepSeek launched its large language model DeepSeek-R1, demonstrating impressive reasoning capabilities in mathematics and programming that rival top-tier models like GPT-4. Notably, its API costs are approximately 20-30 times lower than GPT-4, with significantly reduced training and inference costs.
This development potentially marks a new phase in AI development. Previously, AI models relied heavily on raw computing power, with major tech companies engaging in an AI arms race through massive computational investments. This drove market investment focus toward AI infrastructure companies like NVIDIA. While the market was uncertain about which company would ultimately win the AI revolution, the intense competition ensured steady business for hardware providers like NVIDIA and Broadcom, explaining their sustained stock price growth.
However, DeepSeek-R1's emergence has shown that beyond just scaling up computing power, significant improvements can be achieved through optimizing model architecture and algorithms. The market initially reacted negatively to this news from a computing perspective, with NVIDIA's stock dropping 17% on January 27. We believe 2025 will mark the transition to an inference-focused era in AI, potentially challenging the investment thesis for hardware companies like NVIDIA, especially given their already high valuations.
Yet, the outlook for hardware isn't entirely negative. While training costs are decreasing, the widespread deployment of AI models is driving up inference demands. ASIC chips, which excel in inference tasks, are likely to gain market share relative to GPUs in the long term. The decreased training demand might be offset by increased inference requirements, making ASIC chips particularly promising for 2025.
This development is highly positive for the application layer, as lower inference costs enable widespread AI deployment across various applications. Recent U.S. stock market performance shows software stocks outperforming hardware stocks. We expect the 2025 investment focus to shift from computing hardware to AI applications, with consumer-facing (B2C) applications likely generating more market interest than business-facing (B2B) solutions.
Edge AI products will also benefit, including smart glasses, headphones, and phones that will increasingly integrate AI models. Lower inference costs and improved model performance will enhance the value proposition and user experience of AI hardware products. Interested investors should refer to our previous industry report on AI glasses.
Finally, 2025 could be a breakthrough year for autonomous driving. The integration of large language models reduces implementation barriers for autonomous driving, while automakers can develop subscription-based driver assistance products. This could transform traditional auto manufacturers into hardware+SaaS companies, potentially leading to valuation updates. Similar logic applies to other AI-integrated products like humanoid robots.
Based on the above analysis, here are our predictions for the North American technology sector:
1. The structural investment theme remains in the technology sector, with tech stocks (NASDAQ) outperforming the broader U.S. market (S&P 500)
2. Within the tech sector, AI-related stocks continue to show the strongest performance
3. Software application stocks (IGV) will outperform hardware stocks (SOX)
4. ASIC chip stocks (Broadcom) will outperform CPU stocks (NVIDIA)
5. Autonomous driving theme continues to present opportunities (Tesla, Hesai Technology)
6. Humanoid robotics enters its foundational year for scaling (Tesla, ABB, Amazon)
Moreover, we recommend increasing attention to Chinese stocks. In recent years, the prevailing view was that China had fallen behind in the AI arms race while U.S. stocks led in technological advantages. However, DeepSeek's emergence has prompted the market to reassess this perspective. Additionally, Chinese stocks have been suppressed by macroeconomic headwinds in recent years, leading to declining valuations that now present attractive value propositions.
As of mid-February, the Hong Kong Hang Seng Tech Index has been the best-performing index globally. We recommend focusing on U.S.-listed Chinese stocks, particularly those in AI applications. China has consistently demonstrated excellence in application innovation during previous innovation cycles. Notable examples include:
Alibaba (BABA)、PDD Holdings (PDD)、XPeng (XPEV)、Kingsoft Cloud (KC)
Key individual stocks to watch:
Broadcom (AVGO):
Broadcom (AVGO) is a leading semiconductor company specializing in custom ASIC solutions. While NVIDIA's GPUs dominate AI training, Broadcom's ASIC chips excel in AI inference tasks by offering optimized performance for specific applications. ASICs typically provide 10-20x better energy efficiency, lower costs, and reduced latency compared to general-purpose GPUs. Through partnerships with tech giants like Google and Meta for custom AI inference chips, Broadcom is well-positioned to capture market share as the industry focus shifts from training to inference deployment. The company's ASIC expertise represents a potential challenge to NVIDIA's dominance in certain AI applications.
Snowflake (SNOW):
A cloud-based data platform that enables enterprises to store, analyze, and share massive amounts of data. As an AI-era data warehouse leader, Snowflake is integrating AI capabilities to help customers build and deploy AI models, while its platform serves as a crucial infrastructure for training data and AI applications. The company's recent Cortex feature directly supports AI model deployment and management.
ServiceNow (NOW):
A leading enterprise workflow automation platform that's rapidly incorporating AI into its solutions. The company's AI-powered tools help automate IT services, customer service, and business workflows. ServiceNow's recent AI capabilities, including generative AI features, enable automated ticket resolution, predictive analytics, and intelligent process automation, significantly enhancing workplace efficiency.
XPeng (XPEV):
After a challenging period, XPeng is showing strong signs of turnaround with its advanced autonomous driving technology. The company's XNGP system represents a significant leap in intelligent driving capabilities, now available across China's major cities. With strategic partnerships with Volkswagen and improved vehicle delivery numbers, XPeng is emerging as a key player in smart EVs. The recent P7+ and X9 models demonstrate the company's ability to combine premium design with cutting-edge autonomous features.
Hesai Technology (HSAI):
As a leading LiDAR manufacturer, Hesai is benefiting from the increasing content value of LiDAR in ADAS applications. While early adoption focused on high-end vehicles, the push toward higher levels of autonomous driving is driving LiDAR integration into mass-market vehicles. This trend is significantly expanding the dollar content per vehicle for LiDAR systems. Hesai's cost-effective, high-performance solutions position it to capture this growing market as automakers increasingly view LiDAR as essential rather than optional for advanced driver assistance systems.
Bonds:
A unique situation has emerged in the U.S. bond market: while the Fed is expected to cut rates by 100 basis points, Treasury yields have paradoxically risen by 100 basis points. The yield increased from 3.63% in September to a peak of 4.8%. Markets are heavily pricing in the reflation risks of a potential Trump 2.0, making it feel like the Fed is hiking rates rather than cutting them.
From a technical perspective, there's strong support around the 4.8% level. Over the past few years, this key level has consistently triggered significant rebounds, and currently, yields have retreated to 4.5%. Furthermore, the market appears to be overpricing Trump 2.0 risks. There's typically a gap between Trump's rhetoric and actual policy implementation - looking at historical results, Trump's tariff execution in his first month was less aggressive than anticipated, using it more as a negotiation tool rather than immediate policy.
At current levels, U.S. Treasury investments offer favorable risk-reward ratios. Recommended attention on TLT and TLTW.
Gold:
Gold prices have risen from $2,600 to around $2,900 this year, representing approximately an 8% increase. This surge is partly driven by investors increasing their positions due to concerns about a potential 10% import tariff on gold. However, if the anticipated tariff increase doesn't materialize, gold might face a temporary correction.
The fundamental driver remains the continuous gold purchases by central banks worldwide, strengthening price support. This trend is corroborated by changes in countries' U.S. Treasury holdings - China reduced its U.S. Treasury holdings by $10 billion in December, while Japan cut $27 billion. Since the Russia-Ukraine conflict began, countries have shown serious concerns about the safety of dollar-denominated assets, achieving a form of de-dollarization by increasing gold holdings while reducing U.S. Treasury positions.
For long-term investors, assuming no fundamental changes in global geopolitics, we expect this central bank accumulation trend to continue, highlighting gold's long-term strategic value in portfolios. However, from a technical perspective, gold prices are showing signs of stagflation at high levels, presenting significant short-term risks. Short-term investors are advised to maintain caution.
Recommended focus on GLD.
Risk Factors:
1. Risk of Economic Underperformance: The global economy may not meet growth expectations, particularly given persistent inflation pressures and tight monetary conditions. This risk could lead to earnings downgrades across sectors.
2. Geopolitical Risk Shifts: Rising tensions, including potential conflicts in various regions and trade disputes, could disrupt global supply chains and market stability.
3. Risk of Advanced Technology Disruption (AI etc.): Rapid technological advances, particularly in AI, could fundamentally alter competitive landscapes and business models faster than companies can adapt.
Disclaimer:
The Information presented above is for information purposes only, which shall not be intended as and does not constitute an offer to sell or solicitation for an offer to buy any securities or financial instrument or any advice or recommendation with respect to such securities or other financial instruments or investments. When making a decision about your investments, you should seek the advice of a professional financial adviser and carefully consider whether such investments are suitable for you in light of your own experience, financial position and investment objectives. The firm and its analysts do not have any material interest or conflict of interest with any stocks mentioned in this report.
IN NO EVENT SHALL SAHM CAPITAL FINANCIAL COMPANY BE LIABLE FOR ANY DAMAGES, LOSSES OR LIABILITIES INCLUDING WITHOUT LIMITATION, DIRECT OR INDIRECT, SPECIAL, INCIDENTAL, CONSEQUENTIAL DAMAGES, LOSSES OR LIABILITIES, IN CONNECTION WITH YOUR RELIANCE ON OR USE OR INABILITY TO USE THE INFORMATION PRESENTED ABOVE, EVEN IF YOU ADVISE US OF THE POSSIBILITY OF SUCH DAMAGES, LOSSES OR EXPENSES.
