Alphabet (GOOGL) and NVIDIA (NVDA) remain foundational core holdings, as they continue to process the vast majority of commercial AI tokens and maintain an entrenched hardware-software ecosystem.

For exposure to the emerging Physical AI and robotics market, Tesla (TSLA) offers top-tier vertical integration across manufacturing and software, while specialized hardware makers like UBTECH and Unitree lead global unit shipments.

Within the Chinese AI space, rotate capital away from richly valued software labs and into domestic semiconductor manufacturers like SMIC and Cambricon, which are positioned to supply up to 85% of China's localized compute needs by 2028.

Meta Platforms (META) presents an actionable play on open-source disruption, leveraging broad developer distribution to steadily erode the pricing power of proprietary software competitors.

Investors should aggressively allocate to Energy & Power Grid Infrastructure, as power supply constraints represent the single biggest bottleneck to ongoing U.S. data center expansion.

NVIDIA Corporation (NVDA)

• NVIDIA continues to adapt to competitive threats by releasing models like Nemotron 3, which are low-cost and specifically co-designed to maximize performance on its proprietary hardware.
• The company leverages acquisitions such as Hugging Face to embed its underlying hardware stack directly into open-source developer workflows.
• Geopolitical dynamics and export restrictions have prompted Chinese data centers to reject H200 shipments in favor of domestic alternatives, accelerating China's efforts to build a sovereign supply chain.

Takeaways

• NVIDIA maintains an entrenched ecosystem through hardware-software co-design and dominant token processing market share, though long-term revenue from China faces headwinds as domestic alternatives gain traction.

Tesla, Inc. (TSLA)

• Tesla is uniquely positioned in the emerging "physical AI" and humanoid robotics market by combining proprietary manufacturing capabilities with xAI software models.
• The company is applying this unified approach across physical hardware applications, including autonomous vehicles (cybercabs) and factory robotics.
• Integration between the software brain and industrial production is viewed as a major structural advantage against pure-play software competitors.

Takeaways

• Tesla serves as a core play on physical AI and real-world robotics, benefiting from vertical integration across automated manufacturing, custom hardware, and frontier AI software.

Meta Platforms, Inc. (META)

• Meta continues to pressure closed-source AI labs by releasing low-cost, open models like MuseSpark.
• While open-weight models have traditionally been smaller in parameter size compared to major Chinese releases, Meta's strategy focuses on broad distribution and developer adoption.
• If Meta's open-source models reach performance parity with proprietary frontier systems, it could significantly challenge the pricing power of closed-source alternatives.

Takeaways

• Meta provides exposure to open-source AI distribution, leveraging commoditized model architectures to build extensive developer ecosystems without relying directly on API token fees for monetization.

Alphabet Inc. (GOOGL)

• Alphabet continues to defend its market share with the release of open and efficient models like Gemma 4.
• Alongside OpenAI and Anthropic, Google processes the vast majority (over 90%) of total commercial AI tokens, creating a large buffer against open-weight competitors.
• Incumbent frontier labs continue to see heavy enterprise usage for advanced reasoning, complex math, and scientific workflows that smaller open models cannot yet fully replace.

Takeaways

• Alphabet maintains a strong competitive position in enterprise-grade AI workloads, capturing high token volumes while simultaneously rolling out open models to maintain developer reach.

Chinese Semiconductor & AI Infrastructure Sector

• Chinese frontier AI software labs are trading at elevated valuations between 65x and 100x revenue, making the underlying hardware infrastructure a comparatively more attractive value proposition.
• Due to US chip export controls, Chinese firms are co-designing software around lower hardware bit-rates and memory bandwidth constraints.
• Key domestic infrastructure and chip companies benefiting from state-subsidized localization include SMIC, Cambricon, CXMT, Higon, MetaX, Moore Threads, and Huawei.
• Projections suggest China could domestically produce up to 85% of its compute needs by 2028.

Takeaways

• Investors seeking exposure to China's rapidly advancing AI sector should look at domestic semiconductor, foundry, and memory hardware providers rather than overvalued software model labs.

Humanoid Robotics & Physical AI Sector

• China currently leads the global volume for physical humanoid robotics, accounting for roughly 97% of global humanoid shipments and manufacturing units at rates of one robot every 30 minutes.
• Leading hardware players in the region include Unitree, UBTECH, and AGIBot.
• Frontier AI research labs (such as DeepSeek, which invested in Unitree's initial public offering) are increasingly partnering with and funding physical robotics manufacturers.

Takeaways

• Real-world automation and humanoid robotics are scaling rapidly through Chinese industrial manufacturing pipelines, making supply-chain leaders and hardware-software partnerships attractive targets in the physical AI theme.

Energy & Power Grid Infrastructure

• Grid capacity has become a primary bottleneck for AI expansion in the United States, with power constraints likely to limit data center growth before chip supply does.
• AI energy demand in the U.S. accounts for 50% to 70% of all newly added grid capacity over the past five years, compared to only 1% to 5% in China (which adds power capacity at roughly six times the U.S. rate).
• Sustaining Western AI leadership will require significant capital allocation and government collaboration in domestic power generation, transmission, and grid infrastructure.

Takeaways

• Energy utilities, power generation, and electrical grid infrastructure represent critical structural plays necessary to support long-term data center expansion and AI compute scaling.