Gemma

AMD's Instinct MI300 accelerators are positioned to capture significant market share as hyperscalers seek to diversify their silicon supply away from NVIDIA.

AMD maintains a strong, established foothold in the data center CPU market and possesses a growing ecosystem of AI software and hardware integration.

["Intense competition from NVIDIA's dominant CUDA ecosystem","Potential for customized silicon developed by major cloud providers"]

Microsoft is at the forefront of the generative AI revolution through its partnership with OpenAI and the seamless integration of Copilot across its entire software suite. Azure provides the scalable cloud backbone necessary for the next decade of AI workloads.

Extremely high switching costs due to the deep integration of Windows, Office, and Azure within enterprise environments. The massive installed base provides a unique distribution moat for new AI features.

["Increasing regulatory scrutiny regarding cloud dominance","Significant capital expenditure requirements for AI data centers"]

TSM is the indispensable foundry for the AI revolution, manufacturing the high-end, advanced-node silicon required by Nvidia, AMD, and Apple.

The massive capital intensity and specialized technical expertise required for sub-5nm lithography create an almost insurmountable barrier to entry.

["Geopolitical risk regarding Taiwan and China","High capital expenditure requirements"]

Alphabet's blowout Q2 results, specifically the 82% YoY cloud growth, provide concrete evidence that their AI stack is driving significant revenue.

The company possesses unparalleled data flywheels across Search and YouTube, creating a massive barrier to entry for AI competitors.

["Antitrust litigation","Increasing CapEx"]

As AI chips like NVIDIA's Blackwell demand significantly higher power and cooling density, specialized thermal management becomes a non-discretionary component of data center architecture.

Vertiv has deep institutional relationships and specialized engineering expertise that create high barriers to entry in mission-critical data center cooling.

["High valuation multiples compared to traditional industrials","Sensitivity to global capital expenditure cycles"]

Meta's aggressive open-source approach with Llama and Muse Glimmer creates a standard for the industry, driving developer adoption and ecosystem growth.

Massive network effects across social platforms and a vast data advantage for training models provide a significant competitive edge.

["Antitrust and regulatory scrutiny","Significant capital expenditure on AI infrastructure"]

Broadcom is a critical provider of AI networking and custom silicon (ASICs), positioned as the most significant beneficiary behind NVIDIA in the AI infrastructure build-out.

The company maintains high switching costs through deep integration with hyperscale customers and a dominant position in high-performance networking semiconductors.

["Significant customer concentration among a few hyperscale cloud providers"]

Strong Q1 earnings and government contract growth (Maven) demonstrate successful AI commercialization.

Deeply embedded software in government and enterprise workflows creates high switching costs.

["High valuation multiples","Potential for contract volatility"]

AWS is a primary beneficiary of the AI surge, providing the essential cloud infrastructure and compute services required by hyperscalers and enterprises.

Amazon possesses massive scale, high switching costs within its ecosystem, and a dominant position in both cloud and logistics.

["Upcoming earnings volatility","Regulatory/Antitrust scrutiny"]

Marvell is a critical provider of high-speed connectivity and custom ASIC solutions required for large-scale AI data centers, directly benefiting from the hardware build-out cycle.

The company possesses strong IP in custom silicon and high switching costs within hyperscale environments, creating a significant barrier to entry.

["High customer concentration among major cloud providers","Potential semiconductor cyclicality"]

NVIDIA provides the essential compute backbone for training and inferencing large language models, creating a massive structural demand tailwind.

The CUDA software platform creates high switching costs for developers, effectively locking in the hardware ecosystem.

["Geopolitical sensitivity regarding Taiwan manufacturing","High valuation multiples"]