Come together – AI’s value chain from chips to commerce

Artificial intelligence (AI) has moved quickly from hype to deployment. Since generative models emerged in 2022, corporate investment has accelerated, with US AI-related capex now exceeding US$250bn annually1, a scale already comparable to the internet boom of the 1990s. Analysts expect this to climb further as adoption spreads and competition intensifies. What distinguishes this cycle is not only its speed but also its breadth. Semiconductors, infrastructure and platforms are all being pulled into the buildout, creating reinforcing demand across the value chain2.

AI’s trajectory can be seen in three categories: hardware, infrastructure and adoption. Hardware provides the computational base, infrastructure supplies the power and physical systems that enable scale, and adoption reflects the monetisation of AI across consumer and enterprise platforms. In the United States, the world’s largest technology companies are embedding AI into search, advertising, cloud services, cybersecurity and workflow automation, highlighting how AI is shifting from an R&D expense to a commercial growth driver .

Key takeaways

• Semiconductors are scaling into a new cycle, with revenues projected to rise 38% by 20263, and networking emerging as a critical bottleneck.
• Infrastructure is becoming the constraint, with hyperscaler and other capex commitments, grid equipment shortages, higher transformer costs and aging US power infrastructure.
• Adoption is shifting into a monetisation phase, with technology leaders embedding AI into search, cloud, advertising, and workflow automation, driving potential earnings uplift.

AI Hardware: Scaling the Brains of the System

The semiconductor industry remains the foundation of AI’s growth story. Sector revenue is expected to expand by 38% over 2025-26, supported by hyperscaler investment and the diffusion of AI workloads into various industries. Nvidia, AMD and Broadcom are extending compute clusters to unprecedented scales, with installations now exceeding 100,000 GPUs4 and forecasts pointing to a tenfold increase over the next decade. Networking has become as critical as compute, with Broadcom and Marvell advancing Ethernet switches and Nvidia expanding its Spectrum-X architecture to reduce bottlenecks inside disaggregated data centres.

Memory is entering the AI spotlight in ways not seen before. NAND flash memory chip is projected to account for 34% of the global market by 2029, adding nearly US$30bn in incremental opportunity to the semiconductor industry.5 Supply shortages are already anticipated from 2026 as capacity expansion lags surging demand. Hard drives, once seen as a legacy technology, remain cost-effective for large-scale data storage, suggesting both HDDs and SSDs will benefit from the next phase of AI. Equipment and testing providers are also seeing stronger order books as customers signal more structured long-term investment plans.

AI Infrastructure: Powering the Scale-Up

Artificial intelligence is putting unprecedented strain on physical infrastructure. Oracle has lifted its FY26 capex plans to US$25bn, more than double initial expectations, with its order backlog reaching US$138bn and projected to double again as training and inference demand accelerates. US hyperscalers6 are making similar commitments, with depreciation charges rising as new facilities come online. Global electricity demand growth, which averaged just over 2% annually for the past decade, is now expected to accelerate to 3.5% through 2030 as data centres absorb a larger share of incremental load.7

This expansion is colliding with ageing grid systems. Transformer delivery times have stretched to two years for standard units and up to four years for larger models, while prices are up as much as 80% since 2020 and 70% of US transmission transformers are over 25 years old.8 Suppliers face record backlogs, with Korean firms like HD Hyundai Electric and LS Electric winning multi-billion-dollar contracts, and Eaton, Schneider and Siemens Energy reporting stronger pipelines. At the datacentre level, Vertiv and Super Micro are posting record bookings in servers and cooling systems, while copper and uranium producers stand to benefit from grid upgrades.

AI Adoption: From Tools to Revenue Engines

Artificial intelligence is shifting from pilot projects to scaled monetisation across the largest US platforms. Microsoft has already deployed Copilot into Office 365, with early adoption helping drive the company’s US$250bn+ enterprise software run rate higher, while Azure’s AI services are compounding backlog growth.9 Alphabet is embedding generative AI into search and YouTube ad products, where management estimates AI-driven formats are lifting return on investment for advertisers. Meta is rolling out AI assistants across messaging apps and expanding Advantage+ ad tools, contributing to higher engagement and better monetisation of its 3.2bn daily active users.10

Apple has taken a hardware-first approach, unveiling AI-native iPhones that integrate on-device large language models to differentiate user experience and potentially spark a new replacement cycle. Amazon is embedding AI across AWS, where management recently reported11 record backlog growth tied to training and inference demand, and across e-commerce with AI-enhanced logistics and personalised recommendations. ServiceNow and CrowdStrike are layering AI into enterprise workflows and cybersecurity, improving efficiency and broadening total addressable markets.

Analysts estimate that across these names, AI adoption could raise long-term earnings by 10–50%, with incremental revenue streams from software subscriptions, advertising, cloud services and device sales.12 This breadth shows that AI has moved beyond proof-of-concept to become a material driver of revenue and margin expansion across the US technology complex.

Conclusion: Capturing the Value Chain

AI is unfolding across three interconnected layers. Semiconductors provide the compute foundation, infrastructure delivers the power and systems that enable scale, and platforms are embedding AI into everyday applications, driving monetisation. Together, they form an investment cycle that is global, structural, and self-reinforcing.

For investors, Global X Semiconductor ETF (SEMI) aligns with the hardware backbone, Global X AI Infrastructure ETF (AINF) with the enablers powering the scale-up, and Global X FANG+ ETF (FANG) with the adoption of AI through leading US technology platforms. For those seeking broad exposure across all three, the Global X Artificial Intelligence ETF (GXAI) spans the full value chain in a single portfolio.

ETF

Global X Artificial Intelligence ETF (GXAI)

Global Shares

View