Barclays analysts have highlighted a select group of global technology leaders poised to benefit from what they describe as ‘an unprecedented buildout of artificial intelligence infrastructure.’ According to the bank’s projections, annual AI‑related capital expenditure from hyperscalers and leading AI laboratories could exceed $1 trillion before peaking in 2028. This would reflect one of the largest investment cycles in history.
To assess the breadth of the opportunity, Barclays compiled a universe of more than 400 public and private companies spanning 19 digital and power‑infrastructure categories—from advanced semiconductors and custom accelerators to power systems, networking, and thermal management. In addition, the analysts also noted potential incremental upside from sovereign AI initiatives and activity in China. This could drive further demand for compute capacity.
Barclays notes that consensus estimates for hyperscaler capital expenditures may prove conservative. The bank sees potential upside of more than $300 billion relative to current forecasts.
While capex growth is expected to moderate later in the decade as recursive self‑improvement reduces training intensity, the analysts argue that near‑term infrastructure spending remains in a powerful expansion phase.
Leading AI Compute and Infrastructure
Barclays’ list reads like a who’s who of global tech innovation and domination, although it’s worth noting the absence of both Tesla (TSLA) and Apple (AAPL).
| Barclays’ top 10 AI infrastructure picks |
| Nvidia | AMD |
| Microsoft | Broadcom |
| Alphabet | Alibaba |
| Meta Platforms | Arm |
| Amazon | TSMC |
Source: Barclays
Nvidia (NVDA)
Nvidia remains the dominant force in AI compute, benefiting from its end‑to‑end platform leadership and rapid innovation cadence. The company’s Blackwell and forthcoming Rubin GPU architectures continue to set the industry standard for training and inference performance.
Nvidia’s CEO has disclosed visibility to more than $1 trillion in cumulative orders for these platforms between 2025 and 2027. This underscores sustained hyperscaler demand.
Microsoft (MSFT)
Microsoft is deepening vertical integration within Azure’s AI stack. Moreover, the company is deploying its proprietary Maia AI accelerators and Cobalt CPUs. These chips are designed to improve performance‑per‑watt and lower total cost of ownership for AI workloads.
Recent announcements include the availability of NVIDIA Nemotron models on Microsoft Foundry and the rollout of the NVIDIA Vera Rubin NVL72 supercomputer to Azure data centers. This is strengthening Microsoft’s training and inference capabilities.
Alphabet (GOOG)
Alphabet continues to leverage its internally developed Tensor Processing Units (TPUs) to support both Google Cloud customers and internal AI initiatives.
The company recently introduced updates to its Gemini AI assistant, including a chat import tool and the Gemini 3.1 Flash Live audio model for developers. This is further expanding its AI ecosystem.
Meta Platforms (META)
Meta is executing on an ambitious long‑term compute strategy through its in‑house Meta Training and Inference Accelerator (MTIA) roadmap.
The company is also reportedly exploring a potential partnership with the Adani Group to develop new data centres in India—a move that would significantly scale its global infrastructure footprint.
Amazon (AMZN)
Amazon Web Services continues to advance its purpose‑built accelerator lineup with Trainium and Inferentia chips, designed to optimize price‑performance ratios for high‑volume cloud AI customers.
AWS is reportedly developing AI tools aimed at automating internal sales operations. At the same time, banks such as JPMorgan cite ongoing strength in AWS demand across enterprise clients.
Advanced Micro Devices (AMD)
AMD is gaining traction through its Instinct GPU family and EPYC data center CPUs, which support both training and inference workloads.
The company recently partnered with Celestica to develop the Helios rack‑scale AI platform and signed a multi‑year licensing agreement with Adeia. This is enhancing its data center product roadmap.
Broadcom (AVGO)
Broadcom has established itself as a key supplier of custom silicon, merchant accelerators, and next‑generation networking solutions.
The company, alongside Carahsoft, secured a five‑year, $970 million contract with the U.S. Defense Information Systems Agency and began volume shipment of its Tomahawk 6 switch chip, critical for high‑bandwidth AI clusters.
Alibaba (BABA)
Alibaba continues to invest in domestic compute innovation, developing proprietary GPUs to support internal workloads and Alibaba Cloud customers.
The company recently launched the XuanTie C950—a next‑generation RISC‑V‑based AI chip reported to deliver more than three times the performance of its predecessor.
Arm (ARM)
Arm’s architecture underpins nearly all hyperscale custom silicon efforts, including Microsoft’s Cobalt, AWS’s Graviton, Oracle’s Axion, and Nvidia’s Grace‑Blackwell platforms.
Following its recent “Arm Everywhere” event, analysts upgraded the stock and raised price targets after Arm unveiled its new AGI‑class CPU. This chip is designed for agentic AI workloads.
TSMC (TSM)
TSMC remains the indispensable manufacturing partner for leading‑edge AI chips. It provides advanced process technologies and CoWoS packaging required for high‑bandwidth GPUs.
The company reported 29.9% year‑over‑year revenue growth for the first two months of 2026, reflecting surging demand from AI customers.
Broader ecosystem winners
Barclays also identifies additional compute and infrastructure players, including Intel (INTC), Marvell Technology (MRVL), Qualcomm (QCOM), and Tencent (TCEHY), as strategically important within the next wave of AI‑driven capital investment.
Disclaimer: The author Steven Frazer has a personal interest in Nvidia and Broadcom.
You might also like:
