In the first half of 2026, the five major hyperscale cloud service providers—Microsoft, Amazon, Google, Meta, and Oracle—collectively raised their capital expenditure guidance. After analyzing first-quarter financial reports, Morgan Stanley’s internet equities team projected that the combined capital expenditures of these five giants will reach approximately $800 billion in 2026 and climb to $1.2 trillion in 2027, a significant increase from their previous forecast of $450 billion. Another Morgan Stanley report indicates that hyperscale providers will drive about 40% of Russell 1000 companies’ cash capital expenditures between 2026 and 2028, with the total exceeding $2 trillion.
Bank of America Securities has also sharply revised its forecasts upward. Analyst Vivek Arya’s team predicts that global hyperscale cloud providers’ AI capital expenditures will surpass $800 billion in 2026, a 67% year-over-year increase, and break the $1 trillion mark in 2027. This outlook is based on Alphabet, Microsoft, Meta, and Amazon all reporting quarterly revenues above Wall Street expectations, with AI and cloud service demand as the main drivers—Meta raised its full-year 2026 capital expenditure guidance to $125–145 billion, while Amazon’s AWS division grew 28%, marking its fastest pace since 2022.
A longer-term perspective is equally noteworthy. At Marvell’s 2025 AI Investor Day, the company cited data showing global data center capital expenditures at $435 billion in 2024, projected to reach $593 billion in 2025, and potentially surpass $1 trillion by 2028, with a compound annual growth rate (CAGR) of 20% from 2025 to 2028. If the five largest US tech companies’ combined capital expenditures hit around $650 billion in 2026, these figures would be largely consistent.
At the March 2026 GTC, NVIDIA CEO Jensen Huang announced that the company’s Blackwell and Vera Rubin AI chip families are expected to achieve at least $1 trillion in cumulative revenue by the end of 2027, doubling the previous $500 billion forecast. NVIDIA CFO Colette Kress added on the earnings call that, as agent-based AI begins to penetrate various industries, annual AI infrastructure spending could reach $3–4 trillion by the end of the decade.
The following sections break down the core beneficiaries of this capital expenditure cycle, from upstream to downstream.
GPU Chips and ASICs: NVIDIA Leads, Broadcom and Marvell Benefit in Parallel
NVIDIA (NVDA) is the most direct beneficiary of this spending wave. In fiscal year 2026, NVIDIA reported $215.9 billion in revenue, up 65% year-over-year, with a GAAP gross margin of 71.1%. Data center operations were the clear growth engine, generating $193.7 billion in revenue—a 68% increase—accounting for about 90% of total company revenue. In Q4 alone, data center revenue reached $62.3 billion, up 75% year-over-year and setting a new quarterly record. Hyperscale providers are the largest customer group for NVIDIA’s data center business.
With the Blackwell platform scaling up and the Vera Rubin platform about to ship, NVIDIA is also accelerating growth on the inference side. On the earnings call, Jensen Huang described the current phase as an "inference inflection point," emphasizing that real-time computing needs for running AI systems are becoming a new growth engine. However, not all of the $1 trillion in infrastructure spending will go to GPUs. The ASIC custom chip market is expanding even faster.
Broadcom (AVGO) holds a dominant position in this market. According to Yahoo Finance, Broadcom could capture about 60% of the AI server compute ASIC market by 2027, with ASIC shipments expected to triple. On the earnings call, CEO Hock Tan stated that the AI chip market opportunity from its three largest customers will be between $60–90 billion in fiscal 2027, and Broadcom is poised to secure a "reasonable share." A research report from Soochow Securities notes that Marvell projects global data center capital expenditures to exceed $1 trillion by 2028, with AI acceleration compute accounting for $349 billion. This has led to an upward revision of the ASIC market size to $55.4 billion, with a 2023–2028 CAGR of 53%.
Surging demand for AI servers is also driving growth in storage. HBM and high-performance DRAM have become another area of rigid demand beyond AI chips, with SK Hynix, Micron Technology (MU), and Samsung Electronics as the key beneficiaries.
Server Assembly and System Integration
Once the chips are manufactured, server assembly and system integration take over as the next direct demand center for AI infrastructure spending. Major players include Hewlett Packard Enterprise (HPE), Dell Technologies (DELL), and Super Micro Computer (SMCI). According to IDC, the global data center market is valued at about $347 billion in 2024 and is expected to grow to the $627–650 billion range by 2030, with server infrastructure investment as the primary source of incremental growth.
Networking Equipment: Upgrading AI Cluster Interconnects
AI clusters are scaling from thousands to tens of thousands, even hundreds of thousands of cards, driving demand for high-speed Ethernet and InfiniBand equipment as back-end network connections evolve through both scale-up and scale-out architectures.
Arista Networks (ANET) leads among hyperscale cloud providers with its data center switch products. Cisco (CSCO), the traditional enterprise networking giant, is rapidly pivoting toward the AI data center networking market. Both companies’ switches, routers, and optical modules will directly benefit from the sustained bandwidth expansion expected in 2026–2027.
Data Center Cooling: Accelerated Growth in Liquid Cooling
As AI servers’ power density continues to rise—GPU rack power consumption has jumped from several kilowatts in traditional servers to tens of kilowatts or more—data center cooling technology is shifting rapidly from air to liquid cooling.
According to Grand View Research’s GMI report, the global data center liquid cooling market is expected to reach $3.3 billion in 2025 and $10.55 billion by 2030, with a CAGR of 26.1%. Another research agency forecasts the market at $870 million in 2024, growing to $10.7 billion by 2030, with a CAGR as high as 51.93%. A report from SDIC Securities projects that the global new-build data center liquid cooling system market could exceed $50 billion by 2030, with a 22% CAGR between 2026 and 2030.
In terms of competitive landscape, Vertiv (VRT) is expected to lead the liquid cooling market in 2025 with a share exceeding 11.3%. The top five vendors (Vertiv, Schneider Electric, Rittal, Stulz, Boyd) together hold about 35% market share. Schneider Electric (European ticker: SU) also offers a comprehensive suite of liquid cooling solutions. Previous industry reports estimated Vertiv’s market share in liquid cooling at over 60%, but figures vary by source and GMI’s formal market share report should be considered authoritative.
Power Supply: The Gigawatt-Scale Bottleneck
Power demand for AI data centers is emerging as the most severe bottleneck in infrastructure build-out. An Evercore ISI report notes that announced incremental data center power demand already exceeds 125 GW, with 2026 labeled as a "critical year" for the power industry. SemiAnalysis forecasts that global data center critical IT power demand will surge from 49 GW in 2023 to 96 GW in 2026, with AI consuming about 40 GW. Vertiv projects global data center power demand will reach 140 GW by 2029, a 100 GW increase in just five years.
On the supply side, two types of companies stand to benefit most: independent power producers who can sell electricity to data centers on a spot or long-term contract basis, and utilities with large-scale generation assets.
Vistra Corp (VST) is a standout in this field. In January 2026, Vistra signed a 20-year nuclear power purchase agreement (PPA) with Meta, committing to deliver over 2,600 MW of zero-carbon electricity in phases starting at the end of 2026 and reaching full capacity by 2034. According to Investing.com, Vistra’s diversified generation portfolio and retail business give it significant resilience and flexibility as power demand rises.
NextEra Energy (NEE) is America’s leading renewable energy provider. In March 2026, NVIDIA announced partnerships with six US energy giants—including AES, Constellation Energy, NextEra, and Vistra—to unlock up to 100 GW of idle US grid capacity for AI data centers. Google also signed a 25-year PPA with NextEra to help restart a nuclear plant in Illinois.
Energy Storage: From Backup Power to Grid Interactivity
AI data centers’ power loads are not constant—they fluctuate significantly with batch training schedules and inference requests, creating pronounced peaks and valleys. As a result, energy storage systems serve not just as backup power but also play a key role in grid interaction and power cost management.
Energy storage systems deliver value in three main ways: smoothing out load fluctuations to avoid high peak-time prices; providing grid frequency regulation for revenue; and bridging power supply when grid capacity is insufficient for data centers. This trend is driving sustained order visibility for grid-scale energy storage integrators like Fluence Energy (FLNC).
Data Center REITs: Structural Opportunities in Land and Construction
Physical construction of AI data centers also creates structural investment opportunities in land ownership and data center operations. Data center REITs are the most direct beneficiaries at this stage. WisdomTree analysis notes that hyperscale providers are partnering with companies like Digital Realty and Equinix—the former focusing on large-scale builds with robust power and cooling, the latter emphasizing interconnection hubs for AI training workloads. Data center REITs offer predictable cash flows from long-term leases, typically ranging from 10 to 20 years, and have significant pricing power.
Major data center REITs include: Equinix (EQIX), the world’s largest data center REIT with a market cap of about $108 billion and a presence in all major global markets; Digital Realty (DLR), the largest wholesale data center REIT, dominant in core markets; and Iron Mountain (IRM), which has transitioned from traditional document management to managed data center operations, serving around 240,000 customers across 61 countries.
Full Industry Value Chain Overview
Upstream core: GPU chips led by NVIDIA (NVDA), with $193.7 billion in annual data center revenue; ASIC/custom chips led by Broadcom (AVGO), with the three largest customers’ XPU and networking market totaling $60–90 billion by 2027; storage led by Micron (MU) and others, with HBM demand driving revenue elasticity.
Midstream assembly and interconnect: Server assembly led by Hewlett Packard Enterprise (HPE), Dell (DELL), and Super Micro Computer (SMCI); networking equipment led by Arista (ANET) and Cisco (CSCO).
Downstream support: Liquid cooling led by Vertiv (VRT), with an 11.3% market share; power supply led by Vistra (VST) and NextEra Energy (NEE); energy storage led by Fluence (FLNC); data center REITs led by Equinix (EQIX) and Digital Realty (DLR).
Gate US Stock Trading: Bridging Crypto Assets and US AI Infrastructure Investment
While analyzing AI industry beneficiaries, investors face a practical question: how can they efficiently allocate between crypto assets and US equities? In June 2026, Gate officially launched US stock trading services, allowing users to trade over 10,000 US stocks and ETFs directly with USDT on the Gate platform. This covers major US exchanges and liquidity networks, including NYSE, NASDAQ, NYSE Arca, NYSE American, and BATS, and supports fractional share trading starting from as little as 0.01 shares, lowering the barrier to entry for US equity investment.
Gate’s US stock trading offers three core advantages. First, compliance: Gate has formed a strategic partnership with SEC-registered broker-dealer Alpaca, using a comprehensive clearing agreement to cover the full chain of stock trading—execution, clearing, settlement, custody, dividend distribution, and corporate actions. Second, capital efficiency: users can switch holdings between crypto assets and US equities within the same account, enabling instant participation in AI infrastructure stocks with USDT when crypto markets are volatile. Third, no overnight holding fees: unlike perpetual swaps and CFD products, Gate spot stock trading does not involve funding rates or overnight holding costs, making it more suitable for medium- and long-term allocation. Additionally, Gate’s partner broker is a SIPC member, providing protection for securities assets under certain conditions.
To trade US stocks on Gate, users should: update the Gate App to the latest version (Android supported; iOS requires version 8.21.5 or above); complete platform KYC verification and confirm eligibility based on location; enter the "TradFi" section from the bottom navigation bar and access the US stock area; transfer USDT to the US stock account via the trading or asset page to participate in real-time stock and ETF trading. The service currently supports intraday trading, with plans to gradually expand to 24/7 trading.
Conclusion
Morgan Stanley’s report highlights that the combined capital expenditures of hyperscale providers totaled just $260 billion in 2024, but are expected to approach $800 billion in 2026 and reach $1.2 trillion in 2027—representing more than a fourfold expansion in just three years. From chip manufacturing to power supply, from liquid cooling systems to data center REITs, the entire AI infrastructure value chain is undergoing systemic repricing.
At the same time, this capital expenditure race comes with significant risks. The biggest uncertainty is whether profit growth can keep pace with the intensity of capital spending—if AI service monetization lags expectations, hyperscale providers’ credit quality and cash flow sustainability will face renewed scrutiny. Additionally, the pace of US grid infrastructure expansion and whether chip shortages will squeeze demand in other sectors, such as consumer electronics, pose broader economic transmission risks. Investors should carefully assess the risk-return characteristics of each segment when participating in the AI infrastructure theme and allocate prudently.
