In 2026, the global buildout of artificial intelligence infrastructure is advancing at an unprecedented pace. Morgan Stanley projects that by 2028, nearly $3 trillion in AI-related infrastructure investment will flow through the global economy, with over 80% of this spending still ahead. In 2026 alone, leading global tech companies are expected to spend more than $600 billion in capital expenditures on AI infrastructure. Omdia further forecasts that cumulative global data center investment will approach $1.6 trillion by 2030.
The scale of this capital expenditure is virtually unparalleled in the history of technology. Hyperscale tech companies are projected to spend between $660 billion and nearly $700 billion in capex in 2026. AI infrastructure has evolved from a "nice-to-have" technology investment into a strategic expenditure that defines competitive dynamics. The AI Factory market has crossed an irreversible threshold, transforming into a new industrial paradigm characterized by extremely high capital intensity, pronounced geopolitical attributes, and complex engineering barriers.
For investors, understanding the industrial chain structure and capital flows within artificial intelligence (AI) infrastructure is essential to navigating this technology investment cycle. Starting from the three core hardware tracks—GPU, memory, and networking—this article analyzes the investment value and key players in each sector, using the latest market data and industry logic.
GPU: The "Engine" of Computing Infrastructure
GPUs are the most critical computational units in AI infrastructure and currently account for the largest share of capital expenditure. According to Research and Markets, the global AI infrastructure market is expected to grow from $71.88 billion in 2025 to $90.91 billion in 2026, representing a compound annual growth rate of 26.5%. By 2030, this figure is projected to reach $226.95 billion, with GPUs and accelerator systems driving much of this growth.
Market performance underscores the capital chase for computing infrastructure in the GPU sector. In the early hours of June 30 (Beijing time), all three major U.S. stock indices closed higher, with the Nasdaq Composite up 2.07% to 25,820.14. NVIDIA (NVDA) closed at $194.97, up 1.27%, with a market cap of approximately $4.72 trillion. AMD (Advanced Micro Devices) closed at $539.49, up 3.43%, with a market cap of about $879.7 billion. The Philadelphia Semiconductor Index rose 3.83% that day and is up 93.55% year-to-date.
The investment thesis for GPUs is built on two structural factors. First, demand for computing power in large model training and inference continues to climb—from the expansion of model parameters to the scaling of inference deployment, the curve for computing consumption has yet to peak. Second, supply-side entry barriers are extremely high, including architecture design, manufacturing processes, and software ecosystems (such as CUDA), creating multiple moats that allow leading players to maintain strong pricing power for the foreseeable future.
Among notable players, NVIDIA stands as the undisputed leader in AI computing power, with its product roadmap and customer reach serving as industry benchmarks. AMD continues to make strides in both data center CPUs and GPUs, with its stock up 141.3% year-to-date. Cantor Fitzgerald recently raised AMD’s price target to $700. Additionally, Applied Materials (AMAT), a key supplier of semiconductor manufacturing equipment, surged 10.82% to close at $694.64 on June 29, reflecting ongoing market expectations for chip capacity expansion.
Memory: "Locked-in" Capacity and Pricing Power
If GPUs are the "brain" of AI computation, then high-bandwidth memory (HBM) serves as the "neural fibers" that keep the brain running at high speed. During AI training and inference, memory bandwidth directly determines whether the compute units can be fully fed with data—this is known as the "memory wall" bottleneck.
Demand for high-bandwidth memory is soaring as training and inference models scale up. Industry observers note that major capacity has already been locked in by large customers through 2026 and even 2027, leaving little short-term supply flexibility. This supply-demand imbalance gives memory suppliers significantly greater pricing power, order visibility, and profitability.
Market data confirms the robust outlook for the memory sector. Micron Technology (MU) closed at $1,145.28 on June 29, up 1.14%. SK Hynix, another key HBM player, together with Micron and Samsung Electronics, forms the "iron triangle" of global high-bandwidth memory supply. Samsung’s weight in AI infrastructure portfolios is also significant.
The investment logic for memory differs from that of GPUs: it’s not simply a race for technological leadership, but rather a contest of capacity expansion speed and depth of customer relationships. HBM’s complex manufacturing process and long yield ramp mean that vendors who achieve stable large-scale production first will gain a significant first-mover advantage. Furthermore, as AI inference scenarios explode—with inference compute demand expected to surpass training—requirements for memory capacity and bandwidth will only intensify.
Networking: The "Nervous System" and Next Bottleneck for AI
A growing consensus in the networking sector is that as AI clusters scale, network bandwidth is becoming the next bottleneck. In a May report, Bank of America projected that the AI networking market will reach $316 billion by 2030, up from a previous estimate of $240 billion.
This view is grounded in the fact that AI training clusters are evolving from the thousand-card level to tens of thousands or even hundreds of thousands of cards. At this scale, the efficiency of communication between GPUs directly determines overall compute utilization. The industry’s so-called "zombie GPU" effect—where expensive GPUs sit idle waiting for I/O—has become a major headache for hyperscale customers. Evaluation metrics are shifting from pure FLOPS (floating point operations per second) to first-token latency (TTFT) and vector retrieval speed.
During the 2026 Summer Davos, Ericsson’s Global Senior Vice President, Chris Houghton, remarked that the first wave of AI investment flowed into chips and data centers, but the next winners may be the telecom operators laying fiber and building base stations. He likened networking to the "nervous system" of physical AI—large language models are the brain, robots and drones are the body, and the network enables the brain to control the body.
On the networking equipment side, Broadcom (AVGO) is an essential name. As a core supplier of AI networking chips (such as switch ASICs), Broadcom is well positioned to benefit from the demand for upgraded interconnect bandwidth within data centers. Despite a recent pullback in its share price, firms like Jefferies maintain a "strong buy" rating, with an average price target of about $513.58. On June 29, Broadcom closed at $372.45, up 2.04%.
Additionally, Cisco Systems, a traditional networking giant, is actively transforming to meet the new demands of AI data centers, closing up 3.45% at $117.70 on June 29. Dell, as an AI server system integrator, rose 3.78% to $414.61.
Cross-Sector Comparison and Investment Perspectives
From an industry chain perspective, GPUs, memory, and networking each occupy distinct positions:
GPU sector sits at the very top of the value chain, enjoying the highest gross margins and technology premiums, but also facing the highest valuations and market expectations. NVIDIA’s current price-to-earnings ratio (TTM) is about 29.86. Considering its growth rate, this valuation isn’t extreme among tech giants, but any slowdown in demand growth could trigger a valuation reset.
Memory sector is more cyclical. While HBM shortages may mask the traditional DRAM and NAND cycles in the short term, investors should still monitor supply-demand dynamics as large-scale capacity comes online. Current capacity locked through 2026–2027 provides clear mid-term earnings visibility for this sector.
Networking sector is currently less in the spotlight than GPUs or memory, which could mean greater room for positive surprises. Bank of America’s forecast of a $316 billion market by 2030 suggests that networking may outpace current consensus expectations in compound growth over the next several years.
From a risk perspective, all three sectors face common challenges: a potential slowdown in AI capital expenditure, geopolitical disruptions to supply chains, and technological shifts (such as in-memory computing or optical interconnects) that could reshape the industry landscape. Omdia’s survey of over 200 companies identified four core challenges: ROI and time-to-market, digital sovereignty, AI talent shortages, and systemic engineering complexity. These issues will impact the investment return cycle for each sector to varying degrees.
How to Invest in AI Infrastructure on Gate?
For investors seeking exposure to AI infrastructure opportunities, Gate offers a diverse range of entry points.
Gate has listed over 12,500 stocks, including U.S., Hong Kong, and South Korean equities. The platform now fully supports 24/7 trading for U.S., Hong Kong, and Korean stocks—covering pre-market, regular hours, after-hours, overnight, and even weekends. This means investors are no longer constrained by traditional exchange hours and can adjust their positions more flexibly in response to market dynamics.
Regarding AI infrastructure-related stocks, Gate covers many of the core companies mentioned in this article: NVIDIA (NVDA), AMD (AMD), Micron Technology (MU), Broadcom (AVGO), Applied Materials (AMAT), Cisco (CSCO), and Dell (DELL). Through Gate’s stock trading module, investors can allocate and rebalance these holdings in one place.
Conclusion
By 2026, AI infrastructure has moved beyond conceptual narratives and entered a high-stakes race for capital investment. Hyperscale tech companies are pouring hundreds of billions of dollars annually into weaving together GPUs, high-bandwidth memory, and high-speed networks into a global computing infrastructure.
The GPU sector benefits from the highest technical barriers and the most direct mapping to compute demand, making it the most certain direction at present. The memory sector, with its locked-in supply-demand dynamics, offers the clearest mid-term earnings visibility. The networking sector, still underappreciated by the market, may present the greatest opportunity for positive surprises.
Each sector has its own investment rhythm and risk-return profile. Investors can tailor their allocations based on individual risk preferences and investment horizons. Gate’s 24/7 stock trading and broad coverage provide a flexible and efficient toolkit for executing these strategies.
The AI infrastructure build-out is far from over. As Jensen Huang stated at NVIDIA’s 2026 shareholder meeting, AI infrastructure is the largest construction project in human history. In this multi-year wave of capital expenditure, understanding the structure and rhythm of the value chain may deliver more lasting returns than chasing short-term trends.
FAQ
Q1: What are the main sub-sectors covered by AI infrastructure investment?
The three core hardware tracks are: GPUs (graphics processing units, for AI compute acceleration), high-bandwidth memory (HBM, addressing the "memory wall" bottleneck), and data center networking (solving interconnect and communication for large-scale clusters). Additional areas include data center cooling, power systems, and software orchestration layers.
Q2: Why is networking considered the next big opportunity in AI investment?
As AI training clusters scale from thousands to tens or hundreds of thousands of GPUs, communication efficiency between GPUs becomes the key bottleneck for effective compute utilization. Bank of America forecasts the AI networking market will reach $316 billion by 2030. Networking is likened to the "nervous system" of physical AI, forming the essential infrastructure that brings intelligence from the data center into the real world.
Q3: Can AI infrastructure-related U.S. stocks be traded on Gate?
Yes. Gate has listed over 12,500 stocks across U.S., Hong Kong, and South Korean markets, including core companies like NVIDIA (NVDA), AMD (AMD), Micron Technology (MU), and Broadcom (AVGO). The platform supports 24/7 trading, covering pre-market, regular hours, after-hours, overnight, and weekends.
Q4: What are the main risks currently facing AI infrastructure investment?
Key risks include: a slowdown in AI capital expenditure leading to weaker demand, geopolitical disruptions to the chip supply chain, technological paradigm shifts (such as in-memory computing or optical interconnects) impacting the current landscape, and valuation corrections in overheated sectors. Investors should align their allocations with their own risk tolerance.
Q5: What is the market size outlook for AI infrastructure in 2026?
According to Research and Markets, the global AI infrastructure market is expected to grow from $71.88 billion in 2025 to $90.91 billion in 2026, a 26.5% annual growth rate. Other institutions forecast the market could reach $465 billion by 2033. In 2026 alone, leading global tech companies are projected to spend over $600 billion on AI infrastructure capital expenditures.
