The narrative around AI computing power underwent a profound structural shift in 2026. For the past two years, nearly all market speculation centered on Nvidia’s GPUs—who could ramp up production fastest, who secured more H100 allocations, and who would be first to deliver B200 systems. Yet, faced with the combined $725 billion AI infrastructure capital expenditure commitments from Microsoft, Google, Amazon, and Meta in 2026, no single segment can absorb such a massive flow of capital. The real opportunities lie deep within the entire industry value chain.
From chips to servers, memory to storage, optical networks to the real estate that houses data centers, the six major components of AI infrastructure are transitioning from "training-centric" to "full-stack demand explosion."
Segment One: AI Chips—From Hopper to Blackwell to Rubin, Mapping the Supply Landscape
Nvidia stands as the primary driving force behind the entire AI infrastructure investment chain. The key variables in 2026 revolve around generational product shifts and evolving supply dynamics.
According to the latest industry survey by TrendForce, Nvidia’s high-end GPU shipments are projected to increase by nearly 26% year-over-year in 2026, with Blackwell series shipments rising sharply from 61% to 71% of the total, further consolidating its dominance. Meanwhile, the next-generation Rubin architecture has already been sampled to select top partners, but challenges with HBM4 certification and power management mean large-scale shipments won’t begin until the second half of 2026. This generational cadence has a direct consequence: Blackwell B200 and B300 systems are becoming increasingly difficult to obtain. Wedbush Securities’ recent report highlighted that multiple customers are experiencing longer delivery cycles for Blackwell systems, stating, "Supply tightness at this stage of the lifecycle is unprecedented."
From a revenue standpoint, Nvidia has indicated that the combined visible revenue from the Blackwell and Vera Rubin platforms in 2025–2026 could reach $500 billion. HSBC analyst Frank Lee estimates Nvidia’s sales for Q1 and Q2 of fiscal 2026 at $42.2 billion and $55.4 billion, respectively—both exceeding the market consensus of $42 billion and $46.2 billion.
A frequently overlooked growth signal comes from the inference side. Nvidia is aggressively expanding into AI inference applications, with its new LPU solution expected to see demand for hundreds of thousands of units in 2026, doubling in 2027. For investors, Nvidia is no longer just a "training chip company"—the narrative is shifting from training to inference, broadening its valuation ceiling.
Segment Two: AI Servers—Divergent Growth Paths Among the Three Major OEM Giants
Servers are the core vehicles that convert chips into deliverable computing power. In 2025, global server market revenue hit a record $444 billion, with Q4 alone reaching $125.3 billion—a 52.4% year-over-year increase. Gartner projects server spending will grow another 36.9% in 2026.
Within this market, Dell, HPE, and Supermicro—the three leading suppliers—showed markedly different results in their Q2 2026 financials.
Dell delivered a breakout performance. AI-optimized server revenue for the quarter reached $16.1 billion, up 757% year-over-year, surpassing its total AI shipment revenue of $9.8 billion for all of 2025 in just one quarter. The backlog is even more noteworthy—AI-related orders for the quarter totaled $24.4 billion, with a record backlog of $51.3 billion and over 5,000 customers. Bank of America projects global AI server revenue in 2026 will reach $496 billion, with Dell holding about a 12% market share.
HPE follows a completely different growth logic. Q2 2026 fiscal revenue hit $10.7 billion, up 40% year-over-year, with AI system sales rising 66% to $1.54 billion. Unlike Dell’s scale, 61% of HPE’s $5.9 billion AI order backlog comes from government and large enterprise clients, offering higher margins but slower growth. Bank of America expects HPE’s AI server revenue for fiscal 2026 to reach $6.5 billion.
Supermicro pursues a differentiated technology strategy. Quarterly revenue reached $10.24 billion, with its core strength in direct liquid cooling technology. As single-rack power density in data centers jumps above 240kW, liquid cooling is shifting from an "optional" to a "mandatory" solution, and Supermicro commands about 70% of this niche market.
Evercore predicts Dell and HPE will continue to attract outsized investment from Tier-2 service providers, sovereign entities, and enterprises, with server demand drivers spreading from "top hyperscale customers" to the "long tail of enterprise clients."
Segment Three: Memory—Dual Demand Surge for DRAM and HBM
AI data centers are seeing memory demand expand at an unprecedented pace. The logic is straightforward: bigger models, more parameters, and higher concurrent inference all drive exponential growth in DRAM and HBM (high-bandwidth memory) requirements.
Micron is the bellwether for this trend. Since 2026 began, Micron’s stock price has soared over 237%, nearly ninefold in the past year, with its market cap surpassing $1 trillion. In the latest quarter, revenue climbed 74% year-over-year, net profit more than doubled, and memory prices rose about 40% year-to-date.
Multiple research institutions have issued highly consistent bullish forecasts for memory price trends. Citi analysts expect DRAM prices to rise at least through next year; Gartner offers an even more aggressive short-term outlook—DRAM prices are projected to surge 125% in 2026 alone, with storage chip prices potentially jumping 234%. Crucially, despite soaring prices, buyers are still scrambling for capacity—Micron, Samsung, and SK Hynix are nearly sold out through next year.
AI is shifting from "training-first" to "inference-first," and inference requires far more memory capacity than training. This structural change suggests the memory segment’s boom cycle has significant room to run.
Segment Four: Storage—From "Overlooked Segment" to One of 2026’s Strongest Growth Themes
By 2026, the storage sector has become one of the fastest-growing areas in the AI supply chain. In the US stock market, Seagate, SanDisk, and Western Digital have seen their share prices rise sharply this year, driven by AI infrastructure investment and surging demand for high-capacity storage.
Evercore’s January 2026 report notes that storage and networking are expected to claim a larger share of IT spending, as enterprises face worsening data latency and storage architecture bottlenecks. Storage market growth in 2026 is projected to jump from about 4% in 2025 to roughly 9%.
Seagate, a leader in hard drives, is seeing its products revalued for persistent AI data storage and large-scale cold data storage. SanDisk’s newly launched $2,000 2TB storage card highlights the pricing power of high-end storage solutions.
From an investment perspective, memory and storage are still relatively undervalued—forward P/E ratios remain in single digits or low teens, while the AI-driven demand cycle is far from peaking. As the physical anchor for AI data flows, storage offers high certainty of benefit and a strong valuation safety margin worth ongoing attention.
Segment Five: Optical Networks—Essential for AI Data Center Interconnectivity and Bandwidth Expansion
Optical networking is the most "invisible" yet indispensable component of AI infrastructure. As GPU clusters scale from tens of thousands to hundreds of thousands of units, inter-chip bandwidth, intra-data center fiber transmission, and inter-data center data exchange efficiency all become bottlenecks for overall compute output. Ciena is one of the most closely watched players in this space. The company’s Q4 2025 financials exceeded expectations, with UBS raising its price target to $230, Argus to $280 with a buy rating, and Rosenblatt to $305, positioning Ciena as a key participant in AI networking for data center interconnects.
Lumentum, a core supplier of optical communication components, is also benefiting from rising demand for internal data center optical modules. As 400G modules give way to 800G and even 1.6T, the optical networking segment not only enjoys a strong boom cycle but also clear technological upgrade dividends.
Evercore’s main thesis: investment is spreading from compute to storage and network infrastructure. This "diffusion effect" means optical networking firms like Ciena and Lumentum will claim greater profit share within the industry value chain.
Segment Six: Data Center REITs—The "Landlord" Logic of AI’s Physical Layer
At the end of the AI infrastructure investment chain lies the most physical segment—data center REITs. As Microsoft, Google, Amazon, and Meta continue to ramp up capital spending, the physical space, power capacity, and connectivity of data centers become scarce resources.
Equinix’s latest regulatory filings show that institutional investors like Capital Research Global Investors are still increasing their holdings, while Vise Technologies purchased over 1,000 shares, reflecting professional investors’ ongoing recognition of data center assets as strategic components in diversified portfolios.
Equinix’s core value as a REIT operator is providing colocation and interconnect services to enterprises, cloud providers, and network operators, generating stable recurring rental income and returning value to investors via dividends. Digital Realty, meanwhile, focuses on hyperscale data centers, forming a complementary competitive landscape with Equinix.
For investors seeking cash flow, data center REITs offer the rare dividend yield attribute within the AI theme, giving them unique risk management value in portfolio allocation.
Gate Stock Trading: Key Advantages and Operational Steps
Compared to traditional brokers, Gate’s stock trading feature offers distinct advantages: fractional share investing and low entry barriers. You don’t need to buy whole shares to participate in high-priced AI stocks, significantly lowering the capital threshold for retail investors. Unified asset allocation in a single account allows users to configure cryptocurrencies, US stocks, tokenized equities, and ETFs, eliminating the hassle of switching between multiple platforms.
To get started, users can follow these steps: log in to your Gate account and complete identity verification. On the spot trading page, switch the market to the "stock token" category, search for your target asset, choose from spot, perpetual, or Alpha trading modes, enter quantity and price, and confirm your order. For those seeking genuine US stock trading, the platform now supports direct USDT purchases of US stocks via compliant brokerage channels.
Risk Factors and Structural Reflections on Industry Chain Investment
The AI infrastructure value chain is not a risk-free path to deterministic growth. For a complete investment rationale, the following risk factors must be systematically considered.
Supply-side risk: Challenges with HBM4 certification and power management for Nvidia’s Rubin series may slow shipment pace. Any disruption in the supply chain can ripple through server OEMs and affect the entire chain.
Inventory risk from generational shifts: Rapid iterations from Hopper to Blackwell to Rubin increase price pressure and inventory write-down risk for previous-generation products.
Valuation risk: With Micron’s market cap surpassing $1 trillion and cumulative gains over 237%, it’s worth monitoring whether market optimism for the memory segment is already overextended. Wedbush maintains a positive outlook on Nvidia, but their supply chain checks also highlight the paradox that "no enterprise customer’s AI deployment has slowed or changed due to alternative solutions"—a warning sign to watch.
Demand structure risk: Current AI infrastructure spending is highly concentrated among four hyperscale cloud providers. If any one of them cuts budgets, demand expectations across the supply chain will need to be reassessed.
Macro and policy risk: Ongoing US-China tech controls on advanced chip exports disrupt regional supply chain distribution. The H200’s second wave of demand, triggered by new export policies, underscores how policy variables can significantly impact supply and demand dynamics.
Conclusion
The investment logic for AI infrastructure is undergoing profound transformation. From Nvidia’s Blackwell chip shipments surpassing 70%, to Dell’s AI server backlog exceeding $50 billion, Micron’s market cap breaking $1 trillion, and sustained institutional accumulation in Equinix data centers—the cross-validation across six segments points to a single conclusion: certainty in AI computing power spending is shifting from "chip supply" to "full-stack infrastructure."
