The global AI computing power industry in 2026 is undergoing a comprehensive transformation, spanning from foundational chips to upper-layer applications. Marked by NVIDIA’s Q1 FY2027 earnings report released on May 20, 2026, the world’s leading AI chipmaker shattered industry expectations with $81.6 billion in quarterly revenue—an 85% year-over-year increase. Its data center business surged 92% to $75.2 billion, and the Blackwell architecture officially became the dominant platform for global AI training and inference. At this pivotal moment, SEMI China President Feng Li noted in March that global AI infrastructure spending is set to reach $450 billion in 2026, with inference computing power accounting for over 70% for the first time. Meanwhile, Gate is rapidly advancing the integration of traditional finance and crypto assets—expanding from US stock trading to Hong Kong stock trading and launching "Direct-to-IPO" services—building an investment channel that connects global investors directly to the core assets of the AI computing power value chain.
NVIDIA Q1 FY2027 Earnings: Demand for Computing Power Still Accelerating
For the first quarter of fiscal year 2027, ending April 26, 2026, NVIDIA posted $81.6 billion in revenue, up 20% quarter-over-quarter. This marks the fourteenth consecutive quarter of sequential growth and the third straight year of accelerating year-over-year growth—a clear indicator that demand for AI infrastructure continues to expand and is not a short-lived surge. Notably, free cash flow reached $49 billion, up from $35 billion in the previous quarter. The company responded by announcing an additional $80 billion in share repurchase authorization and a significant increase in its quarterly dividend from $0.01 to $0.25 per share. Judging by the growth in shareholder returns and free cash flow, the AI chip industry is transitioning from an early, high-investment expansion phase to a mature stage characterized by self-sustaining growth.
The data center segment stood out in this earnings report. Data center revenue hit $75.2 billion, up 92% year-over-year and 21% quarter-over-quarter. Within this, data center compute revenue reached $60.4 billion and networking revenue $14.8 billion—both nearly doubling year-over-year. NVIDIA CFO Colette Kress emphasized during the earnings call that growth was mainly driven by the ramp-up of Blackwell architecture, with "especially strong" demand for GB300 and NVL72 systems. From an application perspective, hyperscale cloud service providers generated $38 billion in revenue, about half of the data center total, while the ACIE subsegment—including AI Cloud, industrial, and enterprise clients—brought in $37 billion, up 31% quarter-over-quarter. AI cloud revenue more than tripled year-over-year. This indicates that, beyond a handful of cloud giants, a broader array of AI-native cloud providers and industry clients are emerging as new growth drivers for computing power demand.
Jensen Huang described this shift as the large-scale construction of "AI factories" during the earnings call. "Agent AI has arrived. It’s doing real work, creating real value, and expanding rapidly across companies and industries," he said, capturing the ongoing transformation. He was referring not just to a technological trend, but to a systemic change in industry structure—global data centers are evolving from training clusters into continuously operating AI factories.
Order data provides a forward-looking window into these trends. At the March 2026 GTC conference, NVIDIA’s updated backlog figures showed cumulative orders for Blackwell and Rubin (including networking components) exceeding $1 trillion for 2025–2027. This figure excludes contributions from Hopper, CPU, CPX, and LPX. Meanwhile, the four major US hyperscale cloud providers raised their capital expenditure forecasts for 2026 and 2027 to $812 billion and $968 billion, respectively, providing quantifiable procurement support for NVIDIA’s data center business.
The AI Computing Power Value Chain: Systemic Reconstruction from Chips to Data Centers
Focusing solely on NVIDIA’s performance risks missing the structural features of this industry upgrade. The key lies in a fundamental change in product form—moving from selling individual chips to delivering entire rack systems.
The Blackwell platform has become the main driver of current data center revenue, setting a record for the fastest product adoption in NVIDIA’s history. Competition in the AI server market has shifted from single cards and standalone machines to full racks and even entire data center systems. With the mass production of the GB300 NVL72, the price for rack-level AI solutions has jumped from the previous $200,000 range to between $3 million and $7 million. This shift is driven by the increasing need for system-level synergy in large model training and inference. Chip performance alone is no longer the sole benchmark; memory bandwidth, network interconnects, thermal management, and even rack design and delivery capabilities have all become critical measures of system competitiveness.
Shipment data supports this view. TrendForce’s forecast indicates that Blackwell solutions will account for 71% of NVIDIA’s high-end GPU shipments in 2026, with the GB300/B300 series as the core product line. A June 2026 report by Wedbush Securities further confirmed this trend, noting that Blackwell system supply is tightening, lead times are lengthening, and constraints stem from HBM capacity and advanced packaging bottlenecks. The firm also pointed out that no enterprise customers have slowed or altered their AI deployment plans due to market fluctuations, reaffirming the certainty of computing power investment from the demand side.
A further upstream perspective comes from SEMI China President Feng Li’s March 2026 analysis. She projected that global AI infrastructure spending will reach $450 billion in 2026, with inference computing power surpassing 70% for the first time. This surge will drive robust demand for GPUs, HBM, and high-speed networking chips, eventually filtering up to wafer fabs, advanced packaging, and equipment/materials in the semiconductor supply chain. Analyzing the value chain, growth in computing power demand involves at least three tiers: chip design and manufacturing (NVIDIA, TSMC, Samsung); AI server and system integration (Dell, Foxconn Industrial Internet, Supermicro); and cloud services and computing power operations (Microsoft, AWS, Google Cloud, IREN, and other data center operators). On May 27, 2026, AI data center operator IREN announced the purchase of approximately $1.6 billion in NVIDIA air-cooled Blackwell systems from Dell Technologies to fulfill a five-year, $3.4 billion cloud AI services contract. The new systems are expected to go online in early 2027. IREN previously pivoted from Bitcoin mining to AI data center operations, illustrating how the computing power value chain is attracting a wider range of new entrants.
UBS raised its NVIDIA price target from $245 to $275 in its May 2026 report. The core logic: even if the Rubin platform faces short-term delays due to HBM4 certification and cooling issues, Blackwell demand is sufficient to bridge a one- to two-month gap. In the 2026 product mix, Blackwell Ultra is expected to account for 70% and Rubin 22%. More importantly, UBS projects that in 2027, the Rubin platform will contribute about 68% of revenue. In fact, the Vera Rubin system has already secured a large number of preorders, with mass production set to begin in Q3 2026. Its inference throughput is 35 times higher than Blackwell, and its AI factory revenue potential is tenfold. This data suggests that in 2027 and beyond, the supply ceiling for the computing power value chain will continue to rise.
From a market size perspective, the global generative AI market is expected to reach $120 billion in 2026, with large model parameter counts surpassing 100 trillion and single training runs costing over $5 million. Against this backdrop, the growth trajectory for AI chips is clear: the global AI chip market has exceeded $85 billion, with cloud training chips accounting for 42%, edge inference chips 38%, and endpoint chips 20%. China’s AI chip market is projected to surpass RMB 160 billion in 2026, with domestic substitution accelerating under policy and supply chain security drivers.
Analysts’ long-term forecasts for NVIDIA’s stock price reflect this structural optimism. Following the earnings release, Wall Street firms raised their price targets: Melius Research from $380 to $400, BofA Global Research from $300 to $350, KeyBanc to $300, and Morgan Stanley to $288. Based on an average target of about $307, NVIDIA’s market cap would exceed $7.4 trillion; at $400, it would approach $9.68 trillion. KeyBanc specifically noted that increased Blackwell GPU shipments could add $5–7 billion in incremental revenue. Cantor Fitzgerald analysts further pointed out that Blackwell’s FY2026 capacity is fully booked, with backlogs for 2027 and 2028 continuing to accumulate. The main debate among these firms centers on timing, not direction: the focus is whether the current growth rate can be sustained through 2027–2028, not whether overall AI computing power demand will cool.
Gate Stock Trading Channel: A New Avenue for AI Computing Power Investment
For global investors focused on the AI computing power value chain, directly holding shares of core players like NVIDIA, TSMC, Microsoft, and Amazon has long been the most straightforward way to capture industry growth. However, traditional brokerage accounts are often subject to geographic restrictions, fiat currency exchange hurdles, and inefficient capital transfers—especially for crypto users whose main assets are stablecoins like USDT. Entering traditional stock markets typically requires exiting the crypto ecosystem, cross-border fund transfers, and a series of cumbersome account-opening procedures.
Gate is addressing these pain points through a series of product innovations. On June 11, 2026, Gate officially launched Hong Kong stock trading services, allowing users to trade over 1,000 Hong Kong-listed stocks—including Tencent Holdings, Alibaba Group, Meituan, Xiaomi, BYD, and HSBC—directly with USDT stablecoins. There’s no need to open a separate brokerage account or convert funds to HKD. This marks a significant step in Gate’s evolution from a pure crypto exchange to a multi-asset investment hub.
For the US stock market, Gate has formed a strategic partnership with compliant broker Alpaca, providing users with real US stock spot trading services. This covers more than 10,000 global stocks and ETFs listed on the NYSE and NASDAQ. Users can invest directly using USDT liquidity from their Gate accounts, with support for blue-chip tech stocks like NVIDIA, Apple, and Google. Minimum investment starts at just 0.01 shares, or as little as $10 USDT, far below the whole-share requirements of traditional US brokers.
On the fee front, Gate integrates its VIP tier system into stock trading. Users need only hold $2,000 USDT to upgrade to VIP status and enjoy exclusive trading fees as low as 0.023%, with no platform fees, commissions, or hidden charges. Dividends and corporate actions are automatically distributed in USDT equivalents. The unified account system lets users manage both crypto assets and stock portfolios in a single interface, enabling efficient allocation of funds across different markets.
For users seeking pre-IPO investment opportunities, Gate launched its "IPO Access" service in June 2026. This allows users to submit subscription requests before a company is officially listed and, upon allocation, trade directly on the Gate stock market. The first project, SpaceX (SPCX), has completed its allocation, and more high-profile tech IPOs are on the way—offering investors early access to high-growth, pre-listing companies.
The process for trading US or Hong Kong stocks on Gate is now well established. Users should update the Gate app to the latest version (8.23.5 or above), enter the "TradFi" section and select "Stocks." Choose either the US or Hong Kong stock zone as needed. Transfer USDT from the spot or unified account to the dedicated stock account, then search for and select target stocks such as NVIDIA (NVDA) or related ETFs, enter the order quantity (minimum 0.01 shares), and submit the order. Hong Kong stock trading follows HKEX hours (Monday to Friday, 9:30 a.m. to 4:00 p.m. HKT), while US stock trading now covers a 16×5 time window. All holdings and order records can be viewed centrally in the Gate unified account interface. This unified system ensures real-time fund allocation across asset classes in response to market changes, greatly improving multi-asset strategy execution efficiency.
Conclusion
The AI computing power value chain is undergoing a full-scale upgrade, from foundational chips to system delivery. NVIDIA’s $81.6 billion in Q1 FY2027 revenue and $75.2 billion in data center income are merely financial reflections of this structural transformation. The truly critical trend is that annual AI infrastructure spending is rapidly approaching the trillion-dollar mark, with training, post-training, and inference needs expanding in tandem, and AI factories moving from concept to large-scale reality. Against this macro backdrop, global investors’ demand for exposure to core assets in the computing power value chain continues to rise.
The launch and ongoing evolution of Gate’s stock trading features directly address this shifting demand. Whether it’s US or Hong Kong stocks, secondary market trades, or pre-IPO subscription opportunities, Gate is steadily building a comprehensive investment platform that bridges crypto assets and multiple traditional financial products. For investors tracking NVDA’s share price, Blackwell GPU shipment cycles, or the scale of AI data center capital expenditures, Gate provides a compliant channel to access equity in leading global tech companies directly with USDT. As the AI computing power value chain continues to expand across chip design, server integration, and data center operations, this channel itself will carry even greater investment value and market attention.
