From Blackwell to AI Factories: The Rebuilding of the Computing Power Industry Chain Behind NVIDIA’s $75.2 Billion Data Center Revenue

The global AI computing power industry in 2026 is undergoing a comprehensive overhaul, from underlying chips to upper-layer applications. Marked by NVIDIA's Q1 FY2027 earnings report on May 20, 2026, this leading AI chip manufacturer shattered industry perceptions with a quarterly revenue of $81.6 billion, an 85% year-over-year increase. Its data center business reached $75.2 billion, up 92% YoY, and the Blackwell architecture officially became the main platform for global AI training and inference. At this point in time, SEMI China President Feng Li pointed out in March that global AI infrastructure spending in 2026 will reach $450 billion, 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, from US stock trading to Hong Kong stock trading, and even offering "direct IPO" services, building a investment channel connecting global investors to core AI computing power industry targets.

NVIDIA FY2027 Q1 Earnings: Computing Demand Continues to Accelerate

As of Q1 FY2027 ending April 26, 2026, NVIDIA reported $81.6 billion in revenue, a 20% quarter-over-quarter increase, marking the fourteenth consecutive quarter of sequential growth, and the third year of accelerating YoY growth each quarter. This data itself indicates that the sustained expansion of AI infrastructure demand is not a short-term pulse. Notably, free cash flow reached $49 billion, up from $35 billion in the previous quarter. The company announced an additional $80 billion stock repurchase authorization and a significant increase in quarterly dividends from $0.01 to $0.25. Judging by the growth rate of shareholder returns and free cash flow, the AI chip industry has shifted from an early-stage high-investment expansion phase to a mature growth stage capable of self-sustaining development.

Data centers are the most prominent segment in this earnings report. Data center revenue hit $75.2 billion, up 92% YoY and 21% sequentially. Within this, data center computing revenue was $60.4 billion, and data center networking revenue was $14.8 billion, with YoY growth nearly doubling. NVIDIA CFO Colette Kress explicitly stated during the earnings call that the growth mainly stems from continued volume increases of the Blackwell architecture, with demand for GB300 and NVL72 systems being "especially strong." From an application perspective, revenue from hyperscale cloud service providers reached $38 billion, roughly half of total data center revenue, while the ACIE sub-business—covering AI Cloud, industrial, and enterprise clients—earned $37 billion, up 31% QoQ, with AI cloud revenue tripling YoY. This indicates that, beyond a few cloud giants, more AI-native cloud service providers and industry clients are gradually becoming new growth drivers for computing power demand.

Jensen Huang summarized this transformation at the earnings call as the "scaling of AI factories." "Intelligent AI has arrived. It is doing real work, generating tangible value, and expanding rapidly within companies and industries," he described the current industry shift. What he describes is not just a technological trend but a systemic change in industry structure—global data centers are shifting from training clusters to continuous-operating AI factories.

Using order data as a forward-looking indicator better illustrates this point. At GTC in March 2026, NVIDIA's updated backlog data showed that cumulative orders for Blackwell and Rubin (including network components) from 2025 to 2027 exceeded $1 trillion, excluding contributions from Hopper, CPUs, CPX, and LPX. Meanwhile, U.S. major hyperscale cloud providers' capital expenditure forecasts for 2026–2027 were raised to $812 billion and $968 billion, directly supporting NVIDIA's data center procurement needs.

Full View of the AI Computing Power Industry Chain: Systemic Reconstruction from Chips to Data Centers

Focusing solely on NVIDIA’s performance growth might overlook the structural features of this industry upgrade. The key lies in the fundamental change in product form—from selling a single chip to providing an entire rack system.

The Blackwell platform has become the main driver of current data center revenue, achieving NVIDIA’s fastest product adoption rate in history. The competition in the AI server market has shifted from single cards or single machines to system-level competition across entire cabinets and data centers. After mass production of GB300 and NVL72, the price of full-rack AI solutions has surged from the previous $200k level to between $3 million and $7 million. This change is driven by the increasing system-level coordination required for large model training and inference, where performance of a single chip is no longer the sole criterion. Memory bandwidth, network interconnects, thermal management, and even cabinet design and delivery capabilities have become critical dimensions of system competitiveness.

From the shipment structure perspective, TrendForce and Jibang Consulting’s forecast provides quantitative support for this judgment: in 2026, Blackwell solutions will account for 71% of NVIDIA’s high-end GPU shipments, with the GB300/B300 series as core products. Wedbush Securities’ early June 2026 report further confirms this trend, noting that Blackwell system supply is becoming increasingly tight, with delivery cycles lengthening due to HBM capacity bottlenecks and advanced packaging capacity constraints. The firm also pointed out that no enterprise customer has slowed or changed AI deployment plans due to market volatility, reaffirming demand-side certainty for computing power investments.

From an upstream perspective, SEMI China President Feng Li’s March 2026 analysis highlighted that global AI infrastructure spending in 2026 will reach $450 billion, with inference computing power exceeding 70% for the first time, driving strong demand for GPUs, HBM, and high-speed network chips, ultimately propagating to wafer fabs, advanced packaging, and equipment materials in the semiconductor upstream chain. The industry chain analysis indicates that the growth in computing power demand involves at least three levels: chip design and manufacturing (NVIDIA, TSMC, Samsung), AI servers and system integration (Dell, Foxconn Industrial Internet, Supermicro), and cloud services and computing operations (Microsoft, AWS, Google Cloud, IREN, etc.). On May 27, 2026, AI data center operator IREN announced a purchase of approximately $1.6 billion worth of NVIDIA’s liquid-cooled Blackwell systems from Dell Technologies, fulfilling a five-year, $3.4 billion cloud AI service contract. The new systems are expected to be operational by early 2027. IREN, previously transformed from Bitcoin mining to AI data center operations, exemplifies how the computing power industry chain is attracting broader new entrants.

UBS’s latest report in May 2026 raised NVIDIA’s target price from $245 to $275, with the core reasoning that even if Rubin platforms face short-term delays due to HBM4 certification and cooling issues, demand for Blackwell is sufficient to offset a 1–2 month delay. In 2026, Blackwell Ultra will constitute 70% of product mix, with Rubin at 22%. More importantly, their long-term outlook suggests that in 2027, Rubin will contribute about 68% of revenue. Vera Rubin systems have already secured numerous pre-orders, with mass production starting in Q3 2026, offering 35 times the inference throughput of Blackwell and increasing AI factory revenue potential tenfold. This indicates that supply ceilings in the computing power industry chain will continue to be pushed higher into 2027 and beyond.

From a market size perspective, the global generative AI market reached $120 billion in 2026, with large model parameters exceeding 100 trillion, and training costs surpassing $5 million per session. Against this backdrop, the expansion path for AI chips is becoming clearer: the global AI chip market has exceeded $85 billion, with cloud training chips accounting for 42%, edge inference chips 38%, and terminal chips 20%. China’s AI chip market is expected to surpass RMB 160 billion in 2026, driven by policy support and supply chain security accelerating domestic substitution.

Long-term analyst forecasts for NVIDIA’s stock also reflect this structural optimism. After earnings release, Wall Street firms collectively raised target prices: Melius Research from $380 to $400, Bank of America Global Research from $300 to $350, KeyBanc to $300, Morgan Stanley to $288. Using an average target of about $307, NVIDIA’s market cap would surpass $7.4 trillion; at $400, it would approach $9.68 trillion. KeyBanc specifically pointed out that increased shipments of Blackwell GPUs could generate an incremental $5–7 billion in revenue. Cantor Fitzgerald analysts further noted that Blackwell’s capacity for FY2026 is fully booked, with backlog orders accumulating into 2027 and 2028. These differing views mainly concern timing rather than direction: the focus is on whether the current growth rate can be maintained into 2027–2028, rather than whether overall AI computing demand will cool down.

Gate Stock Trading Channel: A New Choice for AI Computing Power Investment

For global investors interested in the AI computing power industry chain, directly holding stocks of NVIDIA, TSMC, Microsoft, Amazon, and other core targets has always been the most straightforward way to capture industry growth. However, traditional securities accounts often face regional restrictions, fiat currency exchange hurdles, and capital transfer inefficiencies—especially for crypto users holding assets mainly in USDT or other stablecoins, who must go through the cumbersome process of exiting the crypto ecosystem, cross-border transfers, and re-opening accounts.

Gate is addressing this pain point through product innovation. On June 11, 2026, Gate officially launched Hong Kong stock trading services, allowing users to trade over 1,000 Hong Kong-listed stocks directly with USDT stablecoins, including Tencent, Alibaba, Meituan, Xiaomi, BYD, HSBC, and more, without opening separate brokerage accounts or converting funds into HKD. This marks a substantive step in Gate’s transition from a pure crypto exchange to a multi-asset investment platform.

For the U.S. stock market, Gate has partnered with compliant broker Alpaca to offer real stock spot trading, covering over 10k global stocks and ETFs listed on NYSE and NASDAQ. Users can invest directly using USDT liquidity in their Gate accounts, supporting tech blue chips like NVIDIA, Apple, Google, with minimums as low as 0.01 shares and starting around $10 USDT, far below traditional broker’s full-share trading requirements.

In terms of fees, Gate integrates VIP tier systems into stock trading, where holding just 2,000 USDT in assets upgrades users to VIP status, enjoying a minimum trading fee of 0.023%, with no platform fees, commissions, or hidden charges. Dividends and corporate actions are automatically distributed in USDT equivalents, and the unified account system allows management of both crypto assets and stock portfolios within a single interface, enabling efficient capital allocation across markets.

For users interested in pre-IPO opportunities, Gate launched the “Direct IPO” (IPO Access) service in June 2026, allowing users to submit subscription applications before companies go public and trade directly on Gate’s stock market after allocation. The first project, SpaceX (SPCX), has completed allocation, with more high-growth tech IPOs to follow, providing investors with early access to promising private companies.

Operationally, the process for trading U.S. or Hong Kong stocks on Gate has become quite streamlined. Users need to update the Gate app to the latest version (8.23.5 or above), go to the “TradFi” section, select “Stocks,” choose either U.S. or Hong Kong stock zones, transfer USDT from spot or unified accounts into the dedicated stock account, then search for target stocks like NVIDIA (NVDA) or related ETFs, input order quantity (minimum 0.01 shares), and submit. Hong Kong trading follows HKEX hours (Monday to Friday, 9:30–16:00 HKT), while U.S. trading has expanded to a 16×5 schedule. After completing trades, all holdings and orders can be viewed centrally within the Gate unified account. The integrated system ensures funds can be dynamically allocated across different asset classes in real time, greatly improving multi-asset strategy execution efficiency.

Conclusion

The AI computing power industry chain is experiencing a full upgrade from chips to system delivery, with NVIDIA’s FY2027 Q1 revenue of $81.6 billion and data center revenue of $75.2 billion merely reflecting this systemic change in financial data. The key trend is that annual AI infrastructure expenditure is rapidly approaching the trillion-dollar mark, with training, post-training, and inference demands expanding simultaneously, turning AI factories from concept to large-scale construction. Against this macro backdrop, global investor demand for core targets in the computing power industry chain continues to grow.

The launch and ongoing iteration of Gate’s stock trading features directly respond to this shift. Whether in U.S. or Hong Kong markets, secondary trading or pre-IPO subscriptions, Gate is gradually building a comprehensive investment platform covering crypto assets and traditional financial products. For investors watching NVIDIA’s stock trends, Blackwell GPU shipment pace, or AI data center capital expenditure scales, Gate offers a compliant channel to directly access shares of top global tech companies using USDT. As the AI computing power industry chain continues to expand across chip design, server integration, and data center operations, this channel will also carry increasing investment value and market attention.