In Q2 FY2027, data-center revenue reached $89.0 billion, up 117% year over year, driven by the ramp of Blackwell Ultra; Blackwell accounted for the majority of system shipments. NVIDIA reported that next-generation Vera Rubin began production shipments in Q3 FY2027, with both Blackwell and Rubin expected to ship going forward, although supply constraints remain.
NVIDIA increased supply and capacity commitments to $279 billion as of July 26, 2026, up from $119 billion in the prior quarter, primarily for data-center infrastructure, memory, and manufacturing capacity supporting current and future architectures.
During Q2, NVIDIA introduced a business model with select AI cloud partners under which the partners procure NVIDIA infrastructure and NVIDIA commits to cloud services. These typically six-year commitments totaled $36 billion at quarter-end and may generate future revenue-sharing income if specified conditions are met.
China-related constraints continued to limit operations: H200 shipments under U.S. licenses represented less than 1% of the latest quarter’s Data Center revenue, while NVIDIA recorded a $0.4 billion first-half charge for H200 excess inventory and purchase obligations. After quarter-end, NVIDIA reported guarantees of up to $105 billion supporting SB Energy data-center leases for approximately 4.25 gigawatts of IT load serving OpenAI, with the first phase expected in fiscal 2029.
(Filed on May 20, 2026)
+85.2%
74.9%
$13.2B
Data Center operations accelerated: first-quarter revenue reached $75.2 billion, up 92% year over year, driven by the ramp of Blackwell 300 systems and demand for InfiniBand, Spectrum-X Ethernet, and NVLink; Blackwell represented the majority of system shipments. Hyperscalers contributed approximately 50% of Data Center revenue, with the remainder diversified across AI Clouds, industrial, enterprise, and sovereign customers.
NVIDIA expects its next-generation Rubin platform to begin shipping in the second half of fiscal 2027. The company is managing concurrent architecture introductions, which it said may create production, supply-demand, quality, yield, inventory, warranty, and revenue-timing challenges.
NVIDIA invested $18.6 billion in private companies and infrastructure funds during the quarter, including investments in AI model makers, and reported $119 billion of manufacturing, supply, and capacity commitments as of April 26, 2026, primarily to support current and future data-center-scale products. It also had $30 billion of multi-year cloud-service commitments, primarily for research and development.
China data-center operations remained constrained: no Data Center Hopper products were shipped to China during the quarter, versus $4.6 billion of such shipments a year earlier. Although U.S. licenses beginning in February 2026 permit small H200 shipments to specified China-based customers, NVIDIA had generated no revenue under the program by quarter-end, and stated it was effectively foreclosed from competing in China’s data-center market.
(Filed on February 25, 2026)
+73.2%
75.0%
$10.6B
Annual context for fiscal 2026: NVIDIA’s Data Center revenue was $193.7 billion, with compute revenue up 59% and networking revenue up 142% year over year; Blackwell platforms represented the majority of Data Center revenue, and the company began shipping Blackwell Ultra production units, including GB300, in the second quarter.
NVIDIA unveiled the Rubin platform during fiscal 2026, targeting agentic AI and reasoning workloads and expecting production shipments to begin in the second half of fiscal 2027. The company also accelerated releases of its open AI model platforms, including Nemotron for agentic AI and Cosmos for physical AI; these are roadmap and ecosystem developments rather than completed commercial shipments.
In December 2025, NVIDIA entered a non-exclusive license agreement with Groq for language-processing-unit technology and hired certain Groq employees. The arrangement involved $13.0 billion paid at closing and $4.0 billion payable within one year; no customer contracts, existing products, or equity interests were acquired.
NVIDIA expanded its AI infrastructure commitments and supply-chain footprint during fiscal 2026: manufacturing is being extended beyond Asia into the U.S. and Latin America, while outstanding manufacturing, supply and capacity commitments reached $95.2 billion, primarily payable through fiscal 2027. The company also committed $27 billion to multi-year cloud services for research and development and provided $3.5 billion in facility lease guarantees to early-stage partners.
(Filed on November 19, 2025)
+62.5%
73.4%
$11.5B
Data Center revenue reached $51.2 billion in Q3 FY2026, up 66% year over year, with Blackwell Ultra the leading architecture across customer categories; production shipments of Blackwell Ultra platforms, including GB300, began in Q2 and prior-generation Blackwell demand remained strong.
Data Center networking revenue rose 162% year over year to $8.2 billion, driven by NVLink compute fabric for GB200 and GB300 systems, along with XDR InfiniBand and Ethernet adoption; NVIDIA said shipment timing and supply availability varied sequentially.
H20 sales were insignificant in Q3 after U.S. export-license requirements reduced demand; licenses granted in August permitted shipments to certain China customers, but cumulative H20 revenue under those licenses was only approximately $50 million as of the filing. The company had recorded a $4.5 billion H20 excess-inventory and purchase-obligation charge in Q1, not Q3.
NVIDIA continued expanding capacity for AI infrastructure, with $50.3 billion of manufacturing, supply and capacity commitments—substantially payable through FY2027—and $26 billion of multi-year cloud-service commitments. During Q3 it committed to invest $5 billion in Intel, subject to regulatory approval, and entered a letter of intent regarding a potential OpenAI investment; it also entered a facility-lease guarantee for a partner with maximum exposure of $860 million over five years.
