NVIDIA revenue up 85% on Blackwell ramp; China foreclosure persists, $18.6B invested in AI ecosystem

The baseline disclosed a new H20 export licensing requirement imposed April 9, 2025, and a resulting $4.5 billion inventory charge. The current filing does not repeat this announcement because the event is now over a year old and the charge was a one-time item in the prior period. The financial consequence persists in comparative results (gross margin improved YoY because the charge did not recur), but the narrative announcement itself is lifecycle-complete.

NVIDIA now discloses a new USG licensing program (started February 2026) that permits limited H200 shipments to specific China customers, subject to U.S. inspection and a 25% tariff. No revenue has been generated yet, and import approval into China remains uncertain. This is a new development not present in the baseline.

The baseline described the AI Diffusion IFR (published January 2025, rescinded May 2025) and noted a replacement rule was forthcoming with uncertain scope. The current filing does not mention this at all, suggesting either the replacement rule was finalized without material impact or the issue is no longer current. The removal of forward-looking uncertainty language is notable.

NVIDIA now discloses a new product platform, Rubin, expected to ship in H2 FY2027. This is a new forward-looking product roadmap disclosure not present in the baseline, which only mentioned Blackwell Ultra samples in Q2 FY2026.

NVIDIA discloses $18.6 billion in new investments in private companies and infrastructure funds during Q1 FY2027, including AI model makers that may indirectly purchase NVIDIA products. The company also notes investments in publicly-held equity securities with significant fair-value volatility risk. This is a substantial capital deployment not disclosed in the baseline.

NVIDIA introduces a new revenue reporting framework splitting Data Center into Hyperscale (public clouds, large consumer internet) and ACIE (AI Clouds, Industrial, Enterprise), plus Edge Computing. This replaces the prior segment-only view and provides more granular visibility into customer mix and growth drivers.

Data Center revenue nearly doubled YoY ($75.2B vs $39.1B). Hyperscale customers remain ~50% of Data Center revenue, but the current filing explicitly notes zero China Hopper shipments in Q1 FY2027 vs $4.6B in Q1 FY2026, highlighting the ongoing China export impact. The revenue growth is driven by Blackwell 300 and networking products.

Gross margin improved to 74.9% in Q1 FY2027 from 60.5% in Q1 FY2026, primarily because the prior-year $4.5B H20 charge did not recur. Inventory provisions dropped from $5.3B to $1.1B YoY. The current filing notes gross margin was flat sequentially as Blackwell remains the majority of revenue, indicating stable product mix.

R&D expenses accelerated: compute and infrastructure costs up 112% YoY (vs 70% in prior year), engineering development materials up 204% (vs 182% prior year), and compensation up 31% (vs 34% prior year). The sharp increase in compute/infrastructure spending reflects the scale of AI model training and internal infrastructure buildout.

Other income swung from a $180M loss in Q1 FY2026 (driven by unrealized losses in CoreWeave and Arm) to a $15.9B gain in Q1 FY2027 (driven by $13.4B unrealized gains in publicly-held equity and $2.6B in non-marketable equity). This reflects significant fair-value appreciation in NVIDIA's equity investment portfolio.

Share repurchases increased to $20.2B in Q1 FY2027 from $14.5B in Q1 FY2026. The Board approved an additional $80B repurchase authorization in May 2026, bringing total remaining authorization to $38.5B. The quarterly dividend was raised 25x from $0.01 to $0.25 per share, a significant capital return increase.

The baseline disclosed plans to increase U.S.-based manufacturing to strengthen supply chain resiliency. The current filing does not mention this initiative, suggesting either the plan was completed, deprioritized, or is no longer considered forward-looking news. The removal of this strategic disclosure is notable given ongoing tariff and trade policy discussions.

The baseline included a standalone climate change disclosure stating no material impact to date. The current filing omits this section entirely, which is a structural change in disclosure rather than a substantive business development.

NVIDIA increased its commercial paper program limit from $575 million to $25.0 billion, a 43x increase. This provides significantly greater short-term liquidity capacity, though no commercial paper is currently outstanding in either period.

Unrecognized tax benefits increased from $2.5B to $4.5B, an 80% increase. This reflects growing tax uncertainty as NVIDIA's income scales, though the company notes it cannot estimate timing of potential tax liability or settlement.

The baseline disclosed an expected near-term release of valuation allowance on state deferred tax assets, which would reduce income tax expense. The current filing does not mention this, suggesting either the release occurred (and is now reflected in results) or the expectation changed. The removal of this forward-looking tax benefit is notable.

NVDA added a new standalone competition risk factor describing intensifying competition from multiple angles: competitors with fabrication facilities and stronger resources, customers developing their own ASICs optimized for specific workloads that bypass NVDA's full feature set, and cloud providers offering competing AI services. The disclosure highlights that NVDA may struggle to achieve necessary scale in AI cloud services and that customer-developed alternatives could reduce demand for NVDA products.

NVDA added new disclosure that customer ability to deploy NVDA infrastructure depends on availability of data centers, energy, and capital — all of which face multi-year expansion timelines and regulatory/financing hurdles. Less-capitalized customers may be unable to secure financing for large-scale projects. These constraints could delay deployments or reduce AI adoption scale, directly impacting NVDA's revenue even if product demand exists.

Prior filing described an ongoing investigation into whether NVDA's export-control compliance discriminates against China customers. Current filing reports that on September 15, 2025, China's antitrust regulators published a preliminary finding that NVDA's compliance with U.S. export controls (offering degraded products to China) violated the terms of China's Mellanox acquisition approval. This escalates from investigation to preliminary adverse finding, increasing risk of penalties or operational restrictions in China.

NVDA added disclosure that starting February 2026, the U.S. government granted licenses for small H200 shipments to specific China customers, but NVDA has generated zero revenue under this program and does not know if China will allow imports. The license requires U.S. inspection before shipment, triggering a 25% tariff on re-importation to the U.S. NVDA may be unable to pass the tariff to customers, faces potential litigation, and competitive harm. This is a new revenue obstacle with direct cost and margin impact.

Prior filing described potential foreclosure from China data center market as a conditional risk ("may be unable"). Current filing states NVDA is already effectively foreclosed as of Q1 FY2027, unable to deliver a product approved by both U.S. and Chinese governments. NVDA can ship gaming/workstation GPUs but not data center products. The disclosure adds that this foreclosure is helping competitors build larger ecosystems to challenge NVDA worldwide, escalating the competitive harm beyond China alone.

NVDA added disclosure that starting August 2025, the U.S. government granted licenses for certain H20 shipments to China customers, but U.S. officials expect the government to receive 15% or more of revenue from licensed sales. This expectation is not codified in regulation, creating uncertainty. If enforced, this would directly reduce NVDA's net revenue from any licensed H20 sales to China by at least 15%, a material margin impact on an already-constrained revenue stream.

NVDA added disclosure of two new legislative risks: (1) the GAIN AI Act (passed Senate October 2025) would restrict the Trump Administration's ability to modify Biden-era export controls and allow private parties to review/overturn licensing decisions, reducing regulatory flexibility; (2) the Remote Access Security Act (RASA) under consideration would prohibit cloud services to companies with China-based ultimate parents, which could restrict NVDA's cloud service providers and OEMs from serving a broad customer base. Both could materially impact NVDA's business and revenue.

NVDA expanded disclosure on chip tracking/throttling mandates to add that even draft (non-binding) legislation has harmed business: following U.S. proposals for mandatory chip features, China's government publicly questioned whether NVDA's H20 products have built-in vulnerabilities, discouraging customer purchases. NVDA issued a public response denying such vulnerabilities. This shows that legislative proposals alone — before enactment — are creating reputational and demand harm in key markets.

NVDA expanded the list of China-origin open-source foundation models from "DeepSeek or Qwen" to "DeepSeek, Qwen, or KIMMI" and added context that demand for open-source models promotes worldwide use of NVDA products. The change highlights that restrictions on supporting these models would harm NVDA's global business, not just China sales, because open-source models drive broader GPU adoption. The addition of KIMMI reflects the evolving landscape of China-origin models that NVDA's products support.

NVDA updated product architecture language from "Product transitions" to "Introducing or offering multiple architectures concurrently" and narrowed the scope from "Data Center, Gaming, Professional Visualization, and Automotive" to "Data Center and Edge Computing." The current filing adds specific forward guidance: the Rubin platform is expected to start shipping in the second half of fiscal year 2027. This provides concrete timing for the next major architecture launch and signals continued annual cadence, which magnifies supply/demand forecasting complexity.

Prior filing listed customer ability to invest in AI infrastructure as a demand factor. Current filing expands this to "customers' and partners' ability to secure capital and energy and to build complex data center infrastructure timely." The change adds energy availability and infrastructure build timelines as explicit demand constraints, reflecting the new risk factor on data center/energy capacity. This broadens the set of external bottlenecks that could limit NVDA's revenue even if product demand exists.

NVDA updated state AI regulation disclosure to note that new state laws took effect on January 1, 2026. Prior filing described state regulations as under consideration or already enacted; current filing confirms that some are now in force. This moves the risk from prospective to current, with immediate compliance cost and operational impact on NVDA's ability to train, deploy, or release AI models in those states.

NVDA added disclosure that the French Competition Authority is questioning whether gaming GPUs and data center GPUs are separate product categories, an inquiry that may impact export controls on gaming products in France and Europe. If regulators conclude they are not separate categories, gaming GPUs could face the same export restrictions as data center GPUs, disrupting NVDA's supply chain and sales in Europe. This is a new regulatory risk that could expand the scope of existing export controls.

Prior filing described U.S. manufacturing expansion as a plan ("We plan to increase"). Current filing states NVDA is already executing ("We are increasing") and adds that NVDA "may experience delays or difficulties in scaling production as planned." This acknowledges that the U.S. manufacturing ramp is underway but facing execution risk, with potential delays or difficulties that could impact NVDA's ability to meet demand. The shift from plan to execution with explicit risk language signals near-term operational challenges.

NVDA expanded compliance disclosure to emphasize that it provides assistance to authorities on diversion attempts but lacks physical control post-sale and must rely on customer/partner compliance programs. The current filing adds that even "unsubstantiated and untrue" diversion reports can harm NVDA, and adds a new category of reputational risk: "incorrect allegations that legitimate and appropriate business is using supposed 'loopholes' in the export controls." The tone is more defensive, pre-emptively addressing potential criticism of NVDA's compliance approach and framing certain allegations as incorrect.