Key Highlights

  • ASML guidance upgrade signals persistent semiconductor supply constraints supporting Nvidia’s pricing power.
  • Nvidia’s data-center earnings dominance and Blackwell ramp reinforce AI-driven revenue visibility.
  • Hyperscaler capex nearing $700 billion anchors long-term demand outlook for AI infrastructure and GPUs.

Why Nvidia Is In Focus After ASML's Q1 Print

ASML Holding, (NASDAQ:ASML) the sole maker of extreme ultraviolet (EUV) lithography machines essential for fabricating advanced chips, reported first-quarter 2026 results that beat expectations and prompted the Dutch equipment manufacturer to raise full-year net sales guidance to between 36 billion and 40 billion euros, from its prior range of 34 to 39 billion euros. This guidance raise is significant because it reflects sustained, robust demand from foundries and integrated device manufacturers ramping advanced chip production. For semiconductor investors and Nvidia shareholders, the headline matters because ASML's health and capacity expansion are leading indicators of downstream demand from chip manufacturers—chief among them Nvidia.

ASML CEO Christophe Fouquet delivered the critical statement: 'We expect supply will not meet the demand for the foreseeable future.' That message has direct relevance to Nvidia. ASML is the only supplier of the EUV lithography tools needed to produce the advanced logic chips that power Nvidia's data-center accelerators. As ASML ramps production to satisfy backlog orders from Samsung, TSMC, and Intel, Nvidia and its customers benefit. Conversely, any slowdown in ASML's machine shipments ripples downstream to Nvidia's supply chain and product availability.

The immediate context: Nvidia reported fiscal Q4 2026 results in late February with blowout data-center revenue of 62.3 billion dollars, up 75 percent year-over-year, and provided fiscal Q1 2027 guidance of 78 billion dollars in total revenue (plus or minus 2 percent), well above consensus expectations of 72.6 billion dollars. The semiconductor industry remains ensnared in a supply pinch: wafer capacity from foundries, high-bandwidth memory (HBM) from Samsung and SK Hynix, and now—per ASML—the EUV lithography tools themselves are all constrained. ASML's raised guidance signals that this supply crunch will persist through 2026 and beyond, validating the demand thesis and offering comfort to investors concerned about potential oversupply.

For Nvidia specifically, ASML's statement is extraordinarily important. When ASML signals that it cannot build machines fast enough to satisfy demand, that is a sign that the semiconductor industry remains structurally supply-constrained. For Nvidia, it means that TSMC (the exclusive fab partner for Nvidia's advanced chips) will not suddenly find itself with excess capacity that could drive down wafer pricing or lead to aggressive competition from rival chipmakers. This supply-demand imbalance is the foundation of Nvidia's pricing power and the durability of its gross margins.

Most Recent Earnings: Data Center Still the Engine

Nvidia's fiscal year 2026 (ended January 26, 2026) closed with total revenue of 215.9 billion dollars, a 65 percent year-over-year increase. Data-center revenue hit 211.2 billion dollars for the full year, representing a 73 percent jump from fiscal 2025, and now accounts for over 91 percent of total revenue. This concentration reflects the dominance of GPU demand from cloud providers and AI infrastructure builders globally, with limited exposure to legacy markets like gaming and professional visualization.

In Q4 FY2026, Nvidia posted revenue of 68.1 billion dollars (up 73 percent year-over-year) and gross margin of 75.2 percent (non-GAAP), the highest of the fiscal year and a reflection of favorable product mix (Blackwell ramp) and operational leverage from scale. Earnings per share came in at 1.62 dollars (adjusted), beating the estimate of 1.53 dollars. The Q4 beat was driven by strength in Blackwell shipments, which ramped faster than expected, demonstrating flawless execution against manufacturing and logistics challenges.

Throughout fiscal 2026, data-center gross margin remained robust, averaging approximately 73 percent. This is remarkable given the volume of shipments and supply-chain complexity. Typically, when a company scales production as rapidly as Nvidia has (60 to 75 percent year-over-year growth), margin dilution follows as per-unit manufacturing costs increase and average selling prices face pressure. Nvidia has managed this inflection gracefully, a testament to strong underlying demand, limited competitive pressure, and pricing discipline maintained throughout the cycle.

Notably, Nvidia did not assume any China data-center revenue in its near-term outlook, an important caveat given volatile geopolitical dynamics and changing export controls. Even absent the China opportunity, the guidance jump speaks to robust US and international demand from hyperscalers, cloud providers, and enterprise customers. The company is shipping to over 15,000 customer accounts, with concentration among megacap cloud providers but diversification across emerging AI vendors and service providers.

Blackwell Ramp and the Path to Rubin

Nvidia's Blackwell architecture is ramping faster than originally planned. The B200 and GB200 products went into mass shipment in late 2025, and the B300 Blackwell Ultra variant began flowing to customers in January 2026—months ahead of expectations. The B300 features 288 gigabytes of HBM3e memory, up from 180 gigabytes on the B200, enabling larger model training and larger batch inference workloads that require increased memory capacity for context windows and model weights.

As of April 2026, B200 and GB200 hardware remains sold out through mid-2026, with an estimated backlog of approximately 3.6 million units. Cloud providers including Amazon, Google, Microsoft, and Oracle have listed B300 instances on their platforms, though dedicated on-premises lead times stretch 12 to 20 weeks. This supply crunch underscores the magnitude of demand, with customers accepting long wait times and premium pricing for cloud-based spot instances and reservation commitments.

Looking ahead, Nvidia is preparing the Vera Rubin architecture, the next major inflection point. Vera Rubin will debut in the second half of 2026, with production ramping underway. The company expects to commence production shipments in H2 2026, followed by Rubin Ultra in H2 2027. At GTC 2026, CEO Jensen Huang raised the company's long-term revenue opportunity to 1 trillion dollars through 2027, citing accelerating inference economics and larger models for reasoning tasks that require more compute and memory.

The Vera Rubin roadmap is aggressive but credible. Nvidia will have launched four major architectures (H100, H200, Blackwell, Vera Rubin) in three years—an unprecedented pace. This cadence locks in customers and defends market share against competitors. The company is also expanding into inference workloads and rack-scale systems, where gross margins may be lower but unit volume is substantially higher, driving incremental revenue growth.

Hyperscaler Capex: The Demand Side of the Equation

The Big Four cloud providers—Amazon, Google, Meta, and Microsoft—are committing enormous sums to AI infrastructure buildout. Combined guidance for 2026 capex totals close to 700 billion dollars: Microsoft over 100 billion (up from 63 billion in 2025), Google 175 to 185 billion (from 91 billion), Meta 115 to 135 billion (from 72 billion), and Amazon roughly 200 billion (from 131 billion). These are record capex levels and represent a historic shift in technology spending.

Approximately 75 percent of hyperscaler capex in 2026 will fund AI-related infrastructure, implying roughly 450 to 500 billion dollars in AI-specific spending. This capex boom is driven by long-term commitment to build large-scale foundation models, inference platforms, and agent-based systems. The shift toward inference—serving trillions of API calls—means GPUs will be deployed not just for training but for continuous inference, multiplying the installed base of accelerators needed globally.

Microsoft's capex guidance is particularly notable, as the company is ramping faster than peers and has committed to deploying Nvidia chips across Azure regions worldwide. Microsoft benefits from OpenAI's success and has an exclusive partnership to deploy those models in Azure, creating a powerful flywheel: Azure becomes the preferred platform for OpenAI workloads, driving incremental infrastructure demand and requiring more Nvidia chips.

For Nvidia, this capex wave is the linchpin of the bull case. Each exabyte of new AI inference capacity requires thousands of GPUs, thousands of HBM modules, and thousands of networking components. Supply is the constraint, not demand. ASML's raised guidance reinforces the narrative that hyperscalers will not lack for chips in 2026 and that supply-demand dynamics remain favorable.

Competitive Landscape: AMD, Custom Silicon, and the ASIC Question

Nvidia does face competition. AMD is ramping its MI355X Instinct GPU for release in Q2 2026 and has announced the MI400 Series for full-year 2026 shipment. AMD claims the MI355X delivers up to 4.2 times the performance of the prior-generation MI300X and projects 20 to 30 percent performance improvements over Nvidia's B200 in specific inference workloads.

The MI355X offers 288 gigabytes of HBM3e, exceeding Nvidia's B200 at 180 gigabytes and matching the B300. The MI400 Series, expected in H2 2026, promises up to 10 times the performance of the MI355X and will be the first product using AMD's unified UDNA architecture. With 432 gigabytes of HBM4 and 19.6 terabytes per second of memory bandwidth, it will be serious competition in specific use cases.

Despite AMD's technical improvements, several factors favor Nvidia. First, software ecosystem: the vast majority of generative AI models are optimized for Nvidia's CUDA platform. Porting to AMD's ROCm requires engineering effort and potential performance compromises. Second, installed base: Nvidia has sold hundreds of millions of GPUs; enterprises have deep expertise in Nvidia platforms. Third, hyperscaler relationships: the Big Four have committed billions to Nvidia infrastructure with minimal incentive to fragment portfolios with competing vendors.

Custom silicon from hyperscalers—such as Google's TPU, Amazon's Trainium, and Meta's custom ASICs—represents an alternative path optimized for proprietary workloads and internal use cases. However, they serve specific needs within each company and are not sold to third parties. While custom silicon will grow in absolute terms, it is unlikely to displace Nvidia's core business of selling general-purpose GPUs to the broader market.

China Export Controls and Geographic Mix

Export controls on advanced semiconductors remain a material wildcard. The Trump administration has oscillated on policy: halting AI chip sales to China in April 2025, reversing in July 2025, restricting again in April 2026, and permitting H200 sales in December 2025 subject to a 25 percent US government tax on revenues.

CEO Huang announced at GTC 2026 that Nvidia has received purchase orders for H200 processors from Chinese customers and is restarting manufacturing for that market. However, Nvidia's near-term guidance excludes China data-center revenue, suggesting conservative forecasting or operational delays in shipping. The China question is a tail risk to upside, not a core assumption in current guidance.

Historically, China represented roughly 20 to 25 percent of Nvidia's total revenue, though data-center revenue as a percentage of China sales was lower than the company's overall product mix. If export controls fully relaxed, China could represent multi-billion-dollar upside to annual revenue. If controls tighten further, Nvidia would absorb the loss, though US and international markets remain robust enough to hit growth targets.

For now, US and international demand is robust enough that Nvidia does not depend on China for growth targets. But over a multi-year horizon, if export controls relax, China could be significant incremental opportunity. A hardening of restrictions would pressure average selling price and could accelerate custom silicon development in China.

Supply Constraints: HBM, Packaging, Wafers, and Lithography

Nvidia's supply chain hinges on three critical inputs: wafer capacity from TSMC and Samsung, high-bandwidth memory (HBM3e and HBM4), and chiplet packaging capabilities. Each represents a potential bottleneck in the production pipeline. The company has worked extensively with suppliers to ensure no single point of failure, but constraints in any component would cascade through the entire system.

HBM3e remains constrained globally. Samsung and SK Hynix have raised HBM3e prices by nearly 20 percent for 2026 contracts, reflecting tight supply and surging demand. HBM3e is estimated at 66 percent of total HBM output in 2026, down from 87 percent in 2025 as HBM4 production ramps. However, the entire 2026 global HBM4 supply is already sold out and allocated to hyperscalers and cloud providers. Micron's high-bandwidth memory capacity is booked through year-end 2026.

The HBM market is experiencing a 'supercycle'—sustained demand growth exceeding industry historical norms. Each Nvidia GPU requires between 80 and 288 gigabytes of HBM depending on the model. A single data-center cluster with 10,000 GPUs requires upward of 1.8 petabytes of HBM. When hyperscalers deploy hundreds of thousands of GPUs per quarter, HBM vendors cannot keep pace despite aggressive production ramping and capital investment.

Looking forward, Nvidia is signaling interest in 16-layer HBM stacks for delivery as early as late 2026. Current interposers designed for HBM3e cannot physically route HBM4 signal density without crosstalk interference. Transitioning to 16-layer designs requires architectural shifts at memory vendors and Nvidia. This suggests HBM supply will remain a chokepoint through at least 2027. The company is working closely with SK Hynix, Samsung, and Micron on next-generation designs.

At the foundry level, TSMC is the exclusive producer of Nvidia's entire product lineup. TSMC has committed to major capacity additions and signaled accommodation for demand, but allocations remain tight. Wafer lead times for cutting-edge N4 and N3 process nodes extend many months out. ASML's raised guidance signals improving machine supply, but that will translate to higher wafer output with a lag of several quarters.

Risks

Several risks could derail the bull case. First, a deepening US-China trade war or expanded export controls could abruptly cut off material demand. The China opportunity represents billions in downside risk if lost to policy shifts.

Second, if hyperscaler capex disappoints relative to current guidance, or if generative AI proves to be transient rather than durable, GPU demand could contract sharply. While macro evidence supports continued spending, a severe recession or slowdown in enterprise AI adoption could reverse the narrative rapidly.

Third, competitive pressure from AMD and hyperscaler custom silicon could erode Nvidia's gross margins and market share over time. To date, Nvidia's margins remain above 70 percent, but if AMD gains meaningful traction, average selling prices could compress. Nvidia would need to maintain software and ecosystem advantages to defend pricing.

Fourth, supply-chain disruptions—geopolitical tensions, manufacturing accidents, logistics delays—could constrain wafer or HBM availability and upend shipment timelines. ASML's remarks about EUV supply being tight suggest any hiccup would ripple through the industry. A fire at a Samsung or SK Hynix fab, or delays at TSMC, could compress Nvidia's shipments and quarterly results.

Fifth, Nvidia's valuation has expanded significantly. As of mid-April 2026, the stock trades around 196 dollars, with a market cap of 4.77 trillion dollars—the world's most valuable company—up 76 percent year-over-year. Any guidance misstep or slowdown in data-center growth could trigger sharp correction. The stock is priced for perfection.

Near-Term Outlook

For the next 12 to 18 months, the base case remains that AI infrastructure capex will continue to accelerate, demand for Nvidia's accelerators will remain robust, and supply will be the primary limiting factor. Hyperscalers have committed massive budgets and are in early innings of large-scale AI deployment. Blackwell will sustain revenue and gross margin through 2026, with Vera Rubin introduction in H2 2026 setting the stage for the next adoption and refresh cycle.

ASML's raised guidance underscores that EUV lithography supply, the ultimate gating factor for advanced chip production, is improving but will remain constrained through 2026. This is favorable for Nvidia because customers will not face sudden gluts that could lead to massive inventory builds or price cuts. The supply-constrained environment benefits incumbent suppliers like Nvidia disproportionately.

Gross margins may face modest pressure as Blackwell production ramifies and HBM3e costs decline, but Nvidia's track record (above 70 percent in FY2026) suggests the company can defend high profitability even as ASPs normalize. Earnings-per-share growth will be sustained by unit growth and operational leverage. Management has guided for continued data-center growth in mid-to-high double-digit percentages, suggesting a credible path to 300 billion dollars in revenue by fiscal 2028.

The key wildcard is China export policy. If the US permits broader chip sales or if geopolitical tensions ease, upside to addressable market could be significant. If controls tighten, downside risk exists. Management is assuming a conservative posture and not factoring China into base-case guidance.

Summary

ASML's Q1 2026 beat and raised full-year guidance to 36 to 40 billion euros is a strong signal that the AI infrastructure capex cycle is in full swing and supply-constrained through at least 2026. For Nvidia, ASML's statement that 'supply will not meet demand for the foreseeable future' validates the thesis that demand for Nvidia's accelerators—Blackwell and upcoming Vera Rubin—will remain strong and pricing power and margins will be sustained throughout 2026.

Nvidia's fiscal Q4 2026 and full-year results demonstrated the power of the AI accelerator business: 215.9 billion dollars in annual revenue, 65 percent year-over-year growth, and 71 percent gross margin. The fiscal Q1 2027 guidance of 78 billion dollars far exceeds consensus, signaling management confidence in sustained demand. B300 Blackwell Ultra is shipping ahead of schedule, and the Vera Rubin roadmap is on track to debut in H2 2026.

Hyperscaler capex in 2026 will total 700 billion dollars, with roughly 450 to 500 billion allocated to AI infrastructure. This wave—driven by Microsoft, Google, Meta, and Amazon—will sustain demand for Nvidia's chips through Blackwell in 2026 and into the Vera Rubin cycle in H2 2026 and 2027. Each company has committed multi-year budgets with AI investment as a strategic priority.

AMD's MI355X and MI400 will compete, but Nvidia's installed base, ecosystem maturity, and first-mover advantage in inference economics will likely preserve dominant market share through 2026 and beyond. Custom silicon from hyperscalers will grow in absolute terms but will not displace general-purpose GPU demand in the broader market.

Supply-chain challenges—HBM constraints, wafer availability, packaging capacity, and EUV lithography—remain critical, but ASML's raised guidance and Nvidia's Blackwell ramp ahead of schedule suggest the industry is navigating bottlenecks effectively. Memory vendors are ramping production, TSMC is investing in capacity, and ASML is expanding EUV shipments. Supply constraints will persist through 2026 but are unlikely to limit Nvidia's growth trajectory.

China export controls remain a tail risk to upside (if relaxed) and downside (if tightened), but are not factored into current guidance. The core business—selling advanced accelerators to US and international hyperscalers—is strong enough to drive Nvidia toward 300 billion dollars in annual revenue by fiscal 2028, with sustainable margins above 70 percent.

For investors, ASML's Q1 print confirms the secular tailwind behind Nvidia's business. The near-term opportunity lies in Blackwell production ramping, margin persistence, and Vera Rubin rollout. Risks include competitive pressure, deepening trade tensions, slowdowns in AI adoption, and stretched valuation. But the fundamental case—that AI infrastructure capex will remain robust and supply-constrained through at least 2026—remains intact. Nvidia is the primary beneficiary of the AI infrastructure supercycle, and ASML's results validate that cycle's durability and scale.