Evid Invest
EvidInvest
← Back to Blog

NVIDIA's AI infrastructure story after a 63% run: what the filings confirm

·EvidInvest Team
NVDAAI infrastructureSEC filingsAetherData Centersemiconductors

NVIDIA's AI infrastructure story after a 63% run: what the filings confirm

Draft for EvidInvest review. This is financial research content, not investment advice. Built from authenticated Aether financial search over NVIDIA SEC filings, direct SEC filing checks, and Yahoo Finance one-year adjusted-close data. Aether transcript search failed in this research run, so this draft does not rely on earnings-call transcript evidence.

Thesis

NVIDIA is no longer simply a semiconductor company with AI exposure. Its own filings now frame the business around data-center-scale AI infrastructure.

The market has rewarded that story. Based on market data checked in the parent research pack, NVDA adjusted close moved from $137.35 on 2025-06-02 to $224.36 on 2026-06-01, a calculated +63.35% one-year return. The stock also reached a period high of $235.74 on 2026-05-14.

That price move changes the research question. The useful question is not whether NVIDIA has AI momentum. The filings already show it does. The better question is: what does NVIDIA's own SEC evidence confirm, what can only be inferred, and where are the evidence boundaries after the stock has already rerated?

Aether's SEC-first evidence stack points to three connected conclusions:

  1. AI infrastructure is now the dominant growth engine. FY2026 revenue was $215.9B, up 65%, and Data Center revenue was up 68%. Q1 FY2027 revenue was $81.6B, up 85%, while Data Center revenue was $75.2B, up 92%.
  2. The growth is tied to systems, networking, and interconnects, not only chips. NVIDIA cites Blackwell systems, InfiniBand, Spectrum-X Ethernet, and NVLink as Data Center growth drivers.
  3. The upside has real constraints. Customer concentration, data-center availability, energy, customer capital, export controls, and supply-chain dependencies define the risk side of the thesis.

That is where EvidInvest and Aether are useful: turning a broad AI narrative into labeled evidence — confirmed SEC facts, inferred relationships, and claims that still need sourcing.

What changed

NVIDIA's latest filing evidence shows a business increasingly shaped around AI infrastructure buildout.

The FY2026 10-K says revenue reached $215.9B, up 65% from a year ago, and Data Center revenue was up 68%. The Q1 FY2027 8-K earnings exhibit then reported record revenue of $81.6B, up 85% from a year ago, and record Data Center revenue of $75.2B, up 92%.

The Q1 FY2027 10-Q gives the operating explanation: Compute & Networking revenue growth was driven by Data Center products, including the ramp of Blackwell systems and demand for InfiniBand, Spectrum-X Ethernet, and NVLink solutions.

That matters because the NVIDIA story is no longer just "more GPUs." The filing-backed story is GPU compute plus networking, interconnect, systems, packaging, memory, and customer data-center deployment.

Evidence label: confirmed SEC fact for revenue, Data Center growth, Blackwell systems, InfiniBand, Spectrum-X Ethernet, and NVLink as stated drivers. The broader conclusion that NVIDIA is a full AI infrastructure platform is an inferred relationship supported by those disclosures.

Financial growth: enormous, but now widely visible

The numbers are unusually direct.

NVIDIA's FY2026 annual filing reported $215.9B of revenue, up 65%. The latest Q1 FY2027 earnings exhibit reported $81.6B of quarterly revenue, up 85%, with Data Center revenue of $75.2B, up 92%.

The Q3 FY2026 8-K evidence shows the ramp was not isolated to one quarter: revenue was $57.0B, up 62% year over year, and Data Center revenue was $51.2B, up 66% year over year.

The filing evidence therefore supports the core bull-case premise: AI infrastructure demand has converted into reported revenue at extraordinary scale.

But the stock-performance evidence matters too. NVDA was already up +63.35% over the checked one-year period. That does not prove the stock is overvalued. It does mean the article should avoid pretending the market has not noticed the AI infrastructure story.

Evidence label: confirmed SEC fact for reported revenue and Data Center growth. The +63.35% stock move is confirmed market data from Yahoo Finance adjusted-close calculations in the parent research pack. Any valuation judgment — undervalued, overvalued, upside, downside — is needs source/model unless a refreshed valuation model is added.

Customer and end-market signals: hyperscale plus ACIE, but unnamed direct customers

NVIDIA's SEC evidence confirms broad customer demand, but it does not support naming specific hyperscale customers in this draft.

The 2025 10-K says NVIDIA's customers include the world's leading public cloud and consumer internet companies, thousands of enterprises and startups, and public-sector entities. The Q1 FY2027 CFO commentary adds a stronger current mix signal: hyperscale revenue increased sequentially and remained approximately 50% of Data Center revenue, while the remaining 50% came from a diversified set of customers including AI Clouds, industrial, enterprise, and sovereign customers.

NVIDIA also says it is transitioning to a new reporting framework with two market platforms — Data Center and Edge Computing. Within Data Center, it expects to report Hyperscale and ACIE, where ACIE incorporates AI Clouds, Industrial, and Enterprise.

That is a useful evidence boundary. The filings support a broad end-market demand story. They do not, in the evidence retrieved here, identify Microsoft, Amazon, Google, Meta, OpenAI, Oracle, CoreWeave, xAI, or other named customers as specific contributors to NVIDIA revenue.

There is also a concentration signal. In Q3 FY2026, NVIDIA disclosed four direct customers each above 10% of total revenue: 22%, 15%, 13%, and 11%, all attributable to Compute & Networking. In Q2 FY2026, two direct customers represented 23% and 16% of total revenue.

Evidence label: confirmed SEC fact for customer categories, hyperscale/ACIE mix, and unnamed direct-customer concentration. Mapping those unnamed direct customers to named hyperscalers or end users is an inferred relationship and should not be published without additional sources.

Supplier-chain and dependency signals: demand has to pass through a concentrated hardware stack

NVIDIA's filings also show why AI infrastructure upside is not only a demand question.

The FY2026 10-K names key supply-chain dependencies:

  • Wafer foundries: TSMC and Samsung.
  • Memory suppliers: SK Hynix, Micron, and Samsung.
  • Packaging technology: CoWoS.
  • Assembly, testing, and packaging partners: Hon Hai Precision Industry, Wistron, and Fabrinet.

The same annual filing says NVIDIA's supply chain is mainly concentrated in Asia, while the company is expanding into the U.S. and Latin America.

Older filing evidence also shows how demand can convert into operational commitments. NVIDIA has disclosed that in periods of growth it may place non-cancellable inventory orders, pay premiums, or provide deposits to secure future supply and capacity. During the Blackwell transition, the company also disclosed a mask change to improve production yield and inventory provisions for low-yielding Blackwell material.

The operational implication is clear: Blackwell/GB-system upside depends on wafer supply, HBM/memory availability, advanced packaging, assembly/test capacity, networking components, energy, and customer deployment schedules.

Evidence label: confirmed SEC fact for named suppliers, CoWoS packaging language, Asia concentration, and disclosed supply/capacity commitment language. Exact bottleneck severity, supplier-by-supplier revenue exposure, and capacity allocation are needs source.

Risks and caveats

The filing-grounded bull case is strong, but the risk stack is concrete:

  • Valuation/performance risk: NVDA's adjusted close rose +63.35% over the checked one-year period. The AI infrastructure story is already well known.
  • Customer concentration risk: Q3 FY2026 disclosed four direct customers above 10% of total revenue, all in Compute & Networking.
  • Infrastructure availability risk: NVIDIA's own 10-Q says the availability of data centers, energy, and customer capital is crucial, and that energy expansion is a complex, multi-year process.
  • Supply-chain risk: NVIDIA depends on named foundries, memory suppliers, CoWoS packaging, and assembly/test/packaging partners.
  • Product-transition risk: Blackwell ramp complexity has previously shown up in yield/material and inventory-provision language.
  • Export-control risk: NVIDIA recorded H20-related charges and lost shipment opportunity after new export licensing requirements. The Q1 FY2027 outlook explicitly assumes no Data Center compute revenue from China.
  • Evidence limitation: Aether transcript search failed with HTTP 500 / Vespa 400 in the parent research run, so this draft does not use management transcript commentary.

This is why the right framing is not "NVDA is a buy" or "NVDA is expensive." The better framing is: NVIDIA has filing-confirmed AI infrastructure momentum at enormous scale, but future upside still depends on capacity, energy, customer capital, supply-chain execution, export rules, and valuation discipline.

Evidence boundaries for upside and downside

What the SEC evidence confirms:

  • Revenue and Data Center growth are large, recent, and explicit.
  • Blackwell systems, InfiniBand, Spectrum-X Ethernet, and NVLink are named as Data Center growth drivers.
  • Customer demand includes hyperscale plus AI Clouds, industrial, enterprise, and sovereign customers.
  • Customer concentration is material through unnamed direct customers in Compute & Networking.
  • Supply chain depends on named foundries, memory suppliers, CoWoS packaging, and assembly/test/packaging partners.
  • Export controls and product-transition complexity can affect revenue, inventory, and margins.

What is inferred but not directly proven by the retrieved evidence:

  • Specific named hyperscale customer exposure.
  • Exact training-versus-inference revenue allocation.
  • Supplier-by-supplier bottleneck probability or revenue dependency.
  • Whether the current stock price is justified.

Needs additional source before publication:

  • Named customers such as Microsoft, Amazon, Google, Meta, OpenAI, xAI, Oracle, CoreWeave, etc.
  • Supplier capacity facts for TSMC CoWoS, HBM vendors, Hon Hai/Foxconn, Wistron, or Fabrinet beyond NVIDIA's own supplier list.
  • Analyst targets, consensus estimates, or market multiples.
  • Real-time price data beyond the 2026-06-01 endpoint used in the research pack.

How Aether changes the workflow

A normal market search can summarize the NVIDIA AI narrative. Aether's SEC-first workflow produces a cleaner evidence map:

  • 10-K: annual revenue, Data Center growth, supplier dependencies, geographic supply-chain concentration, and risk language.
  • 10-Q: current quarterly drivers, customer concentration, infrastructure constraints, H20/export-control impact, and reporting-framework changes.
  • 8-K: earnings-release exhibits and CFO commentary with Data Center revenue, hyperscale/ACIE mix, and outlook assumptions.
  • DEF 14A: proxy-supported summary language around Compute & Networking, Data Center compute, Data Center networking, and Blackwell systems.
  • Market data: one-year stock performance, which changes the valuation question.

The best AI-stock research is not the loudest story. It is the cleanest evidence boundary:

  • Confirmed SEC fact: Data Center scale, Blackwell/networking drivers, hyperscale/ACIE mix, customer concentration, supplier dependencies, export-control impact.
  • Inferred relationship: named customer economics, bottleneck severity, and exact downstream AI capex conversion.
  • Needs source: named hyperscaler revenue exposure, supplier capacity details, transcript commentary, and valuation conclusions.

Bottom line

NVIDIA's filings confirm a powerful AI infrastructure thesis. The company has extraordinary Data Center growth, explicit Blackwell and networking/interconnect drivers, broad hyperscale and ACIE demand, and a supply-chain map that shows how complex the AI infrastructure stack has become.

But after a +63.35% checked one-year stock move, the investment question is no longer whether NVIDIA is exposed to AI. It is whether the next layer of growth can keep outpacing expectations while navigating customer concentration, data-center and energy availability, export controls, product-transition complexity, and supplier-chain dependencies.

My current evidence-grounded classification:

NVIDIA is a filing-confirmed AI infrastructure leader, not just an AI narrative stock. The bull case is real, but the evidence says it should be researched as an execution, capacity, customer-capex, and valuation question — not as a simple "AI demand is high" story.

That is the gap EvidInvest and Aether are built to close: not headlines, but evidence boundaries.

Source notes

Fair Value Weekly

Get DCF breakdowns, fair value updates, and portfolio ideas for serious investors. No spam, no paywalled teasers.