Nvidia

Nvidia Corporation (NASDAQ: NVDA) designs graphics processors. That line used to describe the whole company. It now describes a footnote: the same silicon that once rendered video games trains and runs the large models behind the current AI buildout, and Nvidia sells the chips, the networking, and a growing software layer that ties a data center together.

TEXXR’s archive holds 1,008 Nvidia-tagged articles running back to the final quarter of 2014 — one of the deepest single-entity records in the corpus. The volume itself is worth reading before the words are. Coverage ran from single digits to the high teens per quarter through most of the 2010s, broke 30 for the first time in 2022, broke 50 for the first time in the third quarter of 2023, broke 100 in the third quarter of 2025, and peaked at 193 articles in the first quarter of 2026 — 2.3 times the trailing eight-quarter average of 85 articles a quarter heading into that peak. Nvidia’s stock did not reach its own all-time high until May 2026, a month into the following quarter: the coverage record ran ahead of the price record, then eased off. TEXXR logged 162 Nvidia articles in the second quarter of 2026 and 143 so far in the third, which is still open.

This is DeadRisk’s mag7 tier: the deepest corpus we hold on any single name, and explicitly not a stock-picking exercise. Robert Shiller’s case, laid out in Irrational Exuberance, was that narratives move markets on their own terms, apart from fundamentals. Nvidia’s coverage record is a twelve-year test of that case. This page tracks the record, not the trade.

The earliest coverage in TEXXR’s record is a niche chip company defending its turf: an ITC patent complaint from Samsung over GPU technology in the last quarter of 2014, GeForce card launches, the Shield Android TV set-top box announced in 2015, a 2018 Spectre disclosure alongside the rest of the chip industry. Through 2019 the record shows a graphics-card and workstation vendor with an autonomous-driving side bet — the Drive PX self-driving platform, announced in 2015 — layered on top.

Two crypto cycles interrupt that story before the AI one starts. SoftBank took a $4 billion stake in Nvidia in 2017, on the strength of gaming and early data-center demand. By 2018 Nvidia was disclosing $289 million in chip sales specifically to cryptocurrency miners, and a mining-driven GPU shortage was national tech-news copy. The pattern repeated in 2021 — Nvidia shipped mining-limited GeForce cards to keep gamers supplied — and broke in 2022: crypto demand collapsed, Nvidia disclosed excess RTX 3000-series inventory, longtime board partner EVGA quit the graphics-card business entirely citing conflicts with Nvidia, and the SEC fined Nvidia $5.5 million for failing to adequately disclose how much of its earlier gaming revenue had actually come from miners.

The Arm acquisition brackets the same stretch and tells a similar story of ambition meeting friction: announced at more than $40 billion in September 2020, the same year Nvidia’s A100 chip opened its data-center era and the company passed Intel as the most valuable US chipmaker, then abandoned in February 2022 after regulators refused to clear it and SoftBank took Arm public instead.

What follows is the AI-datacenter turn most readers already know in outline, but the coverage record gives it a shape: a $1 trillion market cap in May 2023, the week Nvidia raised its revenue forecast on AI demand; $2 trillion nine months later; $3 trillion four months after that, passing Apple, then Microsoft two weeks after that to take the title of world’s most valuable public company. Nvidia reached $4 trillion in July 2025 and $5 trillion that October. An export-control subplot runs through those numbers and never fully resolves: the H800, a cut-down H100 built for China in 2023; the H20 licensing back-and-forth of 2025; and a China-only RTX 6000D that Reuters reported met “lukewarm demand” that September. By May 2026 the record had Nvidia positioned as an AI investor in its own right, with more than $40 billion in equity commitments across the industry — before a Bloomberg-reported $1 trillion pullback in July 2026 put the stock back at an 18-times forward-earnings multiple, its lowest since 2019.

TEXXR’s most recent batch of 500 tracked Nvidia edges runs launch (26.8%), financial (24.8%), and partnership (24.4%) — three-quarters of everything the graph records about the company, with regulatory (6.2%), controversy (4.6%), and competitor (4.2%) edges making up a thin remainder. That is the record of a company still mostly building rather than mostly defending.

But the mix is moving. Nvidia’s edge-arc between the first quarter of 2024 and the first quarter of 2026 shows launch-predicate edges falling from 49% to 29% of the total, while competitor-predicate edges rose from zero to 23% over the same stretch. AMD, Qualcomm, and MediaTek all now show up contesting Nvidia’s data-center CPU position, and Nvidia’s own edges name Intel’s foundry, not just its chips, as a competitive front: Nvidia tested Intel’s 18A process and quietly walked away from it in mid-2026, the same season it bought $5 billion of Intel stock — 214.7 million shares at $23.28 — as part of the two companies’ x86 partnership.

The clearest illustration of where Nvidia’s own balance sheet now runs is CoreWeave. Nvidia owns a 5.96% stake as of March 2025, anchored CoreWeave’s IPO that month with a roughly $250 million order, and added another $2 billion in January 2026 to help CoreWeave deploy Nvidia’s new Vera CPUs and add gigawatts of AI computing capacity. The same investor role shows up at industry scale: more than $40 billion in equity commitments by May 2026, with $30 billion of that in OpenAI alone.

Nvidia’s other recent edges read like a map of the AI-capex supply chain itself: Cadence named Nvidia, TSMC, and Schneider Electric as early users of its AuraStack design tool; Nokia built an AI radio-access-network platform with Nvidia; and Nvidia’s own regulatory edges show it cutting authorized AI-chip customers in Singapore, Malaysia, and Japan by more than half, to close off routes for chips to reach China.

One edge pair is worth reading side by side. In the same week of December 2025, Nvidia’s $20 billion Groq licensing deal appears in the record both as a confirmed acquisition and — after a leaked internal email from Jensen Huang — as a denied one. Nvidia ended up with Groq’s CEO, Jonathan Ross, and roughly 90% of Groq’s staff, without buying the company outright. The graph logged the contradiction honestly; the contradiction is the story.

Positioning: how much of NVDA is sold short, on the record brokers file twice a month. It lags by design — read it as where the crowd stood, not where it stands.

Order flow is signed before it becomes revenue — the earliest of the three views. From the capital-flow ledger.

Signals in the coverage record worth tracking, not trade ideas:

• Quarterly article volume. Does it re-accelerate past the 193-article 2026Q1 peak, or keep reverting toward the 2025 run rate of 85 to 128 a quarter? Two quarters of decline — 162, then 143 — are already logged.
• The competitor-edge share. It went from zero to 23% of Nvidia’s edges between 2024Q1 and 2026Q1. Does it keep climbing as AMD, Qualcomm, and MediaTek edges accumulate, or has the rivalry framing already peaked with the CPU story?
• The Singapore/Malaysia/Japan customer cuts. Still logged as rumored. A status flip to confirmed, or a new regulatory or legal edge attached to it, is the thing to watch for.
• The Groq relationship’s status. A confirmed partnership edge and a denied acquisition edge sit side by side. Whether that resolves cleanly, or produces a new controversy edge, says something about how much of Nvidia’s talent strategy the company will call by its name.
• Second-order financial edges of the CoreWeave kind — equity stakes in the companies buying Nvidia’s chips. More of them, or fewer, is the clearest register the graph offers of how far Nvidia’s balance sheet is stretching into the AI-capex chain it supplies.

Nvidia lives or dies by the same constraint as these names — small equal-weighted groups, one constraint each. Not investment products: see all categories.