Portfolio holding · Thesis memo
$206.64
-$1.19 (-0.6%) · ~90d
Delayed market data · chart by TradingView
NVDA: 5 drawdowns of 30%+ in 20 years (2006 2026) vs. +8.8% over the past year. NVDA is priced as the preferred AI accelerator platform: customers keep choosing CUDA/Hopper-Blackwell for training and inference.
Framing as of Jul 31, 2026
NVIDIA Corporation is a leading provider of AI infrastructure, specializing in data center accelerated computing and networking platforms. They offer GPUs for gaming, professional visualization, and AI workloads, alongside automotive solutions.
Their primary customers include original equipment manufacturers, cloud service providers, and system integrators, positioning them as a key player in the AI hardware market.
So what does the current situation look like for NVDA?
NVIDIA has recently demonstrated strong financial performance, with its latest annual revenue growing 65.5% year-over-year and a gross margin of 71.1%. The company has consistently beaten earnings and revenue estimates, most recently on May 20, 2026, with revenue delivered at $81.61B versus an estimate of $78.42B.
There are no recent news items indicating a shift in hyperscaler spending or customer platform changes.
To judge whether that matters, you need how the company got here.
Over the past year, NVIDIA's stock has seen a +8.8% change, reflecting its position as the preferred AI accelerator platform. The working thesis is that customers continue to choose NVIDIA's CUDA and Hopper-Blackwell platforms for training and inference, with Data Center growth tied to hyperscaler buildouts.
This narrative is supported by consistent revenue and EPS beats, though a significant portion of the current price embeds expectations of multi-year AI infrastructure continuation.
A large move should make you careful - the relevant history explains why people disagree.
NVIDIA's history is marked by significant volatility, with 5 drawdowns of 30%+ over the past 20 years (2006-2026). This base-rate history frames the current debate: whether the stock's valuation assumes this pattern of sharp drawdowns has fundamentally changed, or if it is merely a pause before another cycle.
The NVDA memo turns on what must stay true, and what would prove it wrong.
All checks for NVDA are intact.
NVDA is priced as the preferred AI accelerator platform: customers keep choosing CUDA/Hopper-Blackwell for training and inference, and Data Center growth stays tied to hyperscaler buildout rather than a one-quarter spike.
Below we track the key markers for this stock. If any of them break, that is a mechanical hit to the thesis. A stock drop alone does not count.
Hyperscaler AI infrastructure spend stays supportive of Data Center demand.
Intact·F1: event reviewed; insufficient mechanism detail.
Breaks if: Two major hyperscalers guide AI infrastructure / capex down in the same season.
Where we look: Alphabet · Meta · Microsoft · Amazon earnings windows · Hyperscaler earnings — capex / AI infrastructure commentary
Large customers keep choosing NVDA platforms over alternatives at scale.
Intact·F2: event reviewed; insufficient mechanism detail.
Breaks if: A named large customer discloses a material shift away from NVDA platforms.
Where we look: Customer earnings / 8-K disclosures · Customer earnings / design-win disclosures; NVDA Data Center commentary
Supply and packaging constraints ease without collapsing pricing power overnight.
Intact·F3: Company removes ‘sold out’ / constrained-supply language while competitors flood capacity. — The article states that memory is now AI's biggest bottleneck, implying continued demand and potential supply constraints. (fool.com)
Breaks if: Company removes ‘sold out’ / constrained-supply language while competitors flood capacity.
Where we look: NVDA quarterly call · NVDA earnings — supply, backlog, and pricing language