MY THOUGHTS AHEAD OF NVIDIA EARNINGS
$NVIDIA(NVDA)$ has become the gravity point of this entire market. That’s why next week’s earnings aren’t just about one company -- they’re about whether the AI cycle still has the strength to carry $Invesco QQQ(QQQ)$ & $SPDR S&P 500 ETF Trust(SPY)$ higher.
The stock has already returned 15x since the 2022 bottom & that kind of run naturally invites the “bubble” label. The real question now is whether the fundamentals can keep matching the price.
Here’s what I’ll be watching 👇🧵
1. What the Bears Are Getting Wrong
The way I’m looking at $NVIDIA(NVDA)$ heading into earnings is less about whether they “beat” and more about whether they continue to prove that the demand side of this AI cycle is as durable as bulls believe -- and that the supply bottleneck remains on their side, not the customer’s.
Bears will tell you that Nvidia has become the railroad of the 1880s: overbuilt, over-owned, and destined to underperform once the tracks are laid. They’ll argue that hyperscaler capex can’t keep compounding at this pace forever, that margins at 75% are unsustainable once $Amazon.com(AMZN)$ , $Microsoft(MSFT)$ , $Alphabet(GOOGL)$ & $Meta Platforms, Inc.(META)$ accelerate their own silicon, and that China’s reluctance to commit to H20 orders underscores just how dangerous Nvidia’s geopolitical positioning is.
They’ll even go further and point to the law of large numbers: $4T in market cap on a business that may generate just over $200B in revenue next year is stretched, and no chip company has stayed on top forever -- $Intel(INTC)$ and $Cisco(CSCO)$ are both cautionary tales of leadership that looked unshakable until it wasn’t.
Image
2. Why Nvidia Will Remain the AI Engine
But this is where the business specifics matter. $NVIDIA(NVDA)$ isn’t just selling chips. They’ve embedded themselves into the hyperscaler operating model with CUDA as the de facto standard, with InfiniBand and NVLink holding together the networking layer of the biggest AI factories ever built, and with software subscriptions layered across their stack.
When $Microsoft(MSFT)$ or $Meta Platforms, Inc.(META)$ decides to build a $150B data center that consumes more power than NYC, they aren’t shopping for GPUs à la carte -- they’re buying a vertically integrated system that only Nvidia is currently capable of delivering at scale. Even when in-house chips are announced, they’re not displacing Nvidia but supplementing it, because the reality is there simply isn’t enough supply. That’s why the company continues to guide for supply constraints well into 2026.
The other bear pushback is around the idea that this is a bubble -- that AI spending has already overshot returns, that most enterprises are not seeing profitability from AI deployments, and that a political or social backlash could force governments to throttle expansion. And yes, those risks are real. The MIT report showing 95% of corporate AI pilots failing to produce measurable returns has been circulating widely, and Sam Altman himself has admitted we’re in a bubble-like environment. But the nuance is critical: failures at the corporate edge don’t change the capex math at the top.
The Mag Seven are not spending hundreds of billions to run AI pilots -- they are scaling foundational infra that will underpin robotics, enterprise applications, healthcare, and next-generation automation for decades. For them, Nvidia’s efficiency per watt is the single most important factor, because power, not chips, is the limiting resource. That’s why even at $35K per GPU, Nvidia remains the lowest cost option in terms of revenue per watt, and why hyperscalers have signed 20-year nuclear contracts just to guarantee power to run Nvidia hardware.
Image
3. What to Watch in Nvidia Earnings
The numbers themselves won’t be the story. Markets are already braced for ~$53B in revenue and ~$1.19 EPS, but the real signal comes from guidance. If Jensen frames demand as constrained by power, not price, then the bear case around valuation and margins loses credibility. Hyperscalers are doubling down on AI spend into 2026 -- if NVDA commentary confirms those orders are firm, the $4T market cap looks less like a ceiling and more like a waypoint.
The other tell is Blackwell Ultra. This rollout is Nvidia’s litmus test for whether it maintains its 90%+ data center share or leaves an opening for in-house silicon. If the ramp is on schedule and customers are already pulling forward orders, then Nvidia’s cadence is still running ahead of everyone else’s roadmap -- that’s the difference between a company priced for perfection and one that consistently delivers it.
China sits as the swing factor. The 15% export tax has been well-telegraphed, but investors will want clarity on whether Nvidia sees sustainable demand despite political crosscurrents. Any traction in H20 adoption would flip an overhang into upside optionality. And if not, the market will look for assurance that this demand can be absorbed elsewhere -- which so far has never been an issue.
Underneath all of it is the unit economics. Bears argue 75% margins can’t last, but hyperscalers don’t buy on sticker price, they buy on efficiency per watt. If Nvidia can show that Blackwell Ultra meaningfully outpaces both Hopper and the in-house alternatives, then not only do margins hold, they expand. That is the essence of the chip utility thesis.
For me, the report isn’t about a beat or miss by a billion -- it’s about confirming that Nvidia is still selling every GPU it can make at full price, that the Blackwell ramp is on track, and that hyperscaler capex points to a 2025% CAGR. If those hold, then the drawdowns will just be noise. Volatility is inevitable, but the structural story doesn’t change.
Image
For whom haven't open CBA can know more from below:
🏦 Open a CBA today and enjoy privileges of up to SGD 20,000 in trading limit with 0 commission. Trade SG, HK, US stocks as well as ETFs unlimitedly!
Find out more here：
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
- AuntieAaA·08-25GoodLikeReport