Nvidia Stock Is A Buy On Dips Before Blackwell Arrives In 2025

Nvidia’s stock broke to all-time highs recently, trading at $148 in early November and $147 yesterday. The stock has left many investors wondering “what comes next” after the unrelenting, historic surge that began seven quarters ago.

To help my readers determine where Nvidia’s stock will go next, I’ve been fastidious in my analysis about the company’s outsized AI potential since 2018, tracking Big Tech capex as a proxy for AI demand since 2022, discussing the anomalous earnings and revenue revisions throughout 2023 and 2024, and reporting on never-before published data on supply chain checks as recent as two months ago.

CEO of Nvidia, Jensen Huang, waves during the launch of the supercomputer Gefion at Vilhelm ... [+] Lauritzen Terminal in Kastrup, Denmark, on October 23, 2024. The new AI supercomputer has been established in collaboration with EIFO and NVIDIA and is operated by the Danish Center for AI Innovation (DCAI). The computer is aimed 'at breakthroughs in quantum computing, clean energy, biotechnology and other areas serving Danish society and the world', according to Nvidia. (Photo by Mads Claus Rasmussen / Ritzau Scanpix / AFP) / Denmark OUT (Photo by MADS CLAUS RASMUSSEN/Ritzau Scanpix/AFP via Getty Images)

Ritzau Scanpix/AFP via Getty Images

The thoroughness is needed, however, as rumors from the media and short sellers alike run amuck. Rest assured, as 2025 approaches, supply chain data is giving bullish signals that the new generation of GPUs shipping in full volume by mid-2025 (and beginning to ship in the January quarter) will far exceed the GPU sales we saw in 2023 and 2024 combined.

Regarding my firm’s confidence in tracking supply chain data, when The Information stated Nvidia was experiencing a material delay on the next generation of GPUs, going so far as to state that Taiwan Semiconductor had machines sitting idle, I quickly refuted the report based on supply chain data my firm had been tracking. Those data points continue to indicate Blackwell is ramping. Here is what I stated:

“As of now, there’s a disconnect between next fiscal year’s revenue estimates of $167 billion and the $210 billion in GB200s alone expected to ship next year. Perhaps analysts are waiting for signals the supply chain can produce these outsized orders. So far, so good with the signals we see from TSMC and SMCI’s most recent earnings reports. Foxconn commentary helps, as well.”

Fast forward two months, and next year’s fiscal estimates stand at $185 billion up from $167 billion; showing no material impact from the delay (quite the opposite). Our firm was also able to use that same supply chain data to buy Nvidia in July/August, for an average cost basis of $109. The I/O Fund’s first trade was at $3.15, but we actively track the stock and publish our real-time trade alerts for anyone who feels they missed out on the AI juggernaut.

$5B+ in Blackwell Revenue for Q4

The first item that will determine the strength of the upcoming earnings report from Nvidia has nothing to do with the Q3 results. Rather, what the market will want to know is how much Blackwell revenue is expected in the January quarter. Morgan Stanley has estimates placed at $5 to $6 billion, with this number hitting a ceiling due to supply constraints; however, Piper Sandler sees Blackwell revenue potentially higher, at $5 billion up to $8 billion.

The bigger picture is that Blackwell will ramp by an order of magnitude, eventually exceeding Hopper’s revenue. To quantify this, Hopper has delivered approximately $125 billion in data center revenue since Q1 2023, based on estimates from Trend Force placing Hopper at generating 90% of data center revenue in 2024. Blackwell, on the other hand, is expected to deliver up to $210 billion next year alone, based on estimates for up to 60,000 to 70,000 GB200 NVL72 servers priced up to $3 million each.

Given the company is lapping tough comparables, the growth rate will slow considerably even if Blackwell does ramp from $6 billion per quarter to $60 billion per quarter by late-2026 (Hopper is in its seventh quarter and Blackwell will be in its seventh quarter by late 2026). This is because excellence begets excellence, and thus, Nvidia is competing with itself with each new generation of GPUs. For example, with Hopper, the company reported peak quarterly growth of 262% and 265% earlier this year, yet is expected to slow to the mid-40% for growth as we close out 2025.

Nvidia has multiple levers it can pull and outside forces at play that will help it maintain this 40%+ growth rate. This includes a 1-year product road map, Big Tech’s large appetite for AI spending, and long-term AI GPU market growth from Enterprises and the Consumer, plus a commanding market share position.

By coming to market with upgraded, more powerful GPUs on a now-annual cadence, with Blackwell Ultra, Rubin and Rubin Ultra soon to come, Nvidia will continue to be the largest beneficiary of Big Tech’s AI capex to an unprecedented degree as the company continually raises the bar on performance and TCO upgrades with each new generation.

Additionally, Nvidia has a software moat with CUDA and the cash to reserve chip capacity in bulk at the fab level to maintain an 80% to 85% share of what executives foresee as a $500 billion AI accelerator market by 2028. I first covered these points in my free newsletter when I published: “Here’s Why Nvidia Stock Will Reach $10 Trillion Market Cap by 2030.”

Of these points, one of the most visible is that Nvidia continues to pry away tens of billions in cash – and now hundreds of billions —- from the world’s leading tech companies.

Big Tech Capex to Surpass a Quarter Trillion

All roads lead to Nvidia, and it’s no secret that Big Tech and others are competing to purchase Nvidia’s supply constrained GPUs. Our firm began tracking Big Tech capex as a proxy for Nvidia demand in 2022, and tracking it on a quarterly basis starting in early 2023 – to help gauge AI demand, I continue to track Big Tech capex quarterly closely for our readers.

Our recent checks published in the analysis “AI Spending to Exceed a Quarter Trillion Next Year” reveal that AI spending continues to accelerate, with Alphabet, Amazon, Microsoft, and Meta on track to increase their spend by ~$90 billion YoY in 2024. This does not include xAI, CoreWeave, Oracle and dozens of others who are also spending multiple billions on Nvidia’s GPUs, as well.

To better understand the trajectory of AI spending, let’s take a step back to 2023, where the rapid ascent of ChatGPT at the beginning of the year set the stage for AI to step into the spotlight.

In the first half of 2023, Big Tech spent ~$74 billion on capex. Through Q3, that sum had moved up to ~$109 billion. In the first half of 2024, Big Tech spent nearly $104 billion, a 47% YoY increase. Through Q3, that sum had surged to $170 billion, up 56% YoY.

Big Tech could spend another $70 billion in Q4, based on guidance and comments from executives, who overwhelmingly discussed the need for more AI infrastructure, putting full year capex at ~$240 billion, or nearly 15% higher than the level they were tracking at the start of the year.

Big Tech’s Q4 capex could hit $70B, driven by AI infrastructure demand—pushing 2024’s total to ... [+] ~$240B, up 15% from early-year estimates!

I/O Fund

For 2025, Big Tech has already signaled a willingness to spend substantially more on AI. There is clear ROI for Amazon, Google and Microsoft as they rush to meet the elevated demand that continues to outpace AI capacity in their cloud infrastructures. More broadly, Big Tech and large enterprises are racing to further develop and broaden AI services and models. UBS projects Big Tech will spend ~10% more YoY, placing AI-driven capex at $267 billion; however, if 2024 is any sign, this estimate is too low. This all fits in with longer-term projections from Bank of America that sees a cumulative $700 billion spent on AI through 2026.

Nvidia Has Over 2X Better Margins Compared to Most Mag 7 Stocks

As we go through a lull between the Hopper generation being in its seventh quarter, and Blackwell not yet shipping in volume, our firm will be buying the dips on Nvidia for many reasons – one of them being it’s the market leader on margins. By having a near monopoly on GPUs, Nvidia has incredibly strong pricing power.

The GPUs coming in 2025, called Blackwell, are set to intensify this pricing power with DGX B200 systems reportedly going for up to a 40%+ premium to the previous DGX H100 systems, at $500,000 per server versus the low $300,000s per server, respectively.

While GB200 prices are estimated at $60,000 to $70,000 for a single chip, the NVL36 and NVL72 configurations carry much higher price tags and thus, higher average prices per GB200. For example, the NVL36 is expected to cost ~$1.8 million, and for 18 GB200s (36 B200 GPUs), that comes out to $100,000 per GB200 and additional components. For the NVL72, it works out to ~$83,333 per GB200 and additional components.

While there were concerns about Nvidia’s margins given that management guided for a sequential contraction in gross margins in Q3, the sheer pricing power of Blackwell will ultimately be a non-issue next year.

Nvidia’s operating margin of 62% exceeds second place Microsoft by 17.5 points and third place Meta by 21.9 points; Nvidia is more than double the rest of the Mag 7 including Apple and Alphabet. This is because Hopper’s pricing power versus the Ampere generation: Nvidia’s Compute and Networking operating margin expanded from 28.5% in Q3 FY23 when Hopper reached full production to 71.3% in the most recent quarter even as revenue grew 7x during that seven-quarter period.

Nvidia leads the MAG 7 with a 62% operating margin, driven by Hopper’s pricing power—more than ... [+] double Apple, Alphabet, and others in the group!

YCharts

Nvidia is expected to report roughly 50 bps to 100 bps margin contraction this quarter compared to last quarter, and will see roughly 200 bps to 300 bps margin contraction from its peak growth quarters earlier this year. As stated, the pricing power I foresee from Blackwell will keep the margins strong well into 2025, therefore, any concerns over margins this quarter will be a moot point by next year.

The strong margins combined with the expected growth in AI accelerators has caused some analysts to increase earnings per share substantially as of late. Bank of America increased its EPS estimates for next calendar year from $3.90 to $4.47 and for calendar year 2026 from $4.72 to $5.67.

In February, I wrote an analysis describing how Nvidia’s valuation was “eerily low despite 420% rally since 2023” to help our readers prepare for a higher return in the coming months, which detailed the importance of these revisions.

Ultimately, these revisions make the stock cheaper as it leads to more room in the bottom-line valuation. Despite being fairly straight forward, the velocity of the revisions is the single most important point that short sellers and Nvidia critics cannot seem to understand.

Q3 Earnings Details:

Of all the quarters since Nvidia’s Hopper release, this is the quarter most likely to be lackluster. This is because the impact of Hopper and the H200s are well-known and the Blackwell generation won’t be shipping in volume until Q1 and ramping further into Q2.

I am looking forward to the fiscal year guide in the February call, and am even more excited about the May earnings call when Blackwell’s impact will be better understood.

Nvidia’s Q3 FY2025 Revenue:

Nvidia is expected to report revenue of $32.9 billion for growth of 81.8% at the midpoint. Analyst expectations are higher than management guidance of $32.5 billion at the midpoint, for growth of 79.4%. This is a deceleration from last quarter’s 122.4% growth, and peak growth of 262% and 265% in the April and January 2024 quarters.

As pointed out on EPS, another area where Nvidia is unique is the sheer amount of analyst revisions on the stock. It not only speaks to Nvidia’s dominance in the AI data center to continually surprise the Street, but also to the challenge that analysts face in terms of predicting Nvidia’s persistent revenue surge.

For example, this year alone, analysts originally expected Nvidia to report 33.4% revenue growth and this quarter is now expected to be 81.8% growth, for revisions that total 48.4 points in about six months’ time (more than double the original growth expectations).

This quarter, there is a wide range of expectations with UBS believing Nvidia will beat by as much as $2 billion, for revenue of $34.5 billion to $35 billion for Q3. Piper Sandler foresees a beat of $1.3 billion for Q3, and a beat of $1.5 billion for Q4.

It’s been quite clear for the past two years that analysts do not know how to gauge the growth coming from this company. In 2025, Blackwell is likely to wildly exceed analyst estimates again.

EPS:

This quarter, analysts are expecting EPS of $0.74 compared to EPS of $0.67 last quarter. For nearly two years, the company has beaten EPS estimates by 10% or more, yet in the last quarter, the beat was more muted at 5.7%.

On the topic of Nvidia having 2X better margins than most of the Mag 7, here is a glimpse of how Nvidia compares on EPS with a 35%+ growth rate compared to the Mag 7 reporting half this growth rate through 2026:

Nvidia: 35.5% 2Y revenue CAGR, 35.1% EPS CAGR

Apple: 7.1% revenue CAGR, 14.4% EPS CAGR

Microsoft: 14.2% revenue CAGR; 14.9% EPS CAGR

Amazon: 10.7% revenue CAGR; 22.3% EPS CAGR

Meta: 13.5% revenue CAGR; 12.5% EPS CAGR

Supply Constraints:

This quarter, Nvidia’s CFO Colette Kress, will not offer a full year guide yet have to address the elephant in the room — supply constraints.

The fab that makes Nvidia’s chips, Taiwan Semiconductor (TSMC), is working overtime to boost capacity to meet demand. TSMC’s monthly CoWoS capacity was estimated at ~15,000/month at the end of 2023, and was originally expected to triple to ~45,000 to 50,000/month by the end of 2024 in order to meet such high demand from Nvidia, AMD and other advanced node clients. Now, capacity is expected to rise ~300% to 60,000/month.

TSMC remains committed to significantly boosting CoWoS capacity over the next few years in order to accommodate these accelerated AI GPU timelines from both Nvidia and AMD, with multiple different product lines expected to come to market over the next couple of years. By year-end 2025, CoWoS capacity is estimated to be 80,000 to 90,000/month, per Morgan Stanley, with Nvidia reportedly already reserving half of this capacity.

By the end of 2026, CoWoS capacity is estimated to expand to as much as 140,000 to 150,000/month, representing 10x growth in capacity from the end of 2023.

Source: Beth_Kindig xAI

Foxconn and Quanta are also both signaling strong demand for Blackwell come 2025. Foxconn has said that they see “crazy” demand for Blackwell servers, and forecast AI servers to make up half of their overall server business in 2025. Foxconn has said that initial shipments are on time for Q4 before ramping much faster in Q1, with Quanta saying the same, that initial shipments are on schedule and will ramp in Q1.

Quanta sees triple-digit AI server growth through next year on the back of strong demand, with Deputy Spokesperson Carol Hsu saying that “recent capex guidance from top US hyperscalers also confirmed their aggressive spending on AI in 2025, all from a high base in 2024.”

Nvidia’s China Exposure is Low

Nvidia is the subject of some of the most severe export restrictions from the US due to its integral role in advancing AI computing. Subsequently, the company’s China exposure is among the lowest in the semiconductor sector, leaving it less exposed should we see heightened geopolitical tensions — especially tariffs.

Nvidia’s China revenue was 9.6% in Q1 and 12.2% in Q2, down from the low-20% range in the same quarters in fiscal 2024. For all of FY 2024, Nvidia’s China revenue was 16.9%, down from 21.5% the year prior. Other semi peers are much more heavily exposed to China: Broadcom’s China exposure was 32.2% in FY 2023, Intel’s exposure was above 27%, and Qualcomm and Marvell both had more than 40% of revenue stem from China in FY 2024.

Semiconductor Peers are Quite Weak

Although Nvidia’s fundamentals are a perfect 10, the stock is contending with weak peers, as evidenced by major semiconductor ETFs, SOXX and SMH, not making new highs with the S&P 500.

Retail investors often find out the hard way, even the most perfect stock must contend with market forces beyond its control. This is the primary reason Nvidia’s stock may pullback as Nvidia is holding up the semiconductor market, which has grown unusually weak in the past few weeks. SOXX is 20% of its all-time highs and SMH is 14% of its all-time highs despite the S&P 500 making new highs. In a 1-hour webinar for I/O Fund Members last quarter, I discussed why this is an issue for AI investors and what I’d like to see before I resume buying Nvidia.

Conclusion:

My firm has become well-known for calling Nvidia an AI stock in 2018, and later stating Nvidia would Surpass Apple, and finally that Nvidia will reach a $10 trillion market cap by 2030. Yet, perhaps lesser-known is that I nailed the October 2022 bottom by stating Nvidia was Ready to Rumble on H100 GPUs along with a real-time trade alert for $10.80 on October 13th 2022 a mere 25 months ago.

Here is what I stated at the exact moment Nvidia’s stock bottomed in October after selling off 60% following the August earnings report:

“Today, Nvidia’s AI products serve nearly every enterprise company’s artificial intelligence and machine learning ambitions. The company has an impressive launch schedule starting in October for two flagship products – the RTX 40 Series and the H100 GPU. The timing of these releases is no coincidence as it’s a rapid two months following the crypto/gaming revenue miss. Suffice to say, Nvidia’s management team is prepared to rumble —- putting its very best release in gaming and its most powerful AI chip to-date up against the crypto mining selloff. If history is any indication, the turnaround will only be a matter of time.”

The upcoming earnings report has a few similarities to October of 2022, which is that we are toward the end of a product cycle and the CFO cannot offer fiscal year guidance. Despite the H100s ramping and Nvidia having visibility into that ramp, the CFO was tight-lipped two years ago stating: “Our Data Center yes, we do expect it to grow. It may grow about what we just saw between Q1 and Q2. We’ll continue to look at it.” Therefore, I am not expecting much from the CFO on Blackwell in this report, but that lack of detail will be a distant memory this time next year.

Make no mistake, Nvidia is the best stock of the decade and we are only four years in. The big picture is that Nvidia's trajectory will continue due to two words: pricing power.

Our firm has an aggressive buy plan at key levels should the stock pullback, and we have a backup plan should the stock overcome the peer pressure we are seeing from SMH and meaningfully breakout.The keyword is “buy” but the skillset is patience. My firm has blended cutting-edge analysis alongside careful, patient buys for returns of 3280% since our first tranche. Most importantly, the I/O Fund continues to offer buy zones for those who’d like to participate.

The I/O Fund first called out Nvidia’s AI opportunity in November 2018 with our first trade alert at $3.15 for returns of 3280%. We also provided 9 buy alerts from 2021 – 2022 to buy NVDA stock below $20. The I/O Fund has been closely analyzing lesser-known stocks in AI plus crypto with real-time trade alerts and webinars. For a limited time, get up to $250 off with one of our biggest sales of the year starting Nov 28th. Sign up for our newsletter for more information on the upcoming sale or Follow me on xAI/Twitter.

If you would like notifications when my new articles are published, please hit the button below to "Follow" me.