View this email in your browser

76report

September 2, 2026

MAG7 MONITOR: NVIDIA’s High-Class Problem

The biggest problem NVIDIA (NVDA) now faces would be a dream come true for almost any other company—the business cannot physically keep up with customer demand.

After delivering 110% annual earnings growth in the second quarter, which it reported at the end of last week, NVDA guided the market to approximately 70% revenue growth in the next fiscal year.

This is extremely impressive. For perspective, the average company in the S&P 500 is expected to grow sales around 5% this year, with earnings growth just over 10%.

It is even more impressive because NVDA is already the most valuable business in the world, currently representing 8% of the S&P 500 Index. It is also one of the most profitable.

The company is generating extraordinary growth off an already enormous base.

Investors have responded positively to the good news, which also offered what may be the most important read-across for the entire AI complex.

As we write, NVDA shares are up approximately 8% since the earnings announcement, while tech stocks generally have also edged higher, despite sliding back a bit on turmoil in Iran.

The message from NVDA was not ambiguous—customer demand for AI computing capacity is growing, broadening and deepening.

The main challenge is keeping up with it.

NVDA indicated that customer forecasts point to enough demand for its business to roughly double next year—in other words, grow by 100%.

But management is forecasting revenue growth of only 70% because the company does not believe the supply chain can keep up.

To be fair, the fact that NVDA and the entire AI industry are facing supply constraints is genuinely negative in one respect. They are simply unable to sell as much equipment as customers want to buy.

Supply shortages also contribute to higher input costs, particularly for memory, which will pressure NVDA’s profit margins over the next several quarters.

But there is a much more encouraging side to this story.

Excess demand gives suppliers like NVDA substantial pricing power that should eventually allow them to pass through higher input costs.

More importantly, the existence of significant unmet demand provides a cushion against any potential slowdown in the growth of AI spending.

Valuation still attractive

NVDA has been a holding of our American Resilience Model Portfolio since March 12, 2025, when it was added at around $116 per share amid fears of a demand collapse related to DeepSeek.

Investors at the time were panicking that the Chinese AI startup had just proved that AI models could be developed much more efficiently, implying weaker future demand for AI hardware.

With the benefit of hindsight, those concerns appear truly misplaced.

Although NVDA shares have nearly doubled since then, we believe the stock remains attractively valued, taking into account its growth trajectory, competitive position, and extraordinary free cash flow generation.

The latest NVDA results also lend support to our broader optimism toward AI-related stocks.

When it comes to AI, everyone fears overspending and overbuilding. After all, this is what burst the dotcom bubble in the early 2000s. Those scars remain etched in the market’s collective memory.

But NVDA’s latest results demonstrate that demand for AI is real and growing—and we may indeed still be quite early.

Why the opportunity remains

NVDA generated $96.2 billion of revenue in the latest quarter, up 106% from a year ago. Data Center revenue reached $89 billion, representing growth of 117%.

Management expects total revenue to rise again to approximately $108 billion in the current quarter.

These growth rates would be remarkable for a young company growing off a small base. NVDA is producing them with an equity market value of approximately $5 trillion and quarterly revenue approaching $100 billion.

The scale of the demand is difficult to overstate.

Capital spending by the five largest hyperscalers is expected to increase from nearly $800 billion in 2026 to approximately $1.3 trillion in 2027. Cloud industry backlog already exceeds $2 trillion.

Yet NVIDIA is also becoming less dependent on this relatively small group of technology giants.

Its AI Cloud, Consumer Internet and Enterprise (ACIE) segment—which includes neocloud, industrial and enterprise customers—generated $40 billion of revenue last quarter. Revenue rose 25% sequentially and 138% from a year ago, considerably faster than the hyperscaler business.

The first phase of the AI boom was driven primarily by major hyperscalers like Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL), Meta (META) and a handful of AI laboratories building enormous computing clusters.

The next phase will be much broader.

Enterprises, governments, startups and specialized AI cloud providers are increasingly building or purchasing access to their own AI infrastructure. NVDA believes these non-hyperscaler customers could eventually represent as much as half of its Data Center business.

“Running flat out”

NVDA expects supply to remain a bottleneck through at least the end of fiscal 2028.

This is not simply a shortage of NVDA chips. The constraints extend across memory, advanced semiconductor packaging, networking equipment, optical components, electrical infrastructure, cooling systems and data center capacity.

Our entire supply chain is challenged, and everybody is really running flat out. - Jensen Huang, NVDA CEO (8/26/2026)

We view NVDA’s enormous balance sheet and cash generation as tremendous competitive advantages in an industry scrambling for scarce capacity.

NVDA has made approximately $279 billion of commitments to secure supply and capacity through 2032. The company’s commitments increased by roughly $160 billion during the latest quarter alone, with memory representing well over half of the total.

This is an enormous financial commitment. It also reflects NVDA’s confidence that the demand will be there.

The most visible near-term consequence is pressure on profitability.

Gross margin was 75% last quarter. Management expects it to decline to approximately 74% in the current quarter and bottom between 71% and 72% in the fourth quarter before recovering toward 72% to 73% in fiscal 2028.

Rapidly rising memory prices are the main culprit.

Although only a few percentage points, these lower margins should not be dismissed. NVDA is absorbing some of the cost inflation, and its ability to pass those costs along will be important to monitor.

But the source of the margin pressure matters.

This is not margin compression caused by weak demand, inventory discounting or deteriorating competitive conditions. Memory is becoming scarce largely because the AI buildout itself is consuming so much of it.

NVDA has already announced price increases that should begin offsetting these costs next year.

Capturing more of every AI factory

NVDA is not merely selling more GPUs. It is capturing an increasingly large share of the total spending required to build an AI data center.

The company estimates that its revenue opportunity has expanded as it has introduced next generation computing platforms, growing from approximately $18 billion per gigawatt with Hopper to $25 billion with Blackwell—and potentially $40 billion with Vera Rubin.

That increase reflects NVDA’s evolution from a chip supplier into a provider of comprehensive AI infrastructure.

The Vera Rubin platform includes GPUs, the new Vera CPU, high-speed networking, systems and software. Rubin is already in production, with purchase orders from every major hyperscaler, AI cloud and systems manufacturer.

NVIDIA’s server CPU business provides another example of this widening opportunity. The company already has approximately $20 billion of indicated demand and expects CPU revenue to more than double in fiscal 2028.

Growth changes the valuation equation

A $5 trillion market capitalization naturally sounds expensive. But a company’s size alone does not determine whether its shares are attractively valued.

What matters is the relationship between the share price and the earnings the business ultimately produces.

Based on the most recent closing price, NVDA traded at roughly 23 times projected fiscal 2027 earnings, only marginally higher than the 21 times P/E multiple of the S&P 500. That is far from an excessive valuation for a company expected to generate 60% earnings growth this this year.

The valuation becomes even more striking as we look farther ahead.

Current estimates call for approximately $15 of earnings per share in fiscal 2028 (which corresponds with the 2027 calendar year). This places the stock at only around 15 times forward earnings.

NVDA vs. S&P500

(2-Year Forward P/E - Last 3 Years)

Based on fiscal year 2029 estimates, which incorporate more of the anticipated growth, the multiple falls closer to 10 times earnings.

Those are unusually low multiples for a company with NVDA’s competitive position, margins and long-term opportunities.

If NVDA delivers anything close to the earnings growth now anticipated, its current valuation will prove remarkably modest. Even after the stock’s extraordinary appreciation, earnings expectations are rising so rapidly that the forward valuation is compressing.

NVIDIA also generates enormous free cash flow.

The company returned a record $26 billion to shareholders last quarter, including $20 billion of share repurchases.

NVDA’s financial strength allows it to secure scarce supply, invest in strategic partners (like the reported $13 billion acquisition of Hugging Face, a leader in open-source AI), expand its technology platform, and return excess capital to shareholders simultaneously.

This includes a modest annualized dividend of $1 per share at the moment, which has the potential to grow significantly over time as it only represents a small fraction of current earnings.

What NVIDIA also tells us about AI

NVDA’s latest results reinforce our optimism toward AI-related stocks.

There are legitimate questions about circular financing, extended payment terms and the increasingly aggressive capital commitments being made throughout the AI ecosystem.

But NVDA’s results indicate that the more immediate problem is not too much available computing capacity. It is too little.

AI infrastructure is also beginning to produce measurable economic value.

NVDA noted that as its hyperscaler customers brought more computing capacity online, they simultaneously reported accelerating revenue growth and expanding profit margins.

Inference and agentic AI could make this connection even stronger. Unlike model training, which requires a large burst of computing power upfront, AI agents continuously consume compute as they reason, use software and complete tasks.

AI is now doing productive and useful work. AI is generating profitable tokens. And if we had more compute, we could generate more profitable tokens. - Jensen Huang, NVDA CEO (8/26/2026)

If this relationship holds, compute increasingly becomes productive capital—an investment that allows businesses to generate revenue, automate work and improve productivity.

That would support not only NVDA, but also memory manufacturers, semiconductor equipment companies, networking suppliers, electrical equipment providers, utilities, data centers and the large technology companies building AI services.

There will inevitably be setbacks. No capital spending cycle grows in a straight line, and the enormous sums now being committed will produce both winners and costly mistakes.

But for now, there is considerably more demand for intelligence than the world has the computing capacity to produce.

Investors should also remember that it has been less than four years since the public launch of ChatGPT unofficially ushered in the AI era. We are still in the early stages of a technological revolution that is likely to unfold over the rest of our lives.

MAG 7 — Valuation

MAG7 — Performance

NVIDIA (NVDA): Company Snapshot

Click HERE to learn more about our Model Portfolio subscription plans.

Disclaimer

FOR SUBSCRIBER USE ONLY. DO NOT FORWARD OR SHARE.