AI Boom vs. Tight Liquidity: Will the US Stock Rally Continue?

钛媒体APP

joined discussion ·

NVIDIA is extending the AI bull market, but the market is starting to scrutinize the books

(The author of this article is the US Stock Research Society, published by Titanium Media with authorization)

By the US Stock Research Society

This earnings report from NVIDIA is somewhat like that of a student who has consistently ranked first in class for years

Revenue came in at $81.62 billion, up 85% year-over-year; data center revenue reached $75.2 billion, surging 92% YoY; adjusted EPS was $1.87, up 140% YoY; gross margin stood at 75%, exceeding market expectations; and the company guided for $91 billion in revenue next quarter—another record high.

If this were any other company, it would basically be a knockout punch.

But with NVIDIA, the market reaction was nuanced. Shares dropped more than 3% in after-hours trading. The reason is simple: Wall Street is no longer satisfied with NVIDIA just 'scoring well'—it wants to see whether NVIDIA can keep pushing its results into the stratosphere.

This is precisely what makes NVIDIA unique right now:

It’s no longer just an ordinary tech company or even a typical semiconductor stock—it now functions as the consolidated ledger for the entire AI capital expenditure cycle. Whenever NVIDIA reports earnings, the market isn’t just looking at one company’s revenue; it’s assessing whether the global AI supply chain remains hot, whether cloud providers are still pouring money in, and whether segments like HBM, CoWoS, optical modules, power infrastructure, liquid cooling, and networking chips can continue to rally.

My judgment is clear:NVIDIA’s AI strategy has now entered the phase of financial realization—not just valuation storytelling.

But that’s also where the problem lies. The more it delivers, the more critical the market becomes.

Many companies are now talking up AI in a big way, making it sound as if they’ll completely reinvent themselves over the next decade. But when you look at their financial reports, revenue hasn’t changed much, and neither has profit—AI remains mostly a buzzword in PowerPoint slides.

NVIDIA is different.

The strongest part of this Q1 earnings report is that AI has directly translated into revenue and profit. Data center revenue came in at $75.2 billion, accounting for more than 90% of the company’s total sales. In other words, NVIDIA’s current growth is essentially driven by demand for AI computing power.

More importantly, even within data centers, there’s differentiation emerging.

Computing revenue reached $60.4 billion, up 77% year-over-year; networking revenue hit $14.8 billion, surging 199% year-over-year. I believe these figures are extremely significant. In the past, the market focused primarily on NVIDIA’s GPUs. Now, while GPUs remain the engine, networking, interconnects, switches, and system-level solutions are becoming new sources of profit elasticity.

What does this mean?

What NVIDIA sells is no longer just 'a graphics card'—it’s an entire infrastructure stack for AI factories. Customers aren’t just buying GPU performance; they’re buying training speed, inference efficiency, cluster reliability, energy consumption, deployment timelines, software ecosystems, and total cost of ownership.

Jensen Huang made this point very clearly on the earnings call: customers are building AI factories, and what really matters isn’t the price of an individual GPU, but how many tokens can be produced per watt or how much intelligence can be generated per dollar spent.

That might sound like marketing speak, but capital markets love this kind of messaging. It lifts NVIDIA out of the traditional semiconductor cycle and places it squarely into an infrastructure return-on-investment framework.

Chip companies ride hardware cycles, but AI factory platforms benefit from global capital expenditure. The valuation logic is entirely different.

I’ve always felt that many people underestimate NVIDIA’s ambition.

If it were merely selling GPUs, it would inevitably face two pressures: one from cloud providers developing their own chips, and the other from competitors like AMD, Intel, Broadcom, and Marvell carving up the market.

But what NVIDIA is doing now is positioning itself as the 'compute operating system' of the AI era.

The GPU is just the entry point. Behind it are CUDA, NVLink, InfiniBand, Spectrum-X, BlueField, Dynamo, Vera CPU, the Rubin platform, and an increasingly critical data center networking stack.

During this earnings call, the Vera CPU actually emerged as a very significant new development. Management stated that Vera unlocks a $200 billion market NVIDIA hadn’t previously entered in earnest, and they already see visibility into nearly $20 billion in CPU revenue this year.

If this materializes, NVIDIA won’t be relying solely on GPUs for further growth. It will add a CPU revenue curve on top of its existing business, layered with networking, systems, software, and edge AI.

In my view, this is the most important AI signal for investors in this earnings report:NVIDIA is proactively expanding its valuation anchor.

Previously, the market valued NVIDIA based on Blackwell shipments, HBM supply, CoWoS capacity, and cloud providers’ capital expenditures. Now, it must also factor in Vera, Rubin, data center networking, AI Cloud, sovereign AI, enterprise AI, and physical AI.

NVIDIA has also revised its financial reporting framework, splitting its business into two major platforms: Data Center and Edge Computing, with the Data Center segment further broken down into Hyperscale and ACIE.

This move is more than just a change in reporting methodology.

Hyperscale refers to mega cloud customers such as Microsoft, Amazon, Google, and Meta; ACIE encompasses AI Cloud, Industrial AI, Enterprise AI, and Sovereign AI. NVIDIA is essentially signaling to the market: I’m not solely reliant on a few large cloud providers—I’m expanding my customer base.

This is critically important for valuation.

If NVIDIA were only tied to a handful of major cloud customers, the market would inevitably worry about a slowdown in capital expenditures (Capex). If Microsoft, Meta, or Google announced plans to rein in their Capex, NVIDIA’s valuation would take a hit.

But if ACIE continues to grow—with AI Cloud, enterprise clients, sovereign AI initiatives, and industrial customers picking up the baton—NVIDIA’s growth narrative will shift from a 'big cloud procurement cycle' to a 'global AI infrastructure expansion cycle.'

These are two entirely different valuation frameworks.

The most intriguing part now is that despite delivering a very strong earnings report, NVIDIA’s stock price didn’t surge immediately.

This isn’t because the market failed to understand the earnings report—it’s because the market already understands NVIDIA all too well.

In the past, whenever NVIDIA beat expectations, capital would rush in. That no longer works now—because 'NVIDIA beating expectations' has itself become the consensus forecast.

Bank of America previously highlighted an interesting metric: over the past ten quarters, NVIDIA’s actual revenue has averaged 7% to 8% above management’s guidance. Based on this historical pattern, Q1 revenue would need to reach $83–84 billion to truly exceed buy-side expectations.

NVIDIA reported actual revenue of $81.62 billion—above Wall Street’s average estimate but below the threshold anticipated by the most optimistic investors.

That’s why the stock turned lower in after-hours trading.

The market isn’t saying this earnings report was bad. It’s saying: 'It’s good, but not good enough to justify significantly raising my 2027 and 2028 earnings models.'

This is NVIDIA’s current dilemma—and also a testament to its strength.

Ordinary companies rely on storytelling to repair their valuations, but NVIDIA has entered a phase where it must deliver a positive earnings surprise every single quarter.

The next quarter’s revenue guidance of $91 billion carries the same nuance: it’s above the market consensus but below the most optimistic expectation of $96 billion. The company also explicitly stated that this guidance excludes data center computing revenue from China.

That statement actually embeds an option.

If Chinese revenue recovers in the future, NVIDIA has room for further upward revisions; if China continues to generate no revenue, the company can still support the $91 billion figure through non-China markets. For bulls, this signals absurdly strong demand; for cautious investors, it means China-related uncertainty remains.

In my view, over the next two quarters, investors should focus not on whether NVIDIA will keep talking about AI, but on several harder metrics.

First, can Q2 actual revenue significantly exceed the $91 billion guidance? What NVIDIA needs now isn't just a beat—it's a substantial beat.

Second, can gross margin hold around 75%? With Blackwell ramping up, the Vera Rubin transition, and high costs for HBM and advanced packaging, maintaining gross margin would indicate NVIDIA still has strong pricing power.

Third, can data center networking continue its high growth? Networking revenue grew 199% year-over-year—the clearest sign yet that AI factories are moving from 'buying chips' to 'buying systems.'

Fourth, can ACIE continue to outperform? As long as AI cloud, enterprise AI, and sovereign AI keep scaling, the market will believe NVIDIA isn’t reliant solely on hyperscale cloud providers.

Fifth, can the Vera CPU deliver as management expects? With nearly $20 billion in visible revenue this year, successful execution would mark NVIDIA’s next growth curve.

Finally, China revenue also matters. If H200 or other products regain access to the Chinese market, it would serve as an additional catalyst.

My takeaway from this earnings report is that NVIDIA hasn’t cooled down the AI rally; instead, it has given the AI bull market another breath of life.

But that breath isn’t for all AI stocks.

It continues along the AI infrastructure chain: HBM, CoWoS, advanced packaging, optical modules, Ethernet switching, ASICs, liquid cooling, power supply, transformers, and data centers. Capital will likely no longer focus solely on NVIDIA itself but instead continue spreading into these bottleneck assets.

This is also my final assessment of NVIDIA’s latest earnings report:

AI’s greatest value to NVIDIA isn’t boosting its valuation, but rather reshaping its revenue structure and growth model.

Valuation is merely the outcome; growth is the foundation.

NVIDIA has already proven it didn’t rise on AI hype alone—it has genuinely turned AI into revenue, profit, cash flow, and shareholder returns. Announcing an additional $80 billion in share buybacks and raising its dividend from $0.01 to $0.25 per share isn’t something a company still burning cash for narrative sake could do.

But the market will also become more realistic going forward.

NVIDIA has shifted from being the 'biggest winner in AI' to becoming the 'ultimate validator of the AI investment cycle.' Each quarter, it must not only prove its own strength but also demonstrate that global AI capital expenditures are still rising, cloud providers remain willing to spend, inference demand continues to surge, and networking and CPUs can unlock new markets.

Therefore, I believe NVIDIA’s current investment narrative can be summarized as follows:

It is no longer an AI-themed stock but rather the most critical revenue gateway—the 'infrastructure tax' collector—of the AI era. However, this gateway is too expensive, too crowded, and burdened with excessive expectations, so every earnings report is no longer just about delivering results—it’s an audit by the entire market.

252K Views