This year’s keynote from Jensen Huang arrives at a moment when artificial intelligence is moving from experiment to infrastructure. The early phase of AI, often noted as the training of giant models, has already reshaped the technology landscape. But the next phase is about deploying intelligence everywhere: in enterprise software, robotics, autonomous systems, and global data centers.

Nvidia: The Operating System of the AI Economy

The company is no longer simply a semiconductor firm. Nvidia has quietly become the operating system of the AI economy. In fact, thinking about Nvidia solely as a company that builds chips risks missing the broader vision articulated by Jensen Huang. Since ChatGPT’s release in 2022, Nvidia has positioned itself as being heavily involved in the entire technological stack that modern AI depends on: silicon, networking, software frameworks, developer platforms, and increasingly the tools for building autonomous agents.

And the company has trained us that GTC is where the blueprint is revealed.

The 2026 event showcased the next generation of Nvidia’s AI infrastructure roadmap,³ centered around the Vera Rubin platform, the successor to Blackwell, and its expansion into full-scale “AI factory” systems. Rather than focusing solely on raw compute, Rubin is designed to optimize the economics of inference, dramatically increasing token throughput while lowering cost per token. Nvidia highlighted gains far beyond traditional scaling expectations, with systems engineered to deliver step-function improvements in performance per watt and overall token production.

Critically, Nvidia also unveiled a new architectural direction by integrating Groq’s LPU⁴ technology, featuring massive on-chip SRAM⁵ and deterministic execution, into the Rubin ecosystem. This combination is designed to extend performance into ultra-low-latency, high-value inference workloads, reinforcing Nvidia’s strategy of co-optimizing hardware, software, and system design around the emerging “AI factory” model of computing.

The Economics of AI Is Changing

The first wave of generative AI was about training models, an extremely compute-intensive process dominated by large technology firms. But the long-term opportunity lies in inference, the act of running those models billions or trillions of times every day in real-world applications. Many estimates suggest inference could represent the majority of the future AI compute market.

What’s Happening with Groq?

Nvidia reportedly spent roughly $20 billion to license Groq’s technology and bring its team in-house,⁶ a move that raised an obvious question heading into GTC: What, exactly, did they buy?

The keynote provided a clear answer. What Nvidia acquired was not just another chip, but a fundamentally different approach to inference. Groq’s Language Processing Unit (LPU) is built around deterministic execution and massive on-chip SRAM, minimizing reliance on external memory, such as High-Bandwidth Memory (HBM). The result is a system purpose-built for token generation, delivering ultra-low latency and extremely high tokens-per-second throughput for what were referenced as ‘AI factories’.

But the real insight is how Nvidia is deploying it. Rather than replacing GPUs, Groq’s architecture is being tightly integrated into the Vera Rubin platform, where it handles the most latency-sensitive portion of inference, token decoding, while GPUs manage the heavier compute and memory workloads.

This is a shift in how AI infrastructure is designed. Training created the first wave of AI demand, but inference, running models continuously at scale, is now the economic engine. By combining high-throughput GPUs with low-latency LPUs, Nvidia is effectively disaggregating inference itself.

What they bought wasn’t just speed. Nvidia bought a new layer in the AI stack—one optimized for the economics of intelligence at scale.

Is Nvidia a So-Called ‘Value’ Stock?

The company that we just noted is at the center of the AI story cannot possibly find itself in the ‘value’ style—or could it? In Figure 1, we look at certain statistics that may make a compelling case for this conclusion:

• Nvidia had a forward price-to-earnings (P/E) ratio of 20.4x, which is not hugely different from the overall figure for the S&P 500 Index, shown at 19.9x.
• For the overall S&P 500 Index, the estimated earnings growth for the next 12 months is 26.4%, but that same figure for Nvidia is estimated at 69.7%.
• If we try to look at longer-term estimates, we see that the overall S&P 500 Index is showing 15.9% estimated long-term earnings growth, whereas Nvidia is showing 25.7%.

Our bottom line is this—even if the growth forecasts for Nvidia’s earnings are too high, there is still such a gap between what is predicted for the company versus what is predicted for the broad S&P 500 Index that for investors to impute a similar forward P/E ratio for both, in our opinion, tilts the potential risk/reward trade-off towards Nvidia. Each of the so-called Magnificent 7 companies that has a higher forward P/E ratio than Nvidia also has a significantly lower forward-looking earnings growth forecast. Tesla, admittedly, is a bit of an outlier in that the forward P/E ratio is 182.8x, the estimated earnings growth for the forward 12 months is 9.4%, and then the estimated long-term earnings growth for Tesla shows around 32.9%. While this is higher than Nvidia’s figure, we recognize that Tesla is bringing all sorts of new things to the market to give this a chance to happen. Nvidia is delivering on a more established stream of business execution.

Figure 1: Valuation of the Magnificent 7 versus the Broader S&P 500 Index

Not All Indexes Think Alike

For many, the Nasdaq 100 Index, the total returns of which are tracked by Invesco QQQ Trust, Series 1 (QQQ), has been synonymous with exposure to big tech. But there's an important nuance hiding in plain sight: the Nasdaq 100 Index doesn't explicitly target growth or quality. It simply selects the 100 largest non-financial stocks trading on the Nasdaq exchange.⁷ This means that while QQQ gives investors access to scale, it does not optimize for company fundamentals. As the market begins to differentiate more sharply within the mega cap cohort, index design can become an active decision, even in passive wrappers.

QGRW: Quality Growth for the AI-First Economy

Enter the WisdomTree U.S. Quality Growth Fund (QGRW), a fundamentally driven alternative, designed to track the total return performance of the WisdomTree U.S. Quality Growth Index, which focuses toward the more profitable, higher-growth names in large-cap U.S. equities. There is a direct emphasis on return on equity, return on assets and earnings growth, as compared to Nasdaq 100 Index constituents, which are not included due to any direct requirements like this. For investors who believe the AI boom still has legs, but want a basket built on enduring fundamentals, not just market cap, QGRW may offer the right lens for this next leg of the tech trade.

The Top 10: A Setup for Differentiated Returns

In Figure 2, we look at the top 10 holdings of QGRW against those of QQQ. What stands out starkly is the difference in weight for Nvidia, as of February 28, 2026:

• QGRW had a weight of 13.99%.
• QQQ had a weight of 8.74%.

The reality is we do not know what Nvidia’s return will be, but it is clear that the two strategies have significantly different exposure which could lead to a greater potential for differentiated returns.

Figure 2: Top 10 Holdings for QGRW vs. QQQ

Figures 3a and 3b paint the picture of returns. Since QGRW's inception, which corresponds to the longest common period, QGRW was able to outperform QQQ by nearly 2% per year, it is clear in Figure 3a. However, in the early part of 2026, which has been characterized by a potential rotation away from growth stocks and more towards value stocks, QGRW has underperformed QQQ by about 2%, though both strategies were negative year-to-date as of this writing.

Figure 3a: A Tougher Environment (so far) to Start 2026 for Growth Stocks

Figure 3b: Standardized Returns

Conclusion: Positioning for the Next Phase of AI

Nvidia’s GTC 2026 keynote reinforced a simple but powerful idea: we are no longer investing in companies, but in the infrastructure of intelligence itself. As AI shifts from training to inference, from models to agents, and from tools to autonomous systems, the economic center of gravity is moving toward platforms that can scale intelligence efficiently. Nvidia sits at that center. But as leadership within mega-cap technology begins to differentiate, how investors access that exposure matters. Strategies like QGRW, which emphasize quality and growth alongside scale, may offer a more intentional way to participate in what could be one of the most consequential technology cycles of our time.

Figure 4: Additional Information

¹ Graphics processing unit.

² Source: NVIDIA Corporation. (2026, March 3). NVIDIA CEO Jensen Huang and global technology leaders to showcase age of AI at GTC 2026. NVIDIA Newsroom.

³ References to the 2026 Nvidia GTC keynote are sourced from: Huang, J. (2026, March 16).

⁴ LPU stands for language processing unit.

⁵ SRAM stands for static random access memory.

⁶ Source: Financial Times. (2026, March).

⁷ Source: Nasdaq, Inc. (2024). Nasdaq-100 Index methodology. Nasdaq Global Indexes.