When Do You Start Paying Attention? A Lesson from NVIDIA’s 2018 Financial Results

The Number That Started This

In fiscal 2018, which was the year ended January 28, 2018, NVIDIA generated $9.71 billion in revenue.¹ By fiscal 2026, ended January 25, 2026, that figure had reached $215.94 billion.² That is roughly 22 times larger, a compound annual growth rate of approximately 47% sustained over eight consecutive years.

To put a finer point on it, NVIDIA’s fourth-quarter revenue alone in fiscal 2026, $68.1 billion, was about seven times the company’s entire annual revenue eight years earlier.³

I did not pick 2018 as a starting point at random. That was the year when WisdomTree, globally speaking, began treating artificial intelligence as a distinct, investable theme. We were not early by the standards of Silicon Valley insiders, but we were early relative to the broad market’s eventual embrace of the idea. What strikes me now, looking back, is how little of the technical foundation for what actually happened was visible to us, or to almost anyone outside a narrow research community, at that moment.

The Paper Nobody Outside the Field Was Reading

In June 2017, eight researchers at Google published a paper titled “Attention Is All You Need.”⁴ It introduced the Transformer architecture, which is a way of processing sequences using only attention mechanisms, discarding the recurrent and convolutional structures that had dominated natural language processing until then. The paper is, by later standards, almost modest—just eight pages, a translation benchmark and a somewhat cheeky title.

It is also the architecture underneath essentially every large language model that exists today. GPT stands for Generative Pre-trained Transformer. The “T” is not incidental. Without that 2017 paper, there is no ChatGPT, no Claude, no Astra and no coherent story for why NVIDIA’s data center revenue went where it went.

ChatGPT launched to the public on November 30, 2022.⁵ That is roughly five and a half years after the Transformer paper appeared. For almost all of that interval, the idea existed, the architecture existed, researchers were building on it, and the broader investing public had no reason to notice. There was no product, no headline and no obvious reason for revenue to move. The scale-up that mattered, meaning the compute, the data and the engineering to turn an architecture into a consumer-facing capability, happened quietly, in a window most of us were not watching.

I want to be careful here, because I think this is exactly the point where reasoning like mine tends to go wrong.

It is tempting, in hindsight, to make the story sound more inevitable than it was.

Plenty of architectures and papers from that era went nowhere. The Transformer was not obviously “the one” in 2017, even to many people working adjacent to it. The lesson is not “brilliant papers always compound into trillion-dollar outcomes.” The lesson is narrower and potentially more useful.

Transformative technological shifts are frequently preceded by a multi-year period in which the foundational science is published, understood by specialists and essentially invisible to everyone else, including to capital markets.

Where This Leaves Quantum Computing

I do not know whether quantum computing is in that kind of window right now. Nobody does, and I am suspicious of anyone who claims certainty in either direction. What I can say is that the pattern-matching is uncomfortably familiar. Over the past two years, the research coming out of groups working on error correction, logical qubits and specific problem classes, be it cryptography, chemistry or optimization to name a few, has moved from “interesting in principle” to “here is a specific, falsifiable claim about what a machine with N logical qubits could do.” Much of it is still confined to papers, preprints and specialist conferences. Very little of it has produced a headline that reaches a general audience the way ChatGPT did in November 2022.

That gap, between what is happening in the research literature and what is visible to the broader market, is precisely the gap that existed for transformers between 2017 and roughly 2022. It does not tell us when, or even whether, quantum computing crosses from research curiosity to commercial inflection. Revenue for the sector today is a rounding error next to NVIDIA’s, and the hardware challenges are arguably harder than anything language models faced. But the structural shape of “quiet research phase, then a scale-up that becomes visible all at once” is a pattern worth taking seriously precisely because we have just watched it happen, in real time, with a company whose income statement we can all pull up.

What This Actually Argues For

I am not arguing for a specific quantum computing trade, and I do not think this framework tells you which company, if any, becomes the “NVIDIA of quantum computing.” What it argues for is something more basic.

In my view, it argues for paying attention during the quiet phase, before the evidence is obvious, because the evidence tends to become obvious only after most of the value has already been created.

NVIDIA’s stock did not wait for fiscal 2026 revenue to be reported before it moved. By the time $215.94 billion showed up on an income statement, the repricing had largely already happened.

If there is a lesson in the 2018-to-2026 numbers, it is not “predict the future.” It is “notice which parts of your own research process would have caught this the first time, and are you running that process now, on the fields where you genuinely do not yet know the answer.” Quantum computing could be one candidate.

¹ Source: NVIDIA Corporation. (2018). Annual report (Form 10-K) for the fiscal year ended January 28, 2018. U.S. Securities and Exchange Commission.

² Source: NVIDIA Corporation. (2026). Annual report (Form 10-K) for the fiscal year ended January 25, 2026. U.S. Securities and Exchange Commission.

³ Source: NVIDIA Corporation, 2026.

⁴ Source: Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. arXiv.

⁵ Source: OpenAI. (2022, November 30). Introducing ChatGPT.

Investors should carefully consider the investment objectives, risks, charges and expenses of the Funds before investing. U.S. investors only: To obtain a prospectus containing this and other important information, please call 866.909.9473, or click here to view or download a prospectus online. Read the prospectus carefully before you invest. There are risks involved with investing, including the possible loss of principal. Past performance does not guarantee future results.

You cannot invest directly in an index.

Foreign investing involves currency, political and economic risk. Funds focusing on a single country, sector and/or funds that emphasize investments in smaller companies may experience greater price volatility. Investments in emerging markets, real estate, currency, fixed income and alternative investments include additional risks. Due to the investment strategy of certain Funds, they may make higher capital gain distributions than other ETFs. Please see prospectus for discussion of risks.

WisdomTree Funds are distributed by Foreside Fund Services, LLC, in the U.S.

---

A message from Advisor Perspectives and VettaFi: Discover something new! Click here to register for our upcoming webcasts.

© WisdomTree, Inc.

More ETF Topics >