The blogger who helped spark Nvidia’s $600 billion stock collapse and a panic in Silicon Valley

Table of Contents

Last Friday evening, Jeffrey Emanuel spent hours writing a blog post at home in Brooklyn while his wife took care of their kids, bringing him snacks. He worked late into the night, and by early Saturday morning, he had written approximately 12,000 words.

He took a peek at his personal blog's traffic later that Saturday morning, and noticed that 35 people were reading the post online via his YouTube transcription website.

Then the post suddenly became extremely popular online.

He received nearly half a million views. He also gained approximately 13,000 followers on the platform, increasing from around 2,000 to over 15,000 followers.

In an interview with , Emanuel stated that at one point, the traffic to his website crashed the site itself, causing visitors to share an archived link.

One notable trend was that the city with the most concurrent readers at the end of the night was San Jose, California - a city near Nvidia's corporate headquarters.

Emanuel's argument startled Silicon Valley because he didn't claim that the major US tech companies were dishonest or misleading. His main point was that they were not nearly as advanced or efficient as Wall Street was portraying them to be. The large tech companies had developed and trained their artificial-intelligence breakthroughs with enormous amounts of data and cutting-edge computing resources, which required them to invest in Nvidia's data-center hardware at very high profit margins. Emanuel noted that a China-based company, DeepSEEK, had recently released a premium AI product using fewer expensive chips, thereby achieving the same result as the major AI companies at a fraction of the cost.

On Monday, things really took a turn for the worse. Nvidia's stock took a huge hit, falling by about 12.5% at the opening of the market and keeping on falling after that. By the end of the day, the stock loss had taken almost $600 billion away from Nvidia's market value, making it the biggest one-day decline in market value of any company ever recorded.

It's claimed Emanuel's post was a significant contributor to the stock market downturn and is being considered one of the most impactful short research reports in history.

Emanuel spent the rest of the week fully booked, with hedge funds paying him $1,000 per hour to speak on the phone and share his thoughts on Nvidia and artificial intelligence.

“I'm exhausted, I've almost completely lost my voice,” Emanuel said. “This has been the most surreal experience I've ever had.”

Contrarian call

Emanuel mentioned in his blog post, “The Short Case for Nvidia Stock,” that he has professional experience working in the financial markets. A native of New Rochelle, New York, Emanuel studied math at Reed College in Portland, Oregon, before working on Wall Street, where he held positions at several large investment funds, including Millennium Management and Balyasny Asset Management, two of the biggest multimanager hedge funds.

At 42, Emanuel stated to that he has been fascinated with neural networks since 1998 and is an early adopter of both cryptocurrency and AI. In 2021, Emanuel left the financial industry and formed Pastel Networks, a company focused on blockchain, decentralized storage, AI solutions, and other Web3 tools for developers. While he monitored advancements in Silicon Valley and the stock market, his experience between finance and tech led him to believe Nvidia's stock is overvalued.

Last Friday, he was talking to a friend who works with a company that manages investments about why he thought Nvidia's period of being more successful than others was coming to an end. This was just a few days after a newer version of a machine learning program called DeepSeek was released, and still, no one on Wall Street was paying attention.

All the major banks have an extremely positive stock rating on Nvidia. It's like the blind are leading the blind here - they really have no idea what they're doing," Emanuel told . "Their entire reasoning has now completely fallen out of touch with reality.

reviewed this claim and discovered that as of January 31 in the morning, 61 out of 67 analysts rated Nvidia as a "buy".Six analysts, including one from Deutsche Bank, assigned the stock a "hold" rating. No analysts gave it a "sell" rating.

“They justify their position by saying we consult with industry experts. But that's equivalent to asking a barber if you need a haircut.”

Investments made by individual investors led to a substantial 171.2% return, greatly contributing to a 23.3% increase in the S&P 500 last year.

Any copycats who try to gobble up Nvidia's hardware to build their AI data centers are expected to keep throwing money at it, as long as Nvidia's GPUs remain superior to the competition.

The short case

The main point of Emanuel’s argument is this:

Several leading tech companies believe that the most significant advancements in technology since the widespread adoption of the internet are in the areas of deep learning and AI. As a result, these companies are working to integrate this technology into their operations, which requires building and training their AI systems. This process demands a substantial amount of data and computing power. Nvidia is a key supplier of the necessary hardware, and the margins on its most advanced chips are substantial.

But a few things are changing that may not make this model sustainable, Emanuel says.

One concern is that AI companies have been relying on scaling laws that essentially state the more data used to train a model, the more effective it becomes. However, Emanuel suggests that the industry may be running low on quality data to train these AI models — a potential "data wall" is emerging that could slow down AI progress and consequently require fewer resources for training.

Emanuel also inquired about the fate of the training equipment after the AI is trained. As graphics processing units (GPUs) continually become more powerful and advanced, companies might soon find it economical to discard older equipment once a few years have passed, making them frequently spend more money to obtain the latest technology. Nevertheless, these companies will ultimately demand a return on their significant investment.

Some of these companies, like Alphabet, have also been investing in developing their own semiconductor chips. For a while, Nvidia's hardware has been the best for training AI systems, but this might not last as more companies, such as Cerebras, create better hardware. And other GPU manufacturers like Advanced Micro Devices are updating their drivers to be more competitive with Nvidia's.

DeepSeek launched its own AI, which is on a par with OpenAI's ChatGPT, but the key advantage was that it was trained in less time using fewer chips.

Will DeepSeek signal a doomsday scenario for Nvidia and other AI stocks? Here’s what to consider.

Put all these things together — unsustainable spending and expanding data centers, less data to work with, more efficient hardware, and better competing AI technology — and what you get is a future where it's harder to imagine Nvidia's customers continuing to spend as much as they do on Nvidia hardware.

I thought I was done writing my article when I said to myself, 'I'm convinced,'" Emanuel told . "It wasn't until I realized that all of their major hyperscaler customers were actually making their own competitive chipsets, which were manufactured by Taiwan Semiconductor, and they were about to hit the market. I was thinking, 'Do people realize this?' I don't think they do.

If the cost of training and integrating AI significantly decreases, wouldn't major tech companies like Google, Amazon, and Microsoft still invest heavily in AI technology, spending enormous amounts of money on its development?

This made Emanuel wonder why Nvidia's stock was trading at such a high price based on its earnings.

When you know an investment is going to have only a couple years of huge returns, you wouldn't apply a very high earnings multiple. You certainly wouldn't put a multiple of 30 on it.

Nvidia deemed DeepSeek "an excellent AI advance and a perfect example of Test Time Scaling," however it noted that utilizing a trained model to make predictions on new data, or inference, "requires significant numbers of Nvidia GPUs and high-performance networking."

The aftermath

published on Dec. 26.

Knowledge about DeepSeek was circulating for a few weeks before the market downturn. Meta Chief Executive Mark Zuckerberg mentioned DeepSeek on the Jan. 10 Joe Rogan podcast, and Scale AI Chief Executive Alexandr Wang discussed DeepSeek and competing Chinese AI companies on Jan. 23. Nvidia finished the week with higher stock prices.

What was the factor that led to Nvidia's sudden crash on Monday? Emanuel has another idea that could shed some insight into it.

He drew attention to the fact that many people in San Jose were reading his blog post. He suggested that many of them were Nvidia employees with a large amount of Nvidia stock tied up in employee stock options worth thousands - or even millions - of dollars. He speculated that with such a large sum of money invested in a single asset, many were likely considering whether to hold the stock or cash in on their profits before they disappear. He believes his blog post may have influenced some of them to sell.

A lot of the selling pressure you saw on Monday morning wasn't necessarily what you'd expect. I think a significant portion of that came from shares that had essentially been dormant because they'd been stored in Schwab corporate accounts - specifically, restricted stock units that some employees had received around 2003 for free.

It's debatable whether Emanuel's post directly led Nvidia employees to sell, but given that it went viral within the tech-savvy community on X and considering many of the readers were based in San Jose, it's certainly plausible that it may have had some impact.

Emanuel informed that he had received a large number of meeting requests from personnel at various hedge funds who were impressed by his blog post and desired a discussion with him about AI. This disclosed that people on Wall Street had noticed his blog post and appeared to have made trades on that Monday based on the information he had provided.

Those Wall Street firms were more than happy to pay a pretty penny for Emanuel's consulting services, making it clear they took his ideas seriously. Emanuel had a history of doing contract work as an investment analyst, but after that blog post went viral, he raised his hourly rate to a whopping $500. When his schedule filled up, he further increased it to a staggering $1,000 an hour.

Emanuel has been collecting contracting fees as a result of his bearish forecast on Nvidia. However, he stated that he never actually held a short position in the stock.

When I'm writing about this topic, it's really an exercise in curiosity and learning, not something I'm doing in hopes of turning a profit. I didn't go into this with a specific viewpoint or purpose in mind. In fact, I'm more than optimistic about artificial intelligence than just about anything.

While the points Emanuel made in his blog post could be seen as negative for Nvidia, he still views them as a positive indication of AI's future.