Nvidia Has Been Hammered All Year, The Analysts Who Called the Bottom Just Made Their Move

You would have had an odd impression of Nvidia earlier this spring if you had strolled past the trading desks in Lower Manhattan. The atmosphere surrounding the AI trade had changed, but the screens were still occasionally flashing green. There were fewer phone calls. The conversations became more subdued. By the end of March, the stock had fallen so much that even ardent supporters were starting to mutter the word that everyone in the tech industry despises: correction. Looking back, it’s still difficult to determine if what transpired was a momentary loss of nerve or a sober reset.

When the recovery did occur, it was not gradual. Over a thirty-day period, shares increased by about 19%, and at one point they completed a 32% rally off the late-March low. Early in May, Nvidia was once more vying for its all-time high of $216, which had seemed like a distant memory only a few weeks prior. Investors who persevered through the worst of it are now witnessing the same analysts who predicted the bottom return to prominence with calls that are louder, stranger, and more ambitious.

Never one to back down from a prediction, Brad Gerstner of Altimeter Capital told CNBC that he thinks Nvidia will be the first $10 trillion company. The I/O Fund’s Beth Kindig went so far as to say that the company might be worth $20 trillion by 2030 thanks to a GPU refresh cycle that currently occurs every 12 to 18 months, while custom silicon languishes at three to five. It’s an eye-catching assertion that seems almost too tidy, but the statistics that support it are difficult to ignore.

Think about the last quarter. With the data center segment alone reaching $62.3 billion, up 75%, revenue increased 73% year over year to $68.1 billion. Executives frequently make statements like “demand for AI infrastructure continues to exceed expectations,” but they are rarely supported by margins that rise above 75%, according to CFO Colette Kress. With a first-quarter revenue forecast of $78 billion, plus or minus 2%, which completely ignores Chinese data center compute, the guidance was even more optimistic. Reading between Kress’s lines gives the impression that management is purposefully cautious—almost protective—about how much optimism they put on the page.

In a recent note, Joseph Moore of Morgan Stanley described the atmosphere in data centers, noting that clients frequently test a less expensive, custom accelerator, run it in production, and then covertly return to Nvidia. There is more weight in that one sentence than in any chart. It explains why Nvidia’s share of AI accelerator sales remained stable at 86% through 2025 despite years of ASIC competition from Alphabet, Amazon, Microsoft, and Meta. No matter how much capex the hyperscalers invest in custom silicon, software ecosystems built over two decades don’t collapse in a quarter.

The difference between the macro numbers underneath the stock and the actual experience of holding it is what makes this moment unique. The top five hyperscalers were expected to see a 19% increase in capital expenditures this year. With $805 billion in 2026 and $1.1 trillion projected for 2027, Morgan Stanley believes it is closer to 80% than the current consensus of 60%. Ten years ago, when no one could quite agree whether Tesla was a car company, a battery company, or a story, the company experienced similar waves of doubt. Nvidia is currently in a similar state of uncertainty, serving as both the operating system for an entire build-out and a chip manufacturer.

As this develops, it’s difficult to avoid feeling the peculiar weight of conviction returning to a trade that seemed brittle just a few weeks ago. The analysts who predicted the bottom have already moved, regardless of whether the next leg shows up in three years or never quite materializes as described by Gerstner and Kindig. The remainder of the market is still debating whether or not to follow.