Nvidia is everything good and bad about the US stock market in 2026

All the good and bad things about the US stock market in 2026 can be found in Nvidia. On steroids. 

The character of the AI trade has changed this year, becoming much more zero (even negative!) sum. Traders only seem eager to bid up stocks benefiting from acute AI-driven shortages (like memory), while punishing companies forced to accumulate these inputs at higher prices. And sellers are quick to make an example of the companies potentially disrupted by AI (see: software, or any industry Anthropic has referenced).

The conundrum with Nvidia is that it’s all of the above. It’s a massive buyer of memory chips, which are utilized in its racks, while the GPUs — the starring players in those racks — are persistently in short supply amid hot demand.

It’s been an AI winner, the epicenter of the AI boom, even. But the chip designer’s once unquestioned dominance faces pointed queries given how Google’s Gemini 3 (trained on custom TPUs) drew widespread praise, and OpenAI was reportedly “unsatisfied” with how its chips perform in inference. Meta’s huge deal to buy AI infrastructure from Advanced Micro Devices, the No. 2 in GPUs, also has shares of Nvidia trading lower on Tuesday morning.

With its Q4 earnings due out Wednesday after the close, the chip designer’s fundamentals have been a microcosm of the S&P 500 and the wider market: earnings estimates up, multiples down.

With all these crosswinds, it’s no wonder that Nvidia has struggled to generate sustained momentum so far in 2026.

The Street’s view

Wall Street analysts, for their part, mostly believe that Nvidia will be able to convince investors that these apparent crosscurrents are actually a wind at its back.

Analysts are looking for adjusted earnings per share of $1.53 on sales of a little more than $65.9 billion in Q4.

“Advanced wafer supply, CoWoS, and DRAM allocation have become points of constraint for server builds, but we believe NVDA has largely set its supply for Grace Blackwell and has better positioning vs. peers to work around bottlenecks further ensuring NVDA continues to hold its dominant share position through 2026,” wrote Wedbush Securities analyst Dan Ives.

However, some margin pressure may be in the offing as Nvidia deploys new generations of its GPUs. And, in the coming quarters, it may be difficult to distinguish whether any headwinds to profitability are functions of the Vera Rubin ramp, higher input prices, or some mix of the two.

JPMorgan analyst Harlan Sur expects Jensen Huang and co. to indicate that gross margins will be in the mid-70s in the near term, while noting that, in light of the above factors, confidence surrounding this “remains an open question.”

He also thinks the company will aim to reassure investors that its inference capabilities are robust, countering concerns that custom chips will pose an escalating threat to its dominant market position. To this end, near the end of Q4, Nvidia reached a licensing deal (effectively an acquisition) of AI inference specialist Groq. Sur wrote:

“A broader, more overarching theme that we think has weighed on the stock is the perception of share loss relative to AI ASICs/XPUs, as the aggregate mix of AI workloads rapidly shifts more towards inference (where specialized/custom silicon can be especially beneficial) and away from training (where NVDA is the undisputed leader).”

Continuing, the JPMorgan analyst added:

“On this front, we expect management to emphasize significant gen-on-gen gains in inference performance (as demonstrated by recent third-party benchmarking), and at least lift the veil slightly on products currently in the pipeline that leverage Groq IP for specialized, low-latency inference at scale.”

Why so cheap?

The colossal, far bigger-than-expected capex budgets put forward by hyperscalers are, in a very real sense, Nvidia’s earnings guidance: chips are the biggest line item for data centers.

Why hasn’t Nvidia benefited meaningfully from these investment plans?

The reasons, in my eyes, are twofold.

First, there are more intense AI shortages that commanded investor attention. The obvious example is Sandisk, the best-performing member of the S&P 500 with a 181% year-to-date return (and indeed the best performer of last year). The flash drive seller’s 12-month forward price-to-earnings ratio has gone down during this rally — that is, the shares have become cheaper because of just how much forward earnings estimates have risen.

Second, 2026 investment plans from Nvidia’s biggest customers are great news for the chip designer’s 2026 earnings outlook. But the performance of those tech giants in the stock market is a signal.

They say money goes where it’s treated best. If investors are taking money out of hyperscalers because those companies are pouring it into AI capex with an uncertain return, well, at some point, those executives are also going to do something else with their money in a bid to engineer a better outcome in the stock market.