Nvidia Faces Judgment Day: Will Earnings Ignite a New Supercycle or Trigger a Meltdown?

Nvidia’s third-quarter fiscal 2026 report on Wednesday, November 19 after the close is shaping up as the single most important data point for markets over the final six weeks of the year. The stock just came off an all-time high near $212 on October 29 that briefly pushed its market cap to $5 trillion, but has since pulled back to sit almost exactly on its 50-day moving average – a level that now doubles as either launchpad or trap door depending on how the print lands. With Nvidia carrying roughly a 10% weight in the Nasdaq 100 and viewed as the purest proxy for the global AI infrastructure build-out, this report will be read not just as a company update, but as a referendum on the durability of the entire AI trade. Expectations are high , sentiment is edgy, and even well-timed profit-taking by big holders like Peter Thiel and SoftBank is feeding the “have we come too far, too fast?” debate.

Wall Street is looking for about $1.25 in EPS on roughly $54.8–55 billion in revenue, which would still represent more than 50% year-on-year growth for both the top and bottom line. Nvidia’s own guidance last quarter called for $54 billion in revenue (±2%) and non-GAAP gross margin around 73.5%, and several bullish houses are openly modeling upside: estimates in the $56–57 billion range for the October quarter and $62–64 billion for January are common on the street, with some whisper numbers creeping higher. Guidance will be the real swing factor. Consensus sits near $61–62 billion for the January quarter; anything meaningfully above that, paired with gross margin commentary that keeps the mid-70s narrative intact despite higher memory and component costs, would go a long way toward re-energizing the stock’s march back toward – and beyond – the $5 trillion mark.

Under the hood, investors will again focus on the data center segment, which delivered $41.1 billion of last quarter’s $46.7 billion in total revenue, up 56% year-on-year. That print included a 17% sequential increase in Blackwell data center revenue and record $7.3 billion in networking, which nearly doubled from a year earlier on demand for Spectrum-X Ethernet, InfiniBand, and NVLink. Management also posted records in gaming ($4.3 billion, +49% y/y) and automotive ($586 million, +69% y/y), but for this week’s report the market will mostly care about three lines: total data center revenue, networking, and any early contribution from the newer Blackwell Ultra platform. Analysts have flagged 73.3–73.5% gross margin and rising operating expenses as Nvidia accelerates investments; an upside surprise in revenue with stable margins would reinforce the thesis that scale is offsetting cost inflation.

The bigger story is the forward curve. At Nvidia’s Washington, D.C. GTC conference, CEO Jensen Huang stunned even the bulls by disclosing more than $500 billion in cumulative orders for Blackwell and next-generation Rubin through the end of 2026 – part of what he framed as a $3–4 trillion AI infrastructure spending opportunity this decade. He outlined plans to ship roughly 20 million Blackwell/Rubin GPUs across 2025–26, with Blackwell already in full production in Arizona and Rubin slated for 2H26. Evercore, UBS, Bank of America and others have lifted their 2026–27 data center and EPS estimates on the back of that backlog, with some now modeling mid-to-high-single-digit EPS in 2026 and talking openly about upside if Nvidia and TSMC can unlock more CoWoS packaging and HBM capacity. Wednesday’s call is the first real chance for investors to pressure-test those numbers: look for detailed commentary on units, rack-scale NVL72 deployments, and how quickly Blackwell-based GB300 systems are ramping versus the prior Hopper generation.

Read-throughs from the rest of the ecosystem mostly support the bulls. Hyperscalers – Amazon, Microsoft, Alphabet, Meta – all raised AI-related capex in recent earnings, with some estimates now calling for a nearly 70% year-on-year jump in 2025 spend and further growth of approximately 25% in 2026. Foxconn has already hit its original 2025 AI server revenue target in three quarters. On the customer side, deals keep piling up: Nvidia is backing a $500 million GB300-powered data center in Taiwan for GMI Cloud; it is partnering with HPE on Vera Rubin-based supercomputers for Los Alamos and the U.S. Department of Energy; and new arrangements span everyone from Eli Lilly and CrowdStrike to Nokia, Samsung, Hyundai, Uber and Palantir. The headline OpenAI equity commitment and GPU supply pact, plus investments in Intel integration and various sovereign AI initiatives, underscore Nvidia’s push to entrench its full-stack platform – chips, networking, software, and services – at the center of virtually every high-end AI deployment.

China remains the biggest wild card. Last quarter Nvidia booked essentially no H20 data center revenue there due to export controls, and current consensus and most analyst models assume zero China data center sales going forward. Management has talked about China as a potential $50 billion annual opportunity in a friendlier regime, but Huang has also stressed there are no active plans to sell Blackwell into the country under existing rules. Any hint that Washington may green-light limited Blackwell exports, or that Nvidia has found a compliant configuration that meaningfully re-opens the market, would be a major upside surprise; conversely, fresh restrictions – or signs that local players like Huawei are eroding Nvidia’s edge faster than expected – would fuel the bear case that geopolitics could cap the TAM.

Against that backdrop, the recent selling by Peter Thiel’s Thiel Macro fund and SoftBank is less about fundamentals and more about psychology. Thiel’s exit from roughly 538,000 shares and SoftBank’s sale of a multi-billion-dollar stake are reminders that some sophisticated holders are willing to take chips off the table after a historic run, especially while they fund bets on would-be Nvidia challengers and adjacent AI start-ups. If Nvidia delivers another big beat-and-raise, those sales may be remembered as routine profit-taking; if the company merely meets consensus or guides cautiously, they will be cited as canaries in the coal mine.

For the stock and the broader market, then, the setup is binary. A strong print – think revenue in the high-$50s or better, January guidance well north of $62 billion, reaffirmed mid-70s gross margin trajectory, and continued evidence of “insatiable” AI capex – would likely turn the 50-day moving average into a springboard and re-ignite momentum across the AI complex, with knock-on effects for the Nasdaq 100 and risk assets globally. A miss on the top line, a conservative guide, or any suggestion that hyperscaler spending is slowing or that power and supply constraints are biting harder than expected could flip that level into a trap door, validating AI-bubble worries and inviting a deeper rotation out of growth. For traders and long-term investors alike, Wednesday night is not just another earnings print – it is the next vote on whether the AI super-cycle is peaking or just getting its second wind.