Nvidia reported $115.8 billion of revenue for the quarter ended October 25, up 103% from a year ago and 20% sequentially, against a consensus near $109 billion. Non-GAAP EPS was $2.68 against $2.48. Data center did $107.6 billion. Vera Rubin added close to $20 billion in its first quarter of shipments, which the company calls the fastest ramp in its history. Then came the guide: $130 billion, plus or minus 3%, against $125.1 billion expected. On the headline that is 12% sequential growth and a 3.9% beat on the outlook.
The fourth quarter of fiscal 2027 has fourteen weeks. Q3 had thirteen. Divide through and Q3 ran at $8.91 billion a week, while the Q4 midpoint runs at $9.29 billion a week. That is 4.2% growth in weekly run rate, after three straight quarters of 18% to 20% (Q1 +20%, Q2 +18%, Q3 +20%). The number that matters is the 4.2%, and nobody led with it last night because Colette Kress, five days into the interim CEO job after Jensen Huang’s abrupt resignation on November 12, had bigger questions to answer and answered them.
The margin side makes it sharper. Q4 gross margin is guided to 71% to 72%, down from 74.8%. Kress called Q4 the bottom, tied it to honoring existing purchase orders as Rubin ramps, and held fiscal 2028 at 72% to 73%. Fine. But run the incremental math. Q3 gross profit was about $86.6 billion. The Q4 midpoint (130 at 71.5%) is about $93.0 billion. That is $6.3 billion of extra gross profit on $14.2 billion of extra revenue, an incremental margin of roughly 44% for a company that prints 75% on average. Non-GAAP opex steps up to about $10.7 billion from $8.8 billion, carrying the extra week plus transition costs. Non-GAAP operating income per week falls, from about $5.99 billion in Q3 to about $5.88 billion at the Q4 midpoint. Headline operating income still grows. Weekly operating income shrinks by roughly 2%.
None of that makes the quarter weak. It makes the guide conservative in a specific way. Nvidia beat its own Q3 guide of $108 billion by $7.8 billion, or 7.2%. Apply the same overshoot to Q4 and revenue lands near $139 billion, about $9.96 billion a week, which is 11.8% weekly growth. So the real test of the post-Huang Nvidia is whether the beat-over-guide habit survives a CFO who is now also the CEO and has every reason to under-promise. Kress said on the call the company would be “rigorously evaluating future investments, commitments, and guarantees” and described the neocloud stance as disciplined and selective. A finance chief running the company guides like a finance chief.
That line about guarantees points at the balance sheet, which is where the quarter is less tidy than the income statement. Receivables are $81.0 billion against $38.5 billion in January. On quarterly revenue that is roughly 64 days of sales outstanding versus about 51 days at fiscal year-end, so customers are taking almost two weeks longer to pay. Inventory is $39.7 billion against $21.4 billion. Long-term debt went from $7.5 billion to $32.4 billion. And the carrying value of equity stakes (non-marketable securities of $65.5 billion plus marketable equity of $44.2 billion) rose to $109.7 billion from $35.2 billion, an increase of $74.5 billion in nine months. Part of that is mark-ups on stakes in labs and clouds that have repriced, so it is not all cash out the door, but it sits against nine-month free cash flow of $113.7 billion. A meaningful share of what Nvidia earns is being recycled into the customer base that buys its systems. We argued in an earlier piece on AI financing that the vendor deals sit on real revenue, and the Q3 revenue line supports that. The DSO drift is the part to watch.
The moat question got a cleaner answer than the governance question. Rubin shipping on schedule and generating about $20 billion in its first quarter, with Amkor, IBIDEN, SPIL, Unimicron and the rest of the packaging chain scaling behind it, shows the advantage lives in supply-chain command and rack-scale integration (NVLink, the CUDA stack, the networking attach) rather than in one person on a keynote stage. Two smaller items in the release widen it. MediaTek joining NVLink Fusion means Nvidia’s fabric gets licensed into silicon it does not sell. The agreement to acquire Hugging Face puts Nvidia at the default distribution point for open models, which is the layer where a custom ASIC from Broadcom or Marvell would otherwise try to win inference workloads on price. Hyperscaler in-house silicon remains the real competitive threat, and the 561 billion dollar rotation into AMD and Broadcom started before Huang quit. Q3 does not reverse that rotation. It shows it hasn’t shown up in Nvidia’s order book yet.
China is a clean zero in the outlook. Revenue from customers headquartered in China, Hong Kong included, was $11.1 billion in Q3, with $2.4 billion of Hopper data center shipments, and the Q4 guide assumes no China data center compute at all. Given the export control scrutiny around Singapore sales, that conservatism is the right call and leaves a small upside lever.
The stock closed Monday at $156.14, down about 24% in November and roughly 30% below its October record, and traded around $163 after the call (about 5% higher). That is a market cap near $3.9 trillion. The fiscal 2028 outlook was raised to about 75% growth from 70%, which on a fiscal 2027 base of roughly $424 billion implies around $741 billion of revenue, about 7% above the $692 billion FactSet consensus going in. Hold gross margin in the guided 72% to 73% band, let opex grow faster than this quarter’s, tax at 17%, and FY28 EPS comes out in the $16 to $17 range. At $163 that is under 10 times next year’s earnings for the company that just grew 103%. Street targets run from $130 to $515, averaging somewhere between $232 and $265 depending on the tally, after a round of pre-print cuts (Mizuho to $220 from $250, GF Securities to Hold at $188 from $235). Board chair Harvey Jones said the board does not intend to reinstate Huang, and an independent board committee investigation is still open, with no detail given. The discount is a governance discount and a financing discount. It is not an earnings discount.
Base case is $175 to $200 over the next quarter, as the leadership overhang fades without being resolved and the multiple claws back part of November’s drop. The bull case is $220 to $240, which needs a Q4 that beats the guide by something close to Q3’s 7%, a permanent CEO named, and the board committee closed without a finding that touches the financials. The bear case is a cohort derating to $125 to $140: the weekly deceleration shows up in actual Q4 numbers, DSO keeps stretching, and AI semis (AMD, Broadcom, Marvell) get marked down together as vendor-financed revenue gets a lower multiple. Our preview framed the bar at $109 billion and a margin reset. She cleared the first and confirmed the second. For the wider succession picture see the CEO search and key person risk, and for a year-on-year read, last year’s Q3 at $57 billion, which this quarter doubled.
One number decides which case plays out. When Nvidia reports in February, divide Q4 revenue by fourteen. Above $9.6 billion a week, the beat habit survived the founder and the 4.2% was Kress being careful. At or below $9.3 billion, the guide was the truth, and the 20% sequential era ended in the same quarter as the Huang era.