Nvidia Rejects Jensen’s Comeback—Then Drops a $130B Guidance Bomb
Five hours after its ousted founder publicly pleaded for his job back, Nvidia’s board shut the door. Then the AI chipmaker dropped a guidance bomb that left Wall Street speechless. $NVIDIA(NVDA)$
Colette Kress’s voice was trembling. She stumbled over words, mispronounced JPMorgan, and wrapped up Nvidia’s earnings call ten minutes early—while CUDA creator Ian Buck sat beside her without uttering a single syllable.
For a company that has spent the past three years running victory laps around Wall Street, Tuesday evening’s conference call felt remarkably like a hostage video.
And yet, the numbers Kress was reading off the teleprompter were nothing short of violent: $115.8 billion in quarterly revenue, up 103% year-over-year. $130 billion guided for Q4. Full-year FY28 top-line growth hiked to 75%.
Welcome to the post-Jensen Huang era—where the theater is gone, the executive suite is visibly rattled, and the silicon printing press hasn’t missed a single beat.
Just five hours before Kress took the mic, Huang—who resigned last week over personal conduct violations—broke his silence on X. In a contrite, 32-million-view post, Silicon Valley’s most famous leather-jacketed showman made an unmistakable play for a comeback: "If the Board determines that my return is in the best interest of our shareholders, employees, the company, and our country, I would be honored to answer that call."
It took inaugural Chairman Harvey Jones less than sixty seconds on the call to bury that dream. The board, Jones stated flatly, "does not intend to reinstate Jensen as CEO," confirming that executive search firm Heidrick & Struggles is already searching for a permanent successor.
Wall Street’s reaction? Shares popped more than 5% after hours.
Investors looked past the awkwardness in Santa Clara and priced in a far more profound reality: Nvidia has outgrown its founder. Looking across the company’s past three annual filings and Tuesday’s Q3 print, it is clear that Nvidia is no longer an asset-light chip designer reliant on one man’s charisma. It has quietly morphed into a self-sustaining, $518 billion industrial and financial juggernaut—one that locks down global silicon supply, underwrites its own customers, and runs on institutional inertia.
1. Crossing the $100B Threshold: Rubin’s Record Ramp and the Jevons Paradox
On pure fundamentals, Nvidia left the bears with almost nothing to attack.
Third-quarter revenue of $115.84 billion blew past both internal guidance ($108 billion) and the consensus estimate ($109 billion), adding nearly $20 billion sequentially. Stripping out investment gains, Non-GAAP net income hit $64.87 billion(+109% YoY), translating to an eye-watering 67.2% operating margin—meaning more than 67 cents of every revenue dollar dropped straight to the operating line.
Driving that velocity is a seamless architectural handoff.
Heading into the print, skeptics worried the transition from Blackwell Ultra to next-generation silicon would create an air pocket in demand. Instead, Vera Rubin NVL72 became the fastest-ramping product in company history, generating nearly $20 billion in incremental sequential revenue in its debut shipping quarter.
Delivering 10x the tokens per watt of GB200, Rubin cut customer deployment timelines in half to roughly one month. Meanwhile, standalone Vera CPUs are gaining rapid traction across CoreWeave, Meta, and OpenAI, with management projecting CPU and enterprise revenues to more than double in FY28.
More importantly, management armed bulls with granular ecosystem metrics that directly dismantled the "AI capex bubble" narrative:
-
The Jevons Paradox in Action: Since May, LLM token inference prices have plunged nearly 60%. Rather than curbing hardware spend, cheaper compute triggered an explosion in agentic workloads. Weekly token volume on aggregator OpenRouter surged 35x year-over-year (and 3.5x over the past three months), with agentic traffic accounting for 80% of usage.
-
Old Silicon Commands Premiums: Compute isn't just sold out—it’s appreciating. Rental rates for B300 clusters have climbed more than 60% since launch, while multi-year renewal rates for prior-generation Hopper GPUs are clearing at up to 2x their original contract prices.
-
Backlogs Keep Ballooning: Cloud service provider (CSP) backlogs jumped from $2 trillion to $2.5 trillion in just three months. Frontier model labs—which account for roughly 15% of Nvidia’s data center revenue this year—are proving their commercial viability, led by Anthropic hitting an $81 billion annualized revenue run-rate (ARR) at the end of October.
Within the $107.61 billion Data Center segment (92.9% of total sales), customer diversification is accelerating. While traditional Hyperscalers grew 71% YoY to $60.14 billion, the AI Clouds, Industrial, & Enterprise (ACIE) cohort skyrocketed 195% YoY to $47.47 billion—now representing 44.1% of data center revenue, propelled by sovereign AI (up >3x YoY) and physical AI (tracking >$25 billion for the full year).
Looking ahead, Q4 revenue guidance of $130 billion (±3%) benefits from a 14-week fiscal calendar (one extra week), yet achieves that mark while assuming zero China data center compute revenue (following the halt of Hopper shipments to China, which totaled $2.4 billion in Q3). Raising FY28 revenue growth expectations from 70% to 75% off a much larger base sends a blunt message: demand remains constrained only by power, shells, and supply.
2. No Longer Asset-Light: The Quiet Rewiring of Nvidia’s Balance Sheet
For over a decade, Wall Street prized Nvidia as the ultimate asset-light, fabless compounder—designing high-margin silicon while leaving heavy capex to foundries. Over the past three fiscal years, however, Nvidia’s DNA has quietly shifted.
First, Nvidia has evolved from a chip vendor into a turnkey "AI Factory" builder.
In FY24 ($60.9 billion total revenue), Nvidia primarily sold HGX boards and discrete GPUs. Today, through what it calls "Extreme Co-design," Nvidia ships full rack-scale NVL72 systems integrating 72 GPUs, 36 CPUs, NVLink switches, and liquid cooling (pushing quarterly networking revenue past $21 billion).
Selling full systems multiplies revenue per rack, but structurally resets gross margins lower due to higher pass-through component costs. Non-GAAP gross margin landed at 74.9% in Q3 and is guided to trough at 71.0%–72.0% in Q4 as Nvidia fulfills legacy-priced orders while absorbing surging upstream memory costs, before rebounding to 72%–73% in FY28 under new pricing. Nvidia is willingly trading 200 to 300 basis points of margin for end-to-end architectural lock-in.
Second, an annual product cadence is starving custom ASICs of oxygen. By compressing its data center roadmap to a strict one-year cycle—Blackwell, Blackwell Ultra, Vera Rubin, Rubin Ultra next year, and Feynman in 2028—Nvidia is forcing hyperscalers’ internal ASIC teams (Google’s TPU, Amazon’s Trainium) into a brutal treadmill. By the time a custom chip clears tape-out and software integration over a two-year cycle, Nvidia has already moved the goalposts on cost per token.
Third, Nvidia is deploying its balance sheet as a "shadow bank" to lock down the supply chain.
To secure scarce CoWoS packaging and HBM memory through the end of the decade, Nvidia’s off-balance-sheet supply and capacity commitments surged from $279 billion to $390 billion in Q3 ($203 billion due in FY28 alone). Factoring in cloud agreements and uncommenced data center leases, total commitments now sit at $518 billion.
Simultaneously, Nvidia is funding its own ecosystem: non-marketable equity investments nearly tripled from the start of the year to $65.5 billion (alongside $44.2 billion in public equities and a pending deal to acquire Hugging Face).
3. The $118 Billion Guarantee Question: How Nvidia Built a Firewall
The sharpest analyst pushback on the call centered on vendor financing and credit guarantees. Total off-balance-sheet guarantees reached $118 billion at quarter-end—anchored by $105 billion in credit support for SoftBank-backed SB Energy’s PORTS-Pike campus in Ohio, which will house 4.25 GW of Nvidia infrastructure under 20-year leases to OpenAI starting in FY29.
The bear case is straightforward: if AI monetization stalls or leveraged neoclouds hit a financing wall, is Nvidia on the hook for over $100 billion in stranded liabilities?
Kress used her commentary and Q&A to draw a hard firewall around that risk. She clarified that Nvidia’s guarantees are strictly contingent, not unconditional backstops. Credit support only kicks in phase-by-phase after the partner completes site construction, secures independent debt or equity financing, and installs and operationalizes the compute clusters. If a project stalls or financing falls through, Nvidia’s commitment automatically shrinks or evaporates.
Meanwhile, the cash machine continues to hum. Nvidia generated $48.25 billion in operating cash flow and $43.78 billion in free cash flow in Q3, returning a record $28.1 billion via buybacks and dividends while ending the quarter with $53.5 billion in cash and marketable securities. Management reiterated its plan to exhaust its remaining $227.3 billionbuyback authorization by the end of FY28.
Still, two working capital lines warrant close monitoring: accounts receivable jumped to $81.03 billion (pushing DSO up four days sequentially to 64 days due to back-end-loaded Rubin shipments and extended terms for investment-grade buyers), and inventories climbed to $39.72 billion ahead of the $130 billion Q4 ramp. Healthy in an upcycle, both metrics will be the first to flash yellow if demand ever cools.
4. Less Theater, Same Machine
A conference call that ended ten minutes early revealed a company in transition.
For a decade, Jensen Huang’s leather-jacket keynotes turned Nvidia into a singular tech totem.
On Tuesday, Chairman Harvey Jones’s blunt refusal to reinstate him—paired with Kress’s understated, matter-of-fact sign-off, "We've got this"—marked Nvidia’s pivot from founder-led theater to institutional execution. With TSMC executives visiting Santa Clara this very week to lock in packaging capacity and 7.5 million developers entrenched in CUDA, Nvidia’s moat clearly extends far beyond one charismatic leader.
For whoever Heidrick & Struggles taps as the next permanent CEO, the real test lies ahead over the next two quarters:
1.The Margin Inflection: Proving that Q4’s 71%–72% gross margin is truly the cycle trough, and executing the Q1 FY28 pricing reset back toward 72%–73%.
2.Neocloud Credit Health: Watching whether private capital markets keep funding AI labs and neoclouds at scale—the ultimate safeguard for Nvidia’s $81 billion receivables balance and $118 billion guarantee book.
3.Geopolitical Tightrope: Navigating tightening export controls (with China data center revenue now effectively zeroed out) through the newly formed National Security and Export Control Oversight Committee.
Without Huang on stage, Nvidia has lost a touch of its Silicon Valley romanticism. Judging by the cold precision of its Q3 print, however, the empire’s foundation remains unshaken.
What’s your take on a post-Jensen Nvidia? Is the board’s rejection a sign of institutional maturity, or will Nvidia miss its visionary founder in the next chip war?
With a $130B Q4 guide on the table, are you holding, buying, or taking profits? Drop your thoughts below!
Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.
- MoiraHorace·11-18 16:441.34x historical volatility prices a bigger move, not institutional buying. I'd lean bullish on Nvidia if that guidance checks outLikeReport
