HBS CEO Summit Case Study

Jensen Huang
& The Relentless Rise of Nvidia

From a Denny's booth in 1993 to the world's most valuable company. This is not a story about luck. It's a playbook for market creation under constraint.

Case Number N2-326-056
Session Date February 4, 2026
Instructor Hubert Joly
Location Tata Hall 100
Scroll
Executive Summary

Nvidia's rise from niche gaming chip supplier to $4.4 trillion company was not luck. It was the result of deliberate market creation.

In 1993, three engineers met in a San Jose Denny's to plan a graphics chip company. Thirty-two years later, Jensen Huang remains CEO of what became the world's most valuable company—a 591,000% stock return since IPO.

The case is not about semiconductors. It's about a founder who repeatedly bet the company on markets that did not yet exist—while managing near-death moments, activist pressure, and a culture built for speed. When AI arrived, Nvidia had spent fifteen years preparing for a market no one else believed in.

The central question: What is your CUDA-equivalent bet—the thing that looks like a "zero-billion-dollar market" today, but you might be judged on in 10-15 years?

Breaking — January 12, 2026

To the Roundtable Members

The announcement today regarding the Nvidia and Eli Lilly AI Co-Innovation Lab represents a definitive shift in the case study's narrative. It moves Nvidia from a hardware vendor to a strategic partner in vertical integration.

For this specific audience, the significance lies not in the pharmaceutical details, but in the operational and business model precedents being set.

1

The "AI Factory" Business Model

Relevant: Scott Vertrees (Investment), Christian Frigast (Governance)

Nvidia is no longer just selling chips; it is co-investing capital and talent to build proprietary infrastructure.

The Deal: Nvidia and Lilly are jointly investing up to $1 billion over five years to build a lab in the San Francisco Bay Area. This facility will co-locate Lilly scientists with Nvidia engineers.

Strategic Implication: This validates the "AI Factory" concept—not as a metaphor, but as an asset class. For Scott Vertrees, this signals a shift in CapEx strategy where tech infrastructure becomes a joint venture rather than vendor procurement, potentially altering how value is attributed in M&A or turnarounds.

2

Validation of "Agentic" and "Physical" AI

Relevant: Frédéric Guinot (Industrial), Richard Haykel (Healthcare), John Kim (Retail)

The announcement explicitly operationalizes "Agentic AI"—systems that reason and take action—in a physical setting.

Agentic Wet Labs: The collaboration focuses on connecting "agentic wet labs" with computational "dry labs". AI agents will drive robotic experimentation 24/7 without human intervention, creating a closed-loop continuous learning system.
For Manufacturing (Guinot): This is the proof-of-concept for the "lights-out" factory. If an AI agent can manage complex chemical biology experiments autonomously, the leap to autonomous forging or casting lines is significantly shorter.
For Retail (Kim): The project utilizes Nvidia NiM microservices and AI Blueprints for building agents. This confirms that the software infrastructure for autonomous customer service agents is now enterprise-ready and being deployed in high-stakes environments.
3

Supply Chain Digital Twins (Beyond Biology)

Relevant: Robert Fleishman (Logistics/Conglomerate), Todd Sachse (Construction)

Crucially, this partnership extends beyond drug discovery into manufacturing and supply chain operations.

The Digital Twin: Lilly is using Nvidia Omniverse and RTX PRO Servers to create digital twins of its manufacturing lines.

Operational Utility: They are modeling and stress-testing the entire supply chain to "enhance capacity to manufacture high-demand medications" and reduce downtime. For Robert Fleishman, this is a direct template for how GT Global could simulate automotive logistics or plastic injection molding processes to predict bottlenecks before they occur physically.

4

Technical Validation of the "Rubin" Platform

Relevant: All Members

This is the first major commercial confirmation of the Vera Rubin architecture discussed in your briefing.

The Tech Stack: The new lab and Lilly's AI factory will specifically harness the Nvidia Vera Rubin architecture.

Why it Matters: This confirms that the Rubin platform (and its promise of 10x lower inference costs) is not just a roadmap item but is being actively adopted by Fortune 500 cap-ex heavyweights. This de-risks the technology for other sectors represented at the table.

5

Implications for Healthcare Delivery

Relevant: Richard Haykel (Hospital Leadership)

For Haykel Hospital, this signals a future change in clinical trials and drug availability.

Generative Clinical Development: The partnership aims to use generative AI to accelerate clinical trials and automate medical writing, drastically shortening the time from lab to patient.

TuneLab: Lilly is launching TuneLab, a platform allowing biotech companies (and potentially research hospitals) to access proprietary AI models trained on Lilly's data. This suggests a future where smaller players might access "Big Pharma" AI capabilities via a cloud service to support local research or personalized medicine.

Summary for the Room

This announcement proves that the "Industrial Revolution" narrative Jensen Huang pushes is taking root. It demonstrates that the technology has matured from "chatbots" (generative text) to "doing work" (agentic research and physical manufacturing). For the investors and operators in the room, the takeaway is that AI is moving from an IT expense line to a core operational asset that drives the physical production of goods.

Intelligence Briefing

To the Members of the HBS CEO Summit Roundtable

The following intelligence briefing synthesizes recent developments regarding Nvidia's strategic roadmap, specifically the launch of the Rubin platform and the rise of Sovereign AI, tailored to your specific operational contexts—ranging from heavy industry and healthcare to private equity and consumer retail.

1

Operational Efficiency & "Physical AI"

For: Guinot, Sachse, Fleishman — Manufacturing, Construction, and Logistics Leaders

The most critical development for operators of physical assets is Nvidia's pivot toward "Physical AI" and "Agentic AI"—systems designed to reason, plan, and control physical machinery rather than just generate text.

The Rubin Platform: Announced at CES 2026, the Rubin platform is in full production. Crucially for industrial scaling, it promises a 10x reduction in inference token cost compared to the previous Blackwell generation.
Why It Matters: This drastic cost reduction changes the unit economics of deploying AI for predictive maintenance and digital twins. Nvidia has explicitly partnered with industrial giants like Siemens, Caterpillar, and Eaton to integrate these chips into "AI factories".
Infrastructure Shift: The new architecture is designed to overcome the "energy wall" that currently constrains data centers. For Farinia Group and GT Global, this supports the deployment of "digital twins" that model power, cooling, and manufacturing flows to optimize energy efficiency dynamically.
Robotics: For construction and logistics, Nvidia has launched Alpamayo, an open reasoning model specifically for autonomous machines and Level 4 autonomous driving, which is already being integrated by automotive partners like Mercedes-Benz.
2

Sovereign AI & Infrastructure in Constrained Markets

For: Haykel, Frigast — Healthcare Leadership and Governance/Stewardship

A major shift in 2025–2026 is the decentralization of AI infrastructure via "Sovereign AI," where nations build domestic compute capacity to ensure data residency and cultural independence.

Data Residency & Security: For Haykel Hospital operating in Lebanon, the rise of Sovereign AI offers a potential roadmap for accessing advanced diagnostics without data leaving national or regional borders. Nvidia is actively building sovereign clouds in India (Shakti Cloud), Saudi Arabia, and Japan to process local data locally.
The "Nuclear" Analogy: CEO Jensen Huang has framed this capability as a national imperative, stating, "No one needs atomic bombs, everyone needs AI," positioning domestic compute infrastructure as vital for national security and economic independence.
Cost & Access: The partnership with Yotta in India specifically aims to "democratize access" for startups and researchers using a pay-per-use model, a structure that may eventually trickle down to healthcare systems in other developing markets.
3

ROI, "Real vs. Theatre," and Turnarounds

For: Vertrees, Kim — Investment, QSR, and Digital Marketing

For leaders focused on unit economics and immediate returns, the narrative has shifted from "training" (building models) to "inference" (running them efficiently).

The Efficiency Leap: The Rubin platform's claim of 4x fewer GPUs required for training and significantly lower energy consumption directly addresses the "profitless prosperity" concern where AI costs outweigh revenue.
Agentic AI for Retail: For Baskin-Robbins Japan, the new "Inference Context Memory Storage" platform allows for massive context processing. This enables "Agentic AI" capable of multi-step reasoning—moving beyond simple chatbots to agents that can handle complex customer service or supply chain scheduling tasks with high interactivity and low latency.
Financial Reality Check: Despite the hype, Nvidia faces antitrust scrutiny from the DOJ regarding "tying" arrangements and customer lock-in. However, the company continues to generate massive cash flow, returning $37 billion to shareholders in the first nine months of FY2026 alone. This suggests the AI boom is converting to actual cash, not just valuation spikes.
4

Geopolitics & Supply Chain Risk

For: Fleishman, Frigast, Guinot — Conglomerates and Governance

The supply chain landscape is bifurcating, introducing new risks and opportunities for diversified groups.

Diversification into Intel: In a strategic move to de-risk its reliance on TSMC (Taiwan), Nvidia has finalized a $5 billion investment in Intel, taking a 4% stake. This aligns with a strategy to utilize Intel's U.S.-based foundries (Arizona) for future packaging and production, potentially stabilizing the supply chain for Western markets.
China Strategy Shift: The U.S. government has approved the export of Nvidia's powerful H200 chips to China, but with a transactional twist: the U.S. government will take a 25% cut of sales revenue. However, Nvidia is now demanding full upfront payment from Chinese clients to hedge against sudden policy reversals, creating significant friction for buyers in that region.
Capacity Bottlenecks: Despite these moves, TSMC's advanced packaging (CoWoS) remains the ultimate bottleneck. Nvidia has booked over 50% of TSMC's capacity through 2027, potentially squeezing out competitors and smaller industrial players trying to secure custom silicon.

Summary of Strategic Implications for the Table

Manufacturing/Construction

The technology is moving from "generative" (text/images) to "physical" (robots/twins). The Rubin platform makes these simulations economically viable for the first time.

Healthcare/Regional Ops

"Sovereign AI" infrastructure is being built out globally (Saudi Arabia, India, Japan). Localized, secure compute capacity may soon be available even in complex geopolitical environments.

Capital Allocation

While the DOJ antitrust probe introduces regulatory risk, Nvidia's move to diversify manufacturing to Intel and its focus on lowering inference costs suggests a maturation from experimental capex to operational opex.

$4.4T
Market Cap
World's most valuable company (Aug 2025)
591,000%
Stock Growth
Since IPO in 1999
~80%
AI Chip Share
Dominant market position
32 yrs
Jensen as CEO
From founding to present
01

The Genesis: Forged Through Failure

The Crucible Years

Jensen Huang was born in Taiwan in 1963. His parents sent him to what they believed was a prestigious American boarding school. Oneida Baptist Institute in Kentucky was actually a reformatory. There, he cleaned bathrooms, endured bullying, and developed the resilience that would define his leadership.

A nationally ranked table tennis player, he learned to treat losses as data. He worked at Denny's through high school, teaching himself to take pride in difficult work. "I worked hard! Like, really hard. So I got to be a busboy."

I wish upon you ample doses of pain and suffering.
Jensen Huang's unusual benediction to advice seekers

NV1: The Expensive Education

Nvidia's first product was a technical showcase and a commercial disaster. The NV1 used an unconventional graphics model that required developers to reprogram their games. Doom—the most popular game of the era—turned sluggish and silent on NV1 machines.

The NV2 partnership with Sega collapsed. Huang laid off more than sixty of a hundred employees. The company had six months of runway. Near-death.

RIVA 128: The Reset

With survival on the line, Huang instituted reforms that became permanent operating principles:

Tandem Development: Hardware and drivers built in parallel, shortening cycles by up to a year.

Strategic Investment: Spent $1M (one-third of remaining cash) on a chip emulator to catch design bugs before silicon.

Market Alignment: Abandoned proprietary standards, embraced Microsoft's Direct3D.

The RIVA 128 shipped on time and sold 4 million units. The company survived. The reforms became culture.

The Nvidia Journey

Three decades of near-death experiences, strategic bets, and market creation

1993

Founding at Denny's

Jensen Huang, Curtis Priem, and Chris Malachowsky meet in a San Jose Denny's booth to plan a graphics chip company. Sequoia Capital backs them on the strength of Huang's reputation for hard work.

1995-96

NV1/NV2 Failures

First products are technical showcases but commercial disasters. Sega partnership collapses. 60+ employees laid off. Six months of runway remaining.

1997

RIVA 128: The Reset

Operational reforms. Tandem development. Emulator investment. Direct3D alignment. 4 million units sold. Company survives.

1999

IPO & "GPU" Coined

GeForce 256 launches as the world's first "GPU"—a term Nvidia invented to elevate the category. Stock tops $100 after Xbox deal announcement.

2002-03

NV30 "Leaf Blower"

GeForce FX 5800 Ultra requires a massive, loud fan. Industry punchline. Huang diagnoses root cause as organizational—silos had formed. Reinforces flat structure mandate.

2007

CUDA Launch

$475M investment over four years—one-third of R&D budget—on an unproven use case. Stock drops 80%. Analysts call it "willful destruction of value." Huang calls it a "zero-billion-dollar market."

2012

AlexNet Breakthrough

Deep learning model using just two Nvidia GPUs trounces competition. Google needed 16,000 computers for similar tasks. CUDA bet begins paying off.

2020

Mellanox Acquisition

$6.9B acquisition of Israeli networking firm. Provides high-speed interconnects for linking thousands of GPUs. Annualized revenue eventually exceeds $12B.

2022

Arm Acquisition Fails

$40B deal scuttled by FTC and Chinese regulators. Nvidia forfeits $1.25B prepayment. Would have created unprecedented computing powerhouse.

2022-25

AI Explosion

ChatGPT launches. H100 GPU is exactly what market needs. Blackwell B200 cements dominance. Nvidia becomes world's most valuable company at $4.4T.

02

The CUDA Gambit: Creating Markets

In 2002, a PhD student named Mark Harris demonstrated that GPUs' parallel processing power could solve complex computational problems faster than traditional CPUs. Huang, a believer in discerning "weak signals" from industry noise, decided to fund the development of CUDA—a platform to let researchers harness GPU power.

The Cost of Vision

$475 million over four years. Roughly one-third of Nvidia's R&D budget. For a market that didn't exist.

The stock dropped 80% between October 2007 and November 2008. Activist fund Starboard Value pushed for less CUDA and more buybacks. Wall Street consensus: Huang had blundered by focusing on a distant possibility rather than next quarter's results.

Different ways of different investors saying 'You're different, and that's bad.'
Jensen Huang on Wall Street's reaction to CUDA

The Zero-Billion-Dollar Market

Huang's response to critics was counterintuitive: "Because one of the things you can definitely guarantee is where there are no customers, there are also no competitors."

By creating the market, Nvidia established CUDA as the de facto standard for scientific computing. By 2012—before AI went mainstream—almost 600 institutions were teaching CUDA, and scholars had published over 22,000 CUDA-assisted papers.

The strategic insight: CUDA created switching costs before the AI boom. Researchers trained on CUDA stayed on CUDA. When AI arrived, competitors were years behind. Nvidia had spent fifteen years building the foundation for a market that exploded overnight.

First Principles Thinking

When asked how he spots trends, Huang emphasizes going back to basics:

You can learn how something can be done and then go back to first principles and ask yourself: Given the conditions today, given my motivation, given how things have changed—how would I reinvent this whole thing?
Jensen Huang on his decision-making approach

The CUDA Bet: Then vs. Now

What Wall Street saw vs. what Huang built

Wall Street View (2007-08)

Willful Destruction

  • No commercial customers
  • 1/3 of R&D on unproven use case
  • Deteriorating margins
  • Stock down 80%
  • Activist pressure for buybacks
Reality (2025)

Market Creation

  • De facto AI infrastructure standard
  • ~80% AI chip market share
  • $130.5B revenue FY2025
  • 591,000% stock growth since IPO
  • Competitors 10+ years behind

The Jensen-isms: A Leadership Playbook

Core principles that shaped Nvidia's culture and execution

"We don't steal market share, we create markets"

Forces focus on disruptive innovation that establishes new, high-margin categories where Nvidia sets the standard—rather than fighting for incremental gains in established fields.

"Intellectual honesty"

Mandates a culture of truth-telling and rapid learning from mistakes. Allows rapid diagnosis of root causes without blame, converting each setback into a durable organizational lesson.

"Speed of Light"

Pushes the organization to operate at the absolute limit of what is physically possible, shortening development cycles and enabling rapid response to market opportunities.

"Honing the sword"

Spirited debate leads to the best ideas. Originated in "heated, even furious" technical arguments with co-founders. Ensures ideas are rigorously tested before resources are committed.

"Living between aspiration and desperation"

Maintains a sense of urgency and paranoia. "The company's always in peril and we feel it." Ensures the company never rests on its laurels.

"No one loses alone"

If you are falling behind, inform your team promptly so they can help. Culture emphasizes asking for help while discouraging office politics.

03

The Flat Organization Experiment

60 Direct Reports

Nvidia's structure is unusual for a $4 trillion company. Approximately 60 people report directly to Huang. One-on-one meetings are rare. Autonomy is expected. Information flows freely.

The principle: "The people that report to the CEO should require the least amount of pampering. They should require very little management."

Why It Works

After the NV30 "leaf blower" failure, Huang diagnosed the root cause as organizational. The scrappy, collaborative culture of early days had given way to silos. Problems obvious to consumers were never raised in meetings.

"When you're moving that fast, you want to make sure that information is flowing through the company as quickly as possible—with no barriers and no boundaries."

Love and Torture

Huang's vocabulary is not typical for a Fortune 500 CEO: "I use love fairly abundantly and care, I use abundantly." He's famous for remembering not just employees' names, but their CVs and details of their personal lives.

But candor coexists with care. "Honing the sword" can be unpleasant. One executive noted that "when he's torturing people, he's forcing them to learn a lesson."

I'd rather improve you than give up on you. It's kind of tongue in cheek, but people know that I'd rather torture them into greatness.
Jensen Huang on his approach to developing talent

The Results

Despite the demanding culture, Nvidia has unusually low turnover: 2.7% in FY 2024 and 2.5% in FY 2025. Compensation is high, stock grants are generous, and five employees other than Huang have become billionaires.

Actionable Insights

Lessons from the Nvidia playbook that transfer to any industry facing technological change

01

Market Creation Over Market Share

CUDA built a new category from scratch. By creating a "zero-billion-dollar market," Nvidia established an uncontested standard and deep competitive moat—contrasting sharply with its strategic exit from fiercely competitive mobile chips.

What "zero-billion-dollar market" can we create where our unique capabilities give us an unfair advantage?

02

Failure as Strategic Intelligence

From NV1's market misalignment to NV30's organizational breakdown and the failed Arm acquisition, Nvidia consistently used failure as a tool for systemic improvement. Each setback converted into a durable organizational lesson.

Does our culture punish failure or analyze it for strategic intelligence? Do we have a formal process for blameless post-mortems?

03

Architecture for Speed and Candor

60 senior leaders reporting directly to the CEO maximizes information flow speed. Combined with "honing the sword" (rigorous debate) and "Speed of Light" execution, this prevents the communication breakdowns that plagued NV30.

Where are the information silos and bureaucratic layers that slow critical decisions? Is our culture dominated by consensus, or do we encourage constructive conflict?

04

Long-Term Vision Against Short-Term Pressure

The decade-long ray tracing investment and the multi-year CUDA bet were made despite significant market skepticism and activist demands. By resisting these pressures, Huang made foundational investments that created immense future value.

Are our capital allocation processes optimized for quarterly earnings or decade-long value creation? How do we protect critical long-term initiatives?

The Central Question

Had Huang and Nvidia simply been in the right industry at the right time? Or had they developed a culture uniquely suited for twenty-first century innovation?

The evidence suggests the latter. CUDA wasn't luck—it was a four-year, $475M bet on a market that didn't exist, made against fierce investor opposition. The AI boom didn't create Nvidia's advantage. Fifteen years of preparation did.

As Huang enters his fourth decade leading the company, the question becomes: Has he built a truly self-sustaining culture, or does Nvidia continue to rely on his visionary presence at the top?

What is your CUDA-equivalent bet—the thing that looks like a "zero-billion-dollar market" today, but you might be judged on in 10-15 years?