Intel Made This Exact Mistake in 2018. Now It’s NVIDIA’s Turn to Choose.

In June 2018, Intel’s board held an emergency meeting.

$Intel(INTC)$  

$NVIDIA(NVDA)$  

Their CEO was ousted over an office relationship. With no succession plan in place, the CFO was pushed up as a "temporary placeholder."

In November 2026, NVIDIA’s board held an almost identical meeting.

Eight years ago, that "temporary" decision cost Intel an entire generation of process leadership. Today, NVIDIA stands at the exact same crossroads.

The only difference: Intel made the wrong choice. NVIDIA still has time to get it right.

1. Every Decision Was "Logical," Yet the Company Lost

In 2018, Intel CEO Brian Krzanich resigned after violating company policy. The board appointed CFO Bob Swan as interim CEO, making him permanent in early 2019.

Swan wasn’t a terrible CEO. He protected gross margins and kept shareholder returns flowing. The problem? Every single decision he made was driven by financial logic—at the exact moment Intel desperately needed engineering vision.

 * The 10nm Trap (Stopping the bleeding vs. breaking through): Yield issues predated Swan, but his response defined the next three years. Facing delays, he poured capital into 14nm to protect revenue while taking a conservative stance on 10nm. Financially, this was "pragmatic." In engineering terms, it handed the crown to competitors. While TSMC scaled 7nm and AMD lined up Zen 2 orders, every quarter Intel clung to 14nm widened the gap by a full generation.

 * Apple’s Baseband Business (Winning the spreadsheet, losing the war): In April 2019, after Apple settled with Qualcomm, Swan exited the 5G smartphone modem business and sold the division to Apple for ~$1B. On paper, cutting a single-customer, money-losing unit made total sense. Strategically, it severed Intel’s last deep technical tie with Apple and erased its final foothold in mobile.

 * AI Acceleration (Waiting for clarity): Between 2019 and 2020, while NVIDIA went all-in on CUDA and Tensor Cores, Swan’s team ran the numbers on AI. Their conclusion: the Total Addressable Market (TAM) was unclear and the payback period too long. They were right about the math, but missed a fundamental truth: during a tech paradigm shift, market size is created, not calculated.

The Delayed Fuse:

Swan took over in early 2019, yet Intel’s market share didn’t visibly collapse until 2021–2022. Because semiconductors run on 24-to-30-month product cycles, under-investing in 2019 meant having nothing competitive to sell by 2021.

Financial consequences are linear and immediate. Engineering consequences are exponential and delayed. By the time the financial statements showed the damage, an entire product cycle was already lost.

2. Why a CFO Shouldn’t Lead a Tech Boom

CFOs make great CEOs when an industry matures and needs cost discipline and cash flow management. A technology boom requires the exact opposite.

Look at tech history when non-technical operators take the wheel at an inflection point:

 * Steve Ballmer at Microsoft (2000–2014): Sales/finance background. Protected Windows and Office profits, completely missed mobile and cloud. The stock went sideways for 14 years until cloud-architect Satya Nadella took over.

 * Léo Apotheker at HP (2010–2011): Finance background. Killed webOS within a year because short-term ROI didn't pencil out. Ousted in 11 months; HP never recovered its strategic edge.

Now look at the survivors of tech transitions—all visionary builders:

 * Lisa Su (AMD, 2014): Didn't start by slashing costs; bet the farm on the Zen architecture.

 * Sam Altman (OpenAI): Poured billions into training frontier models before a business model even existed.

 * Jensen Huang (NVIDIA, 2012): Staked the entire company on CUDA while AI was still an academic curiosity.

During a tech explosion, vision isn't a luxury—it’s survival.

3. Back to 2026: NVIDIA’s Temptation

Appointing CFO Colette Kress as Interim CEO is a smart move. She knows Wall Street, understands governance, and will ensure Blackwell and Vera Rubin ship smoothly over the next 18 months while keeping earnings stellar.

But "Interim" and "Permanent" are two entirely different jobs.

The temptation for NVIDIA’s board will be massive: Kress has steadied the ship and the numbers look great—why not keep her on permanently? Or hire another safe, Wall Street-friendly operator?

That is a trap. AI has only just finished its first curve:

 * Curve 1 (2023–2026): Selling Shovels. Everyone needs GPUs. Success relies on compute, capacity, and delivery. The roadmap is already locked in; an execution team can run this.

 * Curve 2 (2026–2029): Selling Ecosystems. AI moves from cloud to edge, text to multimodal, data centers to robotics and autonomy. Winning requires software stacks, system integration, and extreme co-design with TSMC’s next-gen nodes.

 * Curve 3 (2029+): Selling the "OS of AI Infrastructure." Setting the standard for physical-world AI, just as Microsoft defined enterprise computing.

Curve 1 runs on execution. Curves 2 and 3 run on vision.

Who negotiates A16 process co-design with TSMC? Who locks in $100B sovereign AI infrastructure deals in the Middle East? Who defines chip specs for physical AI alongside robotics pioneers? And when the Feynman architecture hits a thermal wall in 2028, who makes the gut call to redesign the cooling rather than throttle clock speeds?

There are no historical regressions or neat ROI models for these decisions. A CFO trained to account for every penny is simply wired differently.

4. The Real Choice: A Caretaker or a Creator?

If NVIDIA posts two more blowout quarters, investors will breathe a sigh of relief: "See? Nothing to worry about."

That is word-for-word what Intel investors said in 2019.

NVIDIA’s 2028 product lineup depends on engineering bets made between late 2026 and early 2027. If the person in the big chair only asks "What’s the NPV of this project?" instead of "What compute architecture will the world need five years from now?", Feynman’s development cycle will quietly stretch from 12 months to 18.

Those extra six months won’t show up in any quarterly footnote. They will silently appear in 2029 as extra line items for AMD or AWS Trainium on hyperscaler purchase orders.

NVIDIA doesn't need another safe pair of hands. It needs a successor with three traits:

 * Can argue with engineers at a whiteboard. Not nodding at slide decks, but asking: "Why aren't we using chiplets here? Why can't we double NVLink bandwidth?"

 * Willing to bet big in the fog. Slapping $5B on the table when market data is still blurry because they see where the world is heading.

 * Defines the game rather than optimizing within it. Sitting eye-to-eye with TSMC, sovereign leaders, and cloud CEOs to shape the next era of computing—not just negotiating purchase orders.

Jensen has stepped down, and Kress has brilliantly caught the falling baton. Now, the board isn't hiring someone to sign off on budgets. They are hiring someone willing to say: "Build it. I’ll take the heat."

That single choice will decide whether NVIDIA continues to rule the second half of the AI era.

# 💰Stocks to watch today?(16 November)

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.

Report

Comment(2)

  • Top
  • Latest