SPAGHETTI DIAGRAMS DON’T LIE

(NVDA), (XLK), (META), (GE), (RTX), (GEV), (ETN), (LLY), (BRK.B), (CEG), (VST)

I once made a six-figure bet on NVIDIA (NVDA) after a lunch in San Francisco where an engineer sketched a data flow diagram on a cocktail napkin. It looked less like elegant math and more like spaghetti that had lost a brief but meaningful fight with gravity. Still, the point landed. 

GPUs weren’t just faster chips. They were the nervous system of artificial intelligence. That napkin is now framed in my office – not out of nostalgia, but because it reminds me that the best investment insights rarely arrive with a PowerPoint deck or a PR team.

Fast forward to today and NVIDIA (NVDA) has become the central bank of the AI economy, issuing compute instead of currency. Everyone sees it. Everyone owns it. Which is exactly why the conversation has to move on. 

The next phase of AI investing has moved past discovering AI and is now about underwriting it. Artificial intelligence is expensive, power-hungry, and deeply impatient with inefficiency. Once you accept that, the market starts to look very different.

Technology’s dominance in 2025 was no mystery. The Technology Select Sector SPDR ETF (XLK) surfed a perfect wave of GPU shortages, hyperscalers panic-buying capacity, and executive teams terrified of being asked why their company lacked an AI strategy. 

With NVIDIA’s (NVDA) Vera Rubin systems scheduled for the second half of 2026, the hardware cycle still has momentum. But momentum eventually meets arithmetic. Margins compress. Alternatives emerge. Owning only the chipmakers at this stage is a bit like buying shovels after the gold rush already has a Starbucks.

The more interesting story is what happens once AI leaves the keynote stage and shows up on the balance sheet. Meta Platforms (META) learned this the hard way. Investors applauded the ambition right up until capital expenditures started resembling a defense appropriation. 

Shares stabilized only after Reality Labs took a very public haircut. The takeaway wasn’t that AI dreams are bad. It’s that markets have a low tolerance for open-ended spending, even when the vision is compelling.

Industrials, to their credit, have been far more pragmatic. Defense names like GE Aerospace (GE) and RTX (RTX) didn’t rally because investors suddenly felt nostalgic for jet engines. They rallied because modern defense is software-driven, sensor-heavy, and energy intensive. 

GE Vernova (GEV) and Eaton (ETN) may never trend on social media, but they are quietly selling the infrastructure that keeps data centers online and grids from buckling. AI doesn’t run on optimism. It runs on electrons, preferably delivered on time.

Healthcare is playing a subtler hand. Eli Lilly (LLY) made headlines thanks to obesity drugs, but behind the scenes AI is reshaping how drugs are discovered, trials are run, and factories are optimized. This isn’t about chatbots diagnosing patients. It’s about replacing guesswork with computation. That kind of shift rarely grabs headlines, but it has an excellent track record of expanding margins.

Financials are benefiting in ways that don’t make for exciting earnings calls. Fraud detection, credit scoring, compliance, and risk modeling are increasingly algorithmic, whether banks advertise it or not. 

Berkshire Hathaway (BRK.B) may weigh on short-term ETF performance with its massive cash position, but across the sector AI is being embedded quietly, the way banks prefer their innovation: useful, defensible, and slightly boring.

Even utilities are enjoying a renaissance of sorts. Data centers are not polite consumers of electricity, and AI workloads don’t pause for ESG debates when latency is on the line. 

Nuclear-adjacent players like Constellation Energy (CEG) and hybrid operators like Vistra (VST) are suddenly popular with hyperscalers who have learned that reliability beats ideology. It’s possible the clean energy transition ends up being financed by machines that refuse to sleep.

Not every sector participates equally. Consumer staples and traditional REITs remain hostage to inflation and rates, with only selective exposure to AI-driven demand. Logistics and specialized real estate will matter eventually, but much of the sector is still waiting for the future to arrive.

As 2026 approaches, the most important question for investors isn’t who talks the loudest about AI. It’s who sends the invoice. The real winners will be the companies selling power, stability, and infrastructure to an industry that consumes capital and electricity with remarkable enthusiasm. 

The napkin on my wall hasn’t changed. The spaghetti diagram still looks ridiculous. But the lesson aged beautifully. In AI, the smartest money still follows the mess.