💰 GPUs Built AI, CPUs Run It

THE MAIN EVENT
GPUs Built AI, CPUs Run It ⚡

For most of the AI boom, GPUs were the entire story.

Nvidia became the face of that era. Its chips trained the models, powered the infrastructure, and turned the company into the most valuable business in the world.

Since the launch of ChatGPT, Nvidia’s data center revenue has exploded from $15B annually to $194B last year.

But AI is entering a new phase.

This week offered a symbolic glimpse of that shift: Alphabet (Google) briefly passed Nvidia in market value before Nvidia rallied back and crossed the $5T mark for the first time since October.

That doesn’t mean Nvidia is losing the AI race. It means the race itself is starting to broaden. AI is still about GPUs. But it is no longer only about GPUs.

The shift comes down to two words: training and inference.

Training is when companies build massive AI models. That remains Nvidia’s kingdom because GPUs are built to process huge amounts of data all at once.

Inference is what happens after the model is built.

Every chatbot response, coding assistant, AI search result, customer service bot, or autonomous agent performs inference. And as AI agents start handling more multi-step tasks, the hardware mix begins to change.

That’s because inference relies far more heavily on CPUs.

If GPUs are the brainpower, CPUs are the coordinators. They handle the step-by-step work: opening apps, pulling data, calling tools, checking outputs, and sending responses.

AMD CEO Lisa Su put it plainly on the company’s earnings call. Running AI models once required roughly one CPU for every four to eight GPUs. Now, she says, that ratio is moving closer to one-to-one.

With enough AI agents, there could eventually be more CPUs than GPUs.

That is the money line.

GPUs built the models, but CPUs may end up running the AI economy around them.

That is why the AI winner circle is starting to expand.

• AMD is the clearest dual-threat: AMD makes both the GPUs that power AI workloads and the CPUs needed to coordinate them, giving it a cleaner shot at selling the full package. Its data center revenue jumped 57% year-over-year to $5.8B, and the company doubled its 2030 CPU server market forecast to $120B from $60B just a few months ago.

• Intel is getting a second look: Its Data Center and AI segment rose 22% year-over-year to $5.1B. Intel largely missed the GPU gold rush, but CEO Lip-Bu Tan noted AI is moving “from foundational models to inference to agentic.” This shift back toward CPU-heavy coordination gives Intel’s Xeon processors fresh relevance.

• Arm is riding the power-efficiency wave: Arm licenses chip designs known for doing more work with less electricity, giving it an edge as AI spreads to devices and as data centers try to control power costs. The company claims it already has 50% market share for CPU compute among top hyperscalers.

• Google is building its own lane: Alphabet’s custom in-house chips (“TPUs”) help it run AI workloads inside Google Cloud more efficiently, lowering reliance on outside suppliers.

• Micron and Samsung are the supporting proof: This week, Samsung joined the elite “four-comma club” with a $1T+ valuation, while Micron crossed $700B. Their rallies show investors are looking beyond GPUs and toward the broader AI hardware stack.

Nvidia sees the shift, too.

At its annual developer conference, Jensen Huang leaned harder into inference and agents, showing off new systems that combine GPUs, CPUs, and other specialized chips into unified AI platforms.

The subtext was clear: Nvidia knows the future is not just selling the best GPU. It is selling the full AI system.

That matters because Nvidia still has the strongest ecosystem, the deepest customer relationships, and the clearest claim as AI’s chip king.

But AMD, Intel, Arm, Google, Samsung, and Micron are each proving there are more ways to win as AI moves from “build the model” to “use the model.”

So no, this isn’t an Nvidia obituary. It’s a reminder that building AI and running AI are two different businesses.

MARKET MOOD
AI Buildout Keeps Pushing Stocks Higher 📈

Winners

DigitalOcean ($DOCN) - Market Cap: $15.7B (Week-to-Date: +46.3%)

DigitalOcean sells cloud computing tools to developers and small businesses. The stock ripped after annualized AI revenue hit $170M, up 221% year-over-year. Its new AI-Native Cloud positions DOCN as both a home for AI agents and a cheaper, simpler alternative to AWS, Azure, and Google Cloud.

Micron ($MU) - Market Cap: $729.0B (Week-to-Date: +19.2%)

Micron blew past a $700B market cap, with the stock now up 127% this year and nearly 700% over the past twelve months. The reason is simple: AI memory demand is overwhelming supply. Micron’s 2026 inventory is fully booked, so customers are signing multi-year prepayment deals. Meta, Microsoft, and Amazon all flagged higher component costs, with memory doing much of the damage. That’s pricing power in plain English.

Corning ($GLW) - Market Cap: $157.0B (Week-to-Date: +15.3%)

Corning rose after landing a multi-year Nvidia partnership that brought attention to a bottleneck I flagged back in March: AI infrastructure needs a lot more fiber. Corning will build three new facilities in Texas and North Carolina, increasing U.S. optical connectivity production by 10x and optical fiber by 50%. The Street is starting to see it less like a sleepy industrial and more like AI infrastructure.

Losers

Planet Fitness ($PLNT) – Market Cap: $3.5B (Week-to-Date: -32.7%)

Planet Fitness beat Q1 numbers, but the outlook did the damage. Revenue rose 22% to $337M, yet management cut 2026 profit guidance to $3.19 per share and cancelled planned membership price hikes. Net new member adds fell 36%, which is the real problem. My theory: casual gym users may be the easiest to lose if GLP-1s make weight loss feel less dependent on a $10 gym membership.

Shake Shack ($SHAK) – Market Cap: $3.0B (Week-to-Date: -29.6%)

Shake Shack got squeezed where it hurts most: beef. Revenue rose 14% to $367M, but operating profit disappointed as margins fell from 12.7% to 10.1%. SHAK’s angle is premium fast-casual, so beef inflation is hard to pass through. Raise prices too much, and a “better burger” starts looking like an easy budget cut.

Shopify ($SHOP) – Market Cap: $145.7B (Week-to-Date: -12.5%)

Shopify’s selloff wasn’t about weak demand. Revenue rose 34%, while gross merchandise volume (total sales running through the platform) rose 35%. The issue was soft guidance and a $581M net loss as the company keeps spending on AI-powered commerce. The market has decided that “growing fast at a loss” needs to become “growing fast at a profit.”

CHART OF THE WEEK
Retirement Planning Is Easier With A Roadmap 🗺️

With stocks near all-time highs, now’s a great time to zoom out and think about the bigger picture.

That’s exactly why I built The Money Maniac FIRE Calculator.

FIRE stands for “Financial Independence, Retire Early,” and the calculator lets you compare multiple paths side-by-side:

• Lean FIRE: Retire as soon as possible by limiting spending to just the essentials.

• Barista FIRE: Semi-retire while part-time income fills the gap.

• Coast FIRE: Stop saving for retirement and let compounding grow your existing portfolio until retirement.

• Fat FIRE: Aim for a more aspirational retirement lifestyle with higher long-term spending.

One of the coolest parts: you can also see how working an extra year or two beyond your target retirement date changes your long-term monthly income.

Try it out here →

FAST FACTS
Yield Traps, Shark Tanks, and Cruise Bugs… Oh My ⚠️

💸 Wall Street’s yield trap: Flashy new ETFs promise annual yields as high as 19%, but when markets turn, both the income and your principal are at risk. [Read]

🦈 The sharks are eating everyone: WSJ found insiders “elite traders” dominate prediction markets, with 0.1% of accounts taking home 67% of profits. [Read]

🦠 This feels too familiar: Health officials are tracing a cruise-linked hantavirus outbreak spanning 23 countries after several deaths onboard. [Read]

🎓 Best cities for new grads: Tampa jumped from No. 26 to No. 2 thanks to strong hiring and lower costs, but you’ll never guess which city took the top spot. [Read]

🍎 Apple may owe you money: Apple agreed to a $250M settlement tied to AI features promoted before they were actually ready on the iPhone 16. [Read]

💼 Jobless claims hit 50-year low: New unemployment filings fell to their lowest level since 1969, suggesting companies aren’t eager to cut workers despite AI fears. [Read]

WORDS TO REMEMBER
Attention Is Too Valuable To Waste 🧠