🥛 A silent contender just emerged 👀

GM. This is Milk Road AI, helping you stay ahead of the markets, the machines, and the megatrends driving it all.

Here’s what we’ve got for you today:

✍️ AMD’s AI engine is finally firing on all cylinders.
🎙️ The Milk Road AI Show: AGI Already Happened... And Almost Everyone Missed It w/ Dr. Alexander Wissner-Gross.
🍪 xAI’s inner circle is shrinking as pressure builds.

Summ (formerly Crypto Tax Calculator) is a tax software built specifically for crypto. Get started with Summ with a 20% discount.

Prices as of 10:00 a.m. ET.

AMD IS FINALLY CASHING THE AI CHECK

In 1992, a man named Archie Karas drove to Las Vegas with $50 in his pocket.

Over the next three years, he went on the greatest run in gambling history, turning that $50 into $40M.

He wasn't playing safe, he was playing all-in on every hand, betting that his momentum was stronger than the house’s math.

Source: Gifdb

In the world of semiconductors, the house has always been one big green name.

You know exactly who I'm talking about.

But Lisa Su just walked up to the table with a stack of chips that would make Archie Karas sweat.

AMD just dropped their Q4 2025 earnings, and it's official:

They’ve moved out of the basement and into the high-roller suite.

What AMD actually does

Before we talk about the billions, you need to know what AMD actually is, besides that company whose stickers are on cheap laptops.

Source: SeekingAlpha

At its core, Advanced Micro Devices is a fabless architect like NVIDIA.

This means they own the high-tech blueprints and the intellectual property, but they outsource the massive, multibillion-dollar headache of actual manufacturing to world-class foundries like TSMC.

This nimbleness allows them to focus purely on out-designing the competition across four massive fronts.

Their data center division is the heavy lifter, where EPYC server CPUs (AMD's server-class CPUs that power the world's largest cloud platforms and data centers) act as the "brains" for cloud platforms.

Meanwhile, Instinct GPUs provide the raw "muscle" needed for massive AI training.

On the client side, the Ryzen brand powers everything from high-end laptops to the new wave of AI PCs that handle smart tasks directly on your device.

Their Gaming business provides the semi-custom silicon beating at the heart of the PlayStation 5 and Xbox Series X, alongside Radeon cards for PC enthusiasts who like their fans to sound like jet engines.

Finally, their Embedded division creates adaptive chips that can be reprogrammed after they are built, living inside everything from Tesla infotainment systems to 5G cell towers and surgical robots.

Why AMD is the new house favorite

In the AI era, AMD’s importance isn't just about speed, it's about breaking the industry’s total dependence on a single vendor.

While NVIDIA is the heavyweight champion with a 92% market share and a software moat called CUDA that keeps developers locked in, they’ve become so dominant that tech giants are starting to feel NVIDIA fatigue.

Source: IOT Analytics

It's like being the only grocery store in town, eventually, people get tired of paying $12 for a gallon of milk.

This leaves a massive gap for a preferred second supplier, and AMD is currently the only company filling it at scale.

Their importance lies in a few key areas, starting with memory supremacy.

Large AI models have a massive appetite for memory, and AMD’s latest Instinct GPUs often pack more high-bandwidth memory (HBM) than NVIDIA's equivalents.

This extra capacity can actually reduce the total number of chips a company needs to buy to run a model, making it a "more for less" play for the big spenders.

Then there is the full-stack play.

While NVIDIA mostly makes GPUs and Intel mostly makes CPUs, AMD is the rare breed that dominates both.

This synergy means they can sell the entire “brain and brawn” package to data centers.

Finally, they offer an open ecosystem.

While NVIDIA's software is a walled garden, AMD's ROCm is open-source, a direct play for developers who want the freedom to switch hardware without spending millions to rewrite their code.

If you think the tech is impressive, just wait until you see the receipts.

THE SOLUTION TO CRYPTO TAXES

Tax season is right around the corner.

And if crypto taxes are already stressing you out, meet Summ.

Summ is a tax software built specifically for crypto.

Why we like it:

3,500+ exchanges, wallets and crypto projects (with DeFi, NFT, staking and airdrop support)
Direct TurboTax integration
Official tax partner of Coinbase and Metamask
Helps sanity-check your new 1099-DA form/s.

The best part? You can get 20% OFF Summ for the first year.

Stop doing your crypto taxes the hard way.

Get started with Summ with a 20% discount.

AMD IS FINALLY CASHING THE AI CHECK (P2)

AMD finally broke into double-digit billions this quarter, delivering a record-shattering $10.2B top line.

Source: Fiscal(.)ai

This is more than a good quarter, it’s a fundamental shift where AMD has officially graduated into a heavyweight player in the AI space.

CEO Lisa Su called 2025 a "defining year", which is CEO-speak for the big bets we made three years ago are finally cashing out.

And if you want to see where that cash is landing, just look at the server racks.

The data center segment, the part that sells parts for giant server warehouses, is now the undisputed heart of the company, bringing in record revenue of $5.3B.

Check out the stats:

Data center now accounts for roughly 52.4% of total sales.
Gross profit margin hit 54.3%, a steady climb from the previous quarter.
Free cash flow reached a record $2.1B, nearly doubling year-over-year.

Source: Fiscal(.)ai

For the first time ever, the AI division is truly stepping out of the shadows.

While AMD doesn't explicitly break out the numbers, analyst estimates suggest that Instinct GPU revenue (the AI muscle) is now pulling weight comparable to the EPYC CPU business (the boring-but-reliable brain).

For a decade, CPUs were AMD’s only ticket to the big leagues.

But the AI boom has basically forced a changing of the guard, where the specialized AI chips are growing so fast they’re catching up to the core business in record time.

With 8 of the top 10 AI companies now using Instinct chips in production, Lisa Su didn't just knock on the door; she kicked it down.

The morning after

If the numbers were so good, why did AMD stock drop 17% after earnings?

Because Wall Street has the trust issues of an ex who just found a suspicious receipt in your pocket.

Analysts discovered that $390M of the revenue beat came from a one-time windfall of MI308 chip shipments to China.

Source: Sherwood

Without this boost, the data center segment would have only just met expectations, and many analysts expected stronger guidance for 2026.

AMD also disclosed a $360M release of previously reserved inventory, meaning they sold chips they had already written off.

This pumped their non-GAAP gross margin to a juicy 57%, but Wall Street realized that without these magic tricks, the clean margin would have been closer to 55%.

Investors want clean receipts, not a magic show.

My verdict

The thesis here is simple: Follow the CapEx.

The big five hyperscalers (Amazon, Google, Microsoft, Meta, and Oracle) have signaled a combined CapEx of roughly $705B for 2026.

Source: @peterwildeford

That is a staggering 74% increase from 2025 levels.

To put that in perspective, the vast majority of those funds is being funneled directly into AI infrastructure like GPUs and servers.

AMD doesn't need to kill NVIDIA to win, they just need big tech to stop putting all their eggs in one basket.

Major cloud players are already diversifying their infrastructure to reduce reliance on a single vendor, and frankly, they don’t have much of a choice because NVIDIA is officially sold out everywhere.

Demand for the latest Blackwell chips is so massive that supply is expected to be tighter than a pair of skinny jeans from 2010 through at least 2026.

Because hyperscalers can't physically buy enough green silicon to satisfy their AI hunger, a massive amount of that spillover demand will land right in AMD’s lap.

Also, while the market panics over short-term China noise, the smart money is looking at the 2030 scoreboard.

Wall Street tends to have the attention span of a goldfish, but the long-term play here is about who owns the foundation of the next decade's computing.

This shift is even hitting the leaderboard in China, where the "House" is losing its edge.

NVIDIA has historically dominated that room, but their market share in China is projected to crater from 40% in 2025 down to just 8% by 2026 as local vendors and trade restrictions move in.

Source: @oguzerkan

This creates a massive vacuum that AMD is perfectly positioned to fill.

Projections show AMD’s share of the Chinese AI accelerator market climbing to 12% by 2026, officially making them the second-largest foreign vendor in the region as Huawei takes the top spot.

Ultimately, the roadmap is too stacked to ignore.

Management expects to grow Data Center revenue by more than 60% annually, scaling the AI business into the "tens of billions" by 2027.

Archie Karas may have lost his $40M by 1995, but he never had a CapEx wave like this to keep him afloat.

Lisa Su has the chips, the customers, and a cash engine that everyone keeps ignoring, I'm calling this a "buy the dip" moment for anyone who likes winning streaks.

Please do not buy all your shares at once.

As always, please dollar cost average (DCA) your entry.

Shares may dip a bit further until the next catalyst hits, so consider starting with a small starter position and scaling in over time.

This isn't financial advice. I’m a writer, not a wizard.

If you make a billion, you don't owe me a dime; if you lose your shirt, don't send me the dry cleaning bill. Do your own homework.

Alright, that’s it for this edition of Milk Road AI.

We want to hear from you.