Marble H1 2026

On June 27, 2026, Marble crossed 2 million USD, an 81.5% return since January 1. Over the same period of time, the SPY returned 7% and the QQQ 15%.

Memory, Bottlenecks, and a sprinkle of Geopolitics.

But the word of the year is Agents.

Around early 2025 something big happened. The tech world began waking up to the fact that AI was getting good enough to prompt itself.

As developers opened their eyes to the wonders of Cursor and Claude Code, so did the market.

Agents work differently from ChatGPT. They’re far more capable, and far more demanding.

Source: Coatue Management (May 2026 Public Markets Update)

The new bottlenecks to the AI expansion have seen legendary runs.

Intel (CPU) +250% YTD

AMD (CPU) +150% YTD

Micron (Memory) +300% YTD

Sandisk (Memory) +750% YTD

All hundred billion to trillion dollar companies.

Adjacent industries such as Connectivity, Energy, and Neoclouds have also benefitted.

The search for the next bottleneck in the AI trade has created an interesting social media culture of stock gurus and microcelebrities.

Outside of the hot AI hardware sector, the broader market has been largely dissapointing.

Consumers remain weak and economic data remains mixed. Throw in geopolitical tension and it’s not surprising to see that outside of AI, the market has been negative since the beginning of the year.

On the exact opposite side of the AI hardware euphoria has been the “SaaSpocalypse”.

Adobe -40% YTD

Atlassian -50% YTD

Hubspot -50% YTD

Figma -50% YTD

Investors are no doubt worried of the future of these companies as Anthropic and OpenAI continue to threaten their businesses.

Reflecting on H1 2026

This was Marble’s best half since our founding in 2023. No doubt a success, yet somewhat bittersweet.

Let’s start with what we did poorly.

Remember Micron? Earlier today Micron traded at $1200.

We bought in at around $600 (Technically indirectly through the DRAM etf).

One year ago Micron was $100. We first looked at Micron at $230 knowing memory was becoming integral to agentic workflows but decided to pass feeling it was up too much… Hopefully you can see why we feel bittersweet. But there are no “could haves” in investing.

The lesson we take away is that as important as it is to be right, it’s also important to dive into the scale of “how” right you are.

Sometimes the market is underpricing by 100% and sometimes the market is underpricing by 1000%.

What we did well

At the start of 2026, Marble had a large cash position, feeling paralyzed by high valuations and geopolitical uncertainty.

We did however begin our heaviest research in the AI hardware space mapping out the entire supply chain which gave us the confidence to re-enter some names.

Had we not re-entered it’s very likely Marble would be flat for the year. Other investments we're proud of:

DOCN (Cloud for startups - Benefactor of agentic usage) +130%

CRDO (Copper wiring allowing GPU’s to communicate faster) +140%

GFS (Foundry for photonics - Using light to send data in GPU’s) +90%

ACMR (Semiconductor wafer cleaning) +120%

BRBR (Protein Powder) +40%

SEZL (Pay in 4) +50%

CGNX (Robotic Vision) + 90%

Credit to some of our amazing analysts for finding these gems.

Looking Forward

As the AI trade begins to stutter a bit, Marble maintains a cautiously optimistic view.

We’re definitely more cautious than we were a few months ago.

One thing that scares us is Tokenomics.

The gist is that AI quality has diminishing returns for many tasks.

Fable 5 might write better emails than GPT 4o, but GPT 4o is already good enough for writing emails.

Every task has a level of “intelligence” where returns begin diminishing.

If Opus 4.8 can already oneshot most productive code, will most companies see the need to spend significantly more for a better model?

We're finally seeing companies actually consider token costs and ROI.

In this sense, we think that many assumptions about margins must come under scrutiny,

Moreover, we believe that companies that can dictate AI economics will benefit while companies who follow the tide might get swept away.

Increasingly, industries insulated from AI are becoming a core part of Marble's attention moving forwards.

Evergreen Industries

Her's an example of what we're talking about.

Senior Living

AI might replace jobs or it might not.

• Seniors have already made their money
• Seniors need specialized living spaces
• Seniors are willing and able to pay for specialized living spaces

We can have AGI or it can all be a huge bubble but these 3 will remain true.

Other Topics of Interest for 2026

Looksmaxxing

It’s clear to us that appearance is becoming more important in the mind of the young generation.

• Protein powder (GenZ loves the gym)
• Plastic surgery
• Nutrition Industry (And GLP-1s)

Weak Consumer

Economic data points at the average consumer not doing very well.

We can see this in weak performances from luxury companies and general consumer titans.

• Pay in 4
• Bulk Stores
• Discount Stores

Agentic Cybersecurity

Cybersecurity is going to be a nightmare once agents are used to attack software.

Which company are companies going to turn to for protection?

SaaSpocalypse Survivors

The market is pricing in doom for all of software. For some companies this is mostly definitely an overreaction.

Which software companies will Anthropic/OpenAI struggle to disturb?

If at-risk companies are looking to protect themselves from AI, who are they turning to?

Robotics

Marble is less bullish on robotics than many. We respect the potential but we feel this is an industry that is too early, similar to quantum computing.

We've observed that markets truly begin to react when earnings begin to beat and that's still a ways off for robotics.

That said, we believe some upstream industries such as sensors, computer vision, LiDar are already heating up.

Japan + China

Japan essentially has many companies that are top dogs in key semiconductor verticals that are now undervalued compared to western counterparts.

China on the other hand is also making waves in AI. With Fable 5 being restricted, it’s clear that the AI wars are being taken on a new level.

The Chinese government just commited 295 billion dollars to domestic AI buildout and is dedicated to ensuring they have a full supply chain.

What’s next for Marble

Historically Marble’s been pretty quiet, but we’re hoping to change that!

We have some interesting stuff cooking up.

• Antidisplacement - A discussion on how to invest to protect your software career against AI

Own the companies trying to replace you

• Monthly recaps on our thoughts on the market