The Frontier File

A public research notebook for the expensive edge of the future.

AI infrastructure. Space. Robotics. Power. Nuclear. Quantum. Emerging technology. One concentrated portfolio — and the reasoning behind the decisions.

Mapping terrain, not handing out treasure maps. Research notes only. Not financial advice.

The Orders Survived. The Prices Didn't.

An AI-infrastructure selloff collides with resilient orders, local data-center restrictions, rising financing costs and new company results. Tower, Coherent, BWXT and the question of who keeps the money.

• AI Infrastructure
• Portfolio Process
• Optics & Networking
• Power & Nuclear
• Memory
• Macro

Where the evidence has been moving.

• № 025November 1, 202615 min read

  Issue 025 — The Buyers Showed Their Hand.

  Microsoft could fund its infrastructure spending from operating cash; Alphabet and Amazon could not. AI hardware suppliers are increasingly taking on financing risk themselves while the Frontier File assesses a large PSQ hedge.

• № 024October 25, 202615 min read

  Issue 024 — The Index Hid the Damage.

  The Nasdaq-100 fell less than one percent, yet the Frontier File portfolio lost 3.48%. SpaceX, Meta and Tesla show three different ways to finance the AI buildout while the PSQ hedge faces a crucial test.

• № 023October 18, 202616 min read

  Issue 023 — The Demand Held. The Prices Didn't.

  TSMC and ASML show AI infrastructure demand is still real. CoreWeave's debt raises the financing question. A reconstructed defensive portfolio avoids part of a brutal week but leaves a nearly 30% PSQ hedge that needs an exit plan.

• № 022October 11, 202615 min read

  Issue 022 — The Pipes Are Full. Who Pays?

  Lumentum says its optical capacity is sold out through early 2029. The Treasury borrowed at its highest 10-year auction yield since 2000, while the Fed discussed AI-related borrowing. The demand looks real. So does the invoice.

Every issue looks for five things.

1. 01What actually changed?
2. 02Which thesis gained or lost evidence?
3. 03Where is the bottleneck?
4. 04What catalyst comes next?
5. 05Does any of it deserve more capital?

The goal isn't to predict every wiggle in the market. It's to document how conviction changes when the evidence changes.

The rules the portfolio actually runs on.

I'm Alex Oklesson, a self-directed retail investor running a concentrated frontier-technology portfolio with a strong growth and asymmetric-upside orientation.

Seasoning, not the steak

Speculative names stay small on purpose. Position size is the risk control.

Catalysts over vibes

Dated events earn weight. Narrative alone does not.

Cheap ≠ distressed

A low price is not a thesis. Broken economics stay broken.

Four tiers

Core · Growth · Speculative · Moonshot — each with its own sizing rules.

Explicit exclusions

What I refuse to own is written down, not improvised.

BWXT: The Quiet Position With the Loudest Opportunity

• The Precision Components Group acquisition adds more than 500,000 square feet of U.S. manufacturing capacity.
• Potential action: an approximately $100 post-earnings addition only if guidance, backlog and cash conversion remain intact. No purchase has been made.

26 issues, June 1, 2026 to now.

The whole record stays public — including the calls that aged badly.