Cerebras Is a Taxi. NVIDIA Runs the Bus Network (Pt1.)

14 min · 25 Aug 2026

Cerebras can generate tokens at extraordinary speed for a handful of users, but NVIDIA’s GPU architecture delivers superior economics at scale. This report explains why the fastest AI inference system does not necessarily make the best mass-market token factory — or the better investment.

4Q26 Portfolio Rebalance: Positioning for a Shift in AI Infrastructure Leadership (Pt.1)

AI spending keeps growing, but the winners are changing. Our 4Q26 rebalance favors Broadcom over NVIDIA and backs suppliers enabling logic folding and optical integration. We also explore Tesla’s physical AI opportunity and the risks to TSMC’s long-term dominance.

38 min read 09 Oct 2026

10 min read 14 Sep 2026

ABBX Week 17: Oracle’s Backlog Matches a Hyperscaler. Its Balance Sheet Does Not.

ABBX Week 17 finds no broad AI bubble. Oracle is the test case: a USD 664 billion backlog, neo-cloud-like equity value, and a lease-and-capex bill that still lands before that backlog turns into cash.

10 min read 30 Aug 2026

The AI Bubble That Isn’t — Yet: What the Completed Q2’26 Data Reveals

Q2’26 ABBX is complete. Parts of the AI market are hot, but the ecosystem is not yet a broad bubble: ROIC still exceeds funding costs, multiples remain below May, and demand is still broadening. The real heat is off-balance-sheet financing, not headline debt.

Read next