The structural bottlenecks that move markets — mapped daily. Who's positioned, who's exposed, how tight the constraint. Not tips. The map underneath the trade.
Somewhere in the supply chain
Joined March 2026
- Tweets245
- Following97
- Followers13
- Likes72
Pinned Tweet
1/ The consensus take on AI compute: “GPU supply will catch up to demand.”
Our pipeline read 139 signals against the actual mechanism and disagrees. The bottleneck isn't design or wafer allocation.
It's foundry throughput hitting a ceiling while inference demand grows non-linearly. 🧵
1
1
26
1/ Next Monday I open the TrendNinja waitlist. Here's how founding access works — and why the first people in win the most. 🧵
1
10
4/ Position is set by signup time — so being early matters even if you never refer a soul. Early + referrals = the winning combo.
1
4
5/ Open Now. Catch it in the first hours — that's when the best positions are still open. → trendninja.app
4
This week I published 5 structural cards, each one a reframe of a consensus narrative:
• GPU: not demand — foundry throughput ceiling
• Minerals: not mining — refining monopoly
• Data centers: not power — permitting
• Electricity: not clean transition — dispatchable-capacity gap
• Plus the falsifier discipline behind all of them
All public. All timestamped. The record starts here.
If this is how you think about markets — theses, not tickers — the full cascades are at trendninja.app. 🥷
1
4
1/ Coal retirements get framed as a clean-energy win. 48 signals reframe it: it's a dispatchable-capacity supply constraint. 🧵
1
3
1/ “Data centers need more power.”
54 signals say the power exists. 165–250 projects are stuck in local permitting and community opposition — not waiting on generation. 🧵
1
4
5/ Falsifier:
FERC co-location pre-emption overriding local zoning, or a major metro fast-tracking AI permits. Backlog would clear in 18–24 months.
1
1
6/ Public + timestamped.
Full cascade for researchers. → trendninja.app
Every card we publish names what would prove it wrong.
The GPU card: 'dissolves if TSMC ships 800MW before 2028.'
The minerals card: 'dissolves if the West builds 10M tonnes of refining in 36 months.'
Most research never tells you its own kill condition. We lead with it.
If the constraint breaks, you'll know the thesis is dead before the price does. That's the entire point.
1
1/ Everyone frames critical minerals as a mining story. "Who digs it up."
Wrong frame. The chokepoint isn't the mine — it's the refinery. And on the minerals that matter most, China controls ~90% of processing. 🧵
1
2
6/ The more interesting layer — who owns non-Chinese processing:
The scarce asset isn't ore in the ground. It's separation and smelting capacity outside China. The vertically integrated players — mine + smelter + refinery under one roof — are the ones the chokepoint rewards, not threatens.
UAMY is the clearest US example: it just commissioned a domestic antimony smelter and holds a sole-source DLA stockpile contract. That's the chokepoint working in your favor.
1
1
7/ The falsifier:
This thesis weakens if the US/EU/India bring large-scale non-Chinese refining online within ~36 months, or China permanently lifts the controls. Watch for a major Western refinery with committed offtake.
The reframe — "it's processing-constrained, not supply-constrained" — is the whole card. The full supplier cascade is what researchers get.
1