Tyler Durden’s Sunday recap on ZeroHedge laid the split out in public. Trump told Dario Amodei, Sam Altman, and Elon Musk the industry should not slow down so the government can regulate it. He said the United States is leading China in AI, that “whoever wins AI, wins,” and that talk of a pause is bringing up things that will not happen. Guardrails were fine in theory. A pause was not.
Obama went the other way. At a Manhattan fundraiser he urged House Democrats, if they take the chamber, to put a public framework on AI. He called the tech too fast and too private. He said he was neither an accelerationist nor a doomer. He wanted kids, safety, and jobs on the 2028 agenda. Minority Leader Hakeem Jeffries, on cue, said decisive action must be taken.
David Sacks, the White House AI and crypto lead, dismissed Amodei and Altman as the “pause the frontier” camp. Sacks’s line was blunt. Do not build a framework that looks like blackmail of the public and the political system. China, he has argued, will not join a global halt. A unilateral slowdown is a gift.
Then the men who run the labs posted. Amodei published “We Must Pace the Frontier.” He warned of large harms in five to ten years. He also said a full pause is a fantasy. Musk wrote three words: “Dario is right.” Altman said independent evaluators with employee-like access were a good idea and that OpenAI would do the same. The pose is safety. The verb is still build.
That is the main idea. The United States is not turning the cluster off. China is not turning the cluster off. Open-weight models from DeepSeek, Alibaba’s Qwen, Moonshot’s Kimi, MiniMax, and Zhipu are already cheap enough to run outside a frontier lab. More models mean more inference. More inference means more racks. More racks mean more megawatts. Megawatts do not travel on hope. They travel on copper.
Why a Political Weekend Matters in the Pit
For two years the American frontier labs — OpenAI, Anthropic, Google, DeepMind, xAI — have been the engine of capital spending. Chipmakers printed money. Equity markets priced a future in which the West sold intelligence by the token. Durden’s piece said the assumption underneath was simple. The frontier stays closed. China stays a cycle behind.
The crack in that story is not a law. It is weights that leak and weights that ship on purpose. Chinese open models, the recap said, can run on a laptop or a local box and still do useful work. They are cheaper per token. They do not wait for a California board to approve a release. Once the weights exist, every factory, ministry, and startup that can buy power becomes a customer of electricity, not of a single U.S. API.
Amodei’s own warning makes the copper case stronger, not weaker. He said the labs are getting better at building the next machines — recursive self-improvement. He said capability could jump in months, not years, and that a model could take over the internet with a persistent botnet. You do not have to buy the sci-fi ending. You only have to buy the spending pattern that follows the fear. If the frontier might jump, no lab that wants to exist in 2028 spends less on compute this winter. If Beijing can buy ninety percent of the quality at ten percent of the cost, as the piece claimed, Washington does not answer with a pause. It answers with more domestic capacity.
A former Anthropic researcher, Jacob Coxon, had already said the labs were racing to self-improving systems and gambling with lives. He asked Washington for a narrow antitrust waiver so rivals could sit in the same room. He wanted talks with Beijing. Amodei called a Chinese lead a grave danger. Musk has sued Altman and still agreed the frontier must be paced by building, not by waiting for Congress. That is not a coalition for less electricity.
The Metal in the Metaphor
People talk about AI as code. The constraint is physical. Training a frontier model is a warehouse of chips, transformers, busbars, and cooling. Serving that model to a hundred million users is a second warehouse, then a third, then a campus next to a gas plant or a hydro line. Each campus is copper in the board, copper in the cable, copper in the substation, copper in the generation tie-in.
The open-weight wave makes the second warehouse larger. A closed API concentrates demand in a few U.S. regions. Cheap local models spread demand to any grid that will sell night power. That is worse for a pause. It is better for long-cycle mine supply. You cannot regulate a weight file at the border the way you regulate an export chip. You can only decide whether your own utilities and miners are allowed to catch up.
Canada sits on the quiet side of that race. It has copper in the ground, hydropower in the north, and a grid that already sells electrons into the U.S. It also has a permitting clock that has killed more projects than geology has. If the political choice in Washington is “full steam ahead,” the bottleneck moves from philosophy to metal and to the line that carries it. Junior and senior Canadian copper names are how public markets express that bottleneck. They are not a pure AI ticker. They are a claim on the wiring.
Friedland’s line still holds in this frame. The world talks about tokens. The world is short holes that can be permitted. Grade has been falling for a generation. Data-center copper is incremental demand on a market that already needed mines for grids, EVs, and air conditioners. A weekend in which the White House rejects a pause is one more reason the demand clock does not reset to zero.
What This Thesis Does Not Say
It does not say buy every copper name on Monday. Mining shares are stocks first. If the AI equity bubble cracks — Durden’s own column joked about circling IPOs — GDX-style beta can take copper developers down with Nvidia. A stink bid on a researched producer is a process. A market order on a headline is not.
It does not say China cannot flood the metal market. It can. It does not say diesel at war prices will spare a Canadian open pit. It will not. Armstrong’s product squeeze and the Red Sea pipes are a cost shock in the same month as this AI fight. High oil can support the inflation case for metal and still crush margins at the mine.
It does not say Obama’s framework dies. A House that flips can still write a bill. Antitrust talk, evaluator access, and export rules can slow a chip or a cluster at the margin. They do not unbuild a substation that is already in a utility queue. The lead time on a mine is longer than a Congress. That is why the weekend still matters. The default in both capitals is more compute, not less.
It does not pick winners among Amodei, Altman, and Musk. Their fight is over who owns the frontier and who wears the safety halo. The grid does not care who wins the essay. The grid cares how many megawatts get signed.
How to Hold the Idea
Treat copper mining equity as a multi-year claim on electrification that now includes an AI arms race the White House just refused to pause. Prefer names with a path to production, a real permit file, and a treasury that survives a year of ugly tape. Separate the explorer that needs a financing window from the producer that already sells concentrate. Size so a 30% drawdown is a bad month, not an exit from the thesis.
Watch three facts, not three tweets. Utility interconnection queues. Announced data-center load in places that can actually deliver power. Mine supply that is permitted, not promised at a summit. Canada’s copper story lives in those three lists. The Obama-Trump split is only the reason the lists do not get shorter this winter.
Conclusion
One idea. The pause lost. The frontier labs said they will keep training. Open weights mean the demand is not confined to a few American campuses. Electricity does not scale on speeches. It scales on copper.
That is not a reason to be reckless in junior paper. It is a reason the long copper clock did not stop on Sunday. Washington argued about safety language. The machines kept asking for power. The power still asks for mines.
Important information
This article uses a September 13, 2026 ZeroHedge recap of public remarks by Donald Trump, Barack Obama, Hakeem Jeffries, David Sacks, Dario Amodei, Sam Altman, and Elon Musk. Those remarks are political and corporate commentary, not a copper forecast. Model-cost comparisons in that recap were not independently audited here. Copper mining equities can fall while the AI build continues. This is not advice to buy or sell any security. Speak with a licensed adviser. The author and publisher accept no liability for actions taken on this article.