Is AI Cracking – or About to Break Out?

AI stocks are rebounding after an end-of-half/holiday rotation and profit-taking move, with the Nasdaq 100 down 4% from mid-June but later jumping as sentiment steadies. The piece points to Bitcoin miners gaining ~56% YTD despite BTC being down nearly 30% YTD, citing grid/energy infrastructure constraints and a 20-year TeraWulf data-center lease by Anthropic (WULF up ~14%). On commodities, Eric Fry reiterates a 2026 copper call, with copper at about $6.20/lb vs a $7.50/lb target; he cites a projected ~150,000 tonne refined-copper deficit and says >$200B of new mining investment is needed, highlighting Freeport-McMoRan (FCX) as a preferred play.

Analysis

The market is re-pricing AI around a scarcer input than chips: delivered megawatts. That shifts the economic rent from pure compute vendors toward companies that already control energized sites, interconnects, and build-ready infrastructure. In that frame, NVDA remains an enabler, but the more asymmetric upside sits one layer down the stack where power-constrained hyperscalers are forced to pay up for speed and optionality.

That favors the best-positioned converters, not the average Bitcoin miner. IREN looks higher quality than the group because the market can underwrite an AI hosting business, while CIFR and WULF still trade with heavier dependence on contract visibility and narrative momentum; single-lease announcements can lift revenue expectations, but they do not yet prove durable EBITDA or free-cash-flow conversion. Second-order winners likely include grid and thermal infrastructure suppliers such as VRT, ETN, and PWR, because every new AI site requires more than land — it needs substations, transformers, switchgear, and time that the market is still underestimating.

FCX remains the cleaner structural expression if the thesis is copper scarcity rather than AI sentiment. The contrarian risk is that the crowd is chasing the same “power bottleneck” story before enough projects have actually cleared interconnection and financing hurdles; if 1-3 month follow-on contract announcements disappoint, the miners’ multiple expansion can reverse faster than the copper complex. Falsifiers: copper losing the $6 handle for a sustained period, AI hosting deals failing to translate into guided revenue, or BTC strength dragging miners back into pure hash-rate beta. Time horizon matters: the miner trade is tactical-to-3 months; FCX is 6-18 months if capex discipline and supply deficits persist.

Key Decisions for Investors

• Tactically long IREN vs short CIFR for 1-3 months: express the view that the market will pay for durable AI infrastructure optionality over generic miner beta. Use a tight stop if IREN loses relative strength after the next contract/news cycle.
• Buy FCX on pullbacks for a 6-12 month hold: this is the cleaner way to own the electrification/copper shortage theme without paying peak multiple for AI names. Falsify below a sustained $6.00/lb copper tape or if FCX guidance stops implying leverage to the price move.
• Treat WULF as a trading vehicle, not a core long: consider taking partial profits into strength unless management can show a repeatable pipeline of long-duration AI leases. The risk/reward worsens quickly if the market starts discounting one-off lease announcements as non-recurring.
• Keep NVDA as a hedge, not the highest-conviction new add: if power scarcity intensifies, chip demand remains intact, but the incremental upside increasingly migrates to infrastructure and power-chain names. Use NVDA strength to finance longs in FCX/IREN rather than chase it outright.
• Watch for a basket trade in power infrastructure on confirmation: ETN/VRT/PWR strength would validate the thesis that the real bottleneck is electrical delivery, not chip supply. If those names fail to participate, the AI-power narrative may be ahead of fundamentals.