Utilities like those in the XLU fund have traded at a premium because investors assumed AI data centers would guarantee decades of extra electricity demand. Texas utilities now freezing or revising down data-center demand assumptions removes the core reas

Verdict: a compelling AI-capex unwind story still has no tradeable trigger

The verdict is a clean wait. The strongest point for the bearish-XLU thesis is concrete: Microsoft's June-quarter free cash flow fell about 71% year over year to $19.6B, with capex of $115.9B consuming roughly 63 cents of every operating-cash dollar — the fragility the idea's $1.3 trillion hyperscaler-spending argument relies on. The strongest point against is structural: the entry rules never triggered in the last 60 months over 1,225 evaluated bars, and one compiled condition (price at or below 0) can never be satisfied, so no entry can fire until the rule set is corrected. Microsoft's own counter-evidence is real too — debt-to-equity at a record low 0.07, a 30.2% return on equity, and a dividend raised to $0.91 per quarter — which weakens the near-term capex-cut story even as margins compress. What would flip the verdict to actionable: a decisive XLU close below the $42.36 support rung with the SMA (20) at $43.42 decisively under the SMA (50) at $44.37 and rising ADX (14) of 31.1, ideally paired with another sequential free-cash-flow drop in the September filing. Until then, monitor, don't buy.

Trade now: XLU setup is armed but not triggered

Nothing to execute yet — this is a watch-list setup, and the entry rules on XLU (1d) have not all triggered. Two of three conditions are live: ADX (14) is 31.1, above the 20 threshold, and the SMA (20) at $43.42 sits just below the SMA (50) at $44.37, so the bearish cross is in progress (status: near, gap of about $0.94). The remaining condition — price at or below 0 — is structurally out of reach at the current XLU close of $43.08, so treat it as a built-in filter rather than a level to wait for. Until the momentum cross confirms cleanly, "wait" means no position and a daily check on the two moving averages. If the setup triggers, the framework is explicit: a hard stop at a 2.4% loss on the position, a take-profit at 4.8%, and a secondary exit if price breaks below the second-ranked support level. That produces an effective reward:risk of roughly 2:1 on the trade structure, with position size capped at 25% of capital under fixed-risk sizing. The thesis itself is advancing independently of the trigger. Per the SEC XBRL data, Microsoft's June-quarter free cash flow fell about 71% year-over-year to $19.6B and operating cash flow fell about 70% to $55.4B, while insider filings show net open-market selling of roughly $27.0M. Those numbers support the idea's argument that AI capex — forecast at $1.3 trillion by 2027 with only one hyperscaler expected to produce positive free cash flow — is vulnerable to cuts. No robust alternative parameter setup was established for this strategy, so the published thresholds are what you get; do not tune them yourself. One honest caveat: the setup is a rules-not-triggered watch — the rules ran on real daily bars but did not open an entry in the evaluated windows, so today's job is monitoring, not buying.

The $1.3 Trillion Bill Comes Due: Why the Utilities Premium Is Cracking

The idea argues that the speculative electricity-demand premium in utilities like XLU should unwind because Texas utilities are freezing or revising down data-center demand assumptions — and the spending behind that demand is less durable than assumed. Per the Yahoo Finance piece from September 7, 2026, "XLU's AI Power Story Crumbles as Texas Freezes Data-Center Demand," the core justification for utilities trading at a premium — guaranteed AI-driven electricity demand — is already being walked back by the very parties who were supposed to underwrite it. The fundamentals make the second leg of the argument concrete. Microsoft, a top-three holding in SPY at roughly 4.3% weight, generated $182.9B in operating cash flow in fiscal 2026 but spent $115.9B in capex, leaving free cash flow at just $67.0B — capex is consuming about 63 cents of every operating-cash dollar. Per the September 6 Yahoo Finance report, six hyperscalers are projected to spend $1.3 trillion on capex in 2027 and only one is forecast to have positive free cash flow. When the biggest cloud buyer is plowing nearly two-thirds of its operating cash into data centers and earning a fiscal-2026 revenue growth rate of just 3.3%, the capex program is discretionary and cuttable — exactly the fragility the idea's bearish direction relies on. There is also a margin compression signal. Microsoft's fiscal Q4 (ended June 30, 2026) gross margin slipped to 67.2% from 67.9%, operating margin fell to 45.1% from 46.8%, and net margin to 39.7% from 40.3% — sequential declines consistent with a capital-intensity problem rather than a pricing-power story. That matters for XLU because the logic is circular: if AI spending is squeezing the spenders' margins…

Scores

• Conviction score breakdown: 51
• Thesis support: 65
• Trade readiness: 25
• Risk quality: 55
• Trigger proximity: 55
• Fundamentals trend: 55

Watch items

• XLU — XLU close vs SMA (20)/SMA (50) cross
• XLU — ADX (14)
• XLU — XLU support[2] level
• MSFT — MSFT free cash flow (next XBRL filing)
• MSFT — MSFT insider net open-market activity