AI's Best Quarter Since 2020 Splits the Buildout From the Bubble
As the Supreme Court shields the Fed but politicizes the SEC and CFTC, hard private capital keeps underwriting data-center power even where megacap-capex ROI is in doubt.
The macro regime is unchanged, and the honest read is that most of today’s batch is stale news already digested (April–June oil/Iran events, the June 17 FOMC, May retail sales, mid-June claims) rehashed under fresh datelines. The regime remains: a higher discount rate (30Y near multi-year highs, 2Y at 4.10%, December-hike ~52% on Kalshi) sitting on a deflating energy-inflation impulse (5Y breakeven 2.26, core PCE undershoot to 3.4%), with the AI trade bifurcated between a confirmed infrastructure/chip layer and an ROI-doubted megacap-capex layer, and a private-credit cascade that has stratified without converting to spreads.
Three things are genuinely new or sharpened. First, the Supreme Court’s split rulings: keeping Cook at the Fed (preserves monetary-policy credibility) while stripping SEC/CFTC commissioners of removal protection (politicizes securities/commodities regulators, a two-way overhang for CME’s CFTC litigation and a marginal crypto-regulatory tailwind). Second, the AI-power-capital cluster deepened materially in a single day (Bloom-Brookfield 5x to $25B, KKR-EDF $4.2B, Schneider-Cognite $3.1B), the strongest infrastructure-layer-breadth confirmation yet against the chip-layer ROI doubt. Third, cybersecurity graduated to a confirmed durable-AI-demand subsegment (PANW and CRWD both record quarters on AI-agent identity security), validating both legs of the PANW-vs-CRM pair. The quarter closed with the S&P/Nasdaq posting their best quarter since 2020 on AI, with breadth broadening to a Dow record even as the Magnificent 7 lagged.
The single most important open item is unchanged: whether the megacap-capex-ROI doubt (the ~$720B 2026 AI spend, OpenAI IPO delayed to 2027, the efficiency-budget shift) reaches the confirmed chip layer at the late-July NVDA/MSFT capex guide. Nothing today resolved it.
Supreme Court Splits: Fed Protected, SEC/CFTC Politicized
The Court blocked Cook’s firing and, in a separate ruling, held the President can remove SEC and CFTC commissioners at will, overturning the 91-year Humphrey’s Executor framework for those agencies. The bifurcation matters: it establishes that the Fed’s structural independence is treated as constitutionally distinct, which reduces the tail of a captured FOMC forced to cut against 4.2% CPI, marginally supporting the long end’s inflation-credibility premium under Warsh.
The under-priced leg is the SEC/CFTC ruling. A directly-removable CFTC injects two-way risk into CME’s ongoing litigation against the agency over perpetual futures, the dominant regulatory overhang on that name; a politically-aligned commission could resolve it either direction. A more compliant SEC raises the odds the semiannual-reporting proposal (Form 10-S) advances, which would reduce disclosure frequency and is a structural negative for the disclosure-monetizing data franchises (MCO, SPGI, MSCI, ICE data). It also tilts crypto-regulatory posture favorable, consistent with the CLARITY Act (advanced 15-9 in Senate Banking) and Trump’s disclosed $580M crypto income, a marginal positive for COIN and HOOD. This is a governance/regulatory-structure shift, not a discrete market mover; weight it as a slow-burn.
AI-Power Capital Cluster Deepens in a Single Day
Three large power/industrial-AI transactions landed together: Bloom Energy–Brookfield expanded their AI-infrastructure partnership 5x to $25B, KKR agreed to buy EDF’s North American power operations for $4.2B, and Schneider Electric agreed to buy industrial-AI firm Cognite for $3.1B. This is the strongest single-day infrastructure-layer-breadth confirmation to date. The mechanism: data-center power is being underwritten with hard private capital (on-site generation, grid infrastructure, industrial-AI software) at the same time megacap-capex multiples are compressing on ROI doubt. The physical buildout is not decelerating even where the chip-layer valuation is under scrutiny.
This reinforces the GEV/VST/CEG power-generation overweight and the ETN electrical-equipment picks-and-shovels read, and sharpens the GEV-vs-ORCL pair: bankable contracted power demand (GEV sold-out gas capacity through 2030, RPO $163.3B) against leveraged fragility (Oracle’s worst week since 2001, ~$130B debt, negative FCF funding the same buildout). One caveat: higher-for-longer rates raise the financing cost of these negative-FCF/capital-intensive projects, a partial offset that favors contracted/merchant names (VST/GEV) over dilutive regulated names (AEP/AEE).
Cybersecurity Confirmed as a Durable AI-Demand Subsegment
Palo Alto and CrowdStrike both posted record quarters driven by identity-security demand as AI agents proliferate. Two independent record prints establish this as a confirmed within-AI theme, distinct from the capex-ROI doubt hitting megacap spenders and distinct from the application-SaaS displacement (CRM, WDAY, NOW). The causal chain: AI-agent proliferation multiplies machine identities and attack surface, which drives identity-security spend. This is a rare application-adjacent segment where AI adoption is net-additive to demand rather than a displacement threat. It validates both legs of the PANW-vs-CRM pair and is reinforced by Accenture’s $4.18B cybersecurity bookings sitting inside an otherwise weak consulting print. PANW’s CyberArk ~$21B pivot (integration and ~$29B goodwill-impairment risk) keeps conviction disciplined despite the record quarter.
Developing Themes
AI Infrastructure vs Megacap-Capex Bifurcation: Sharpened, Not Resolved
The chip layer confirmed hard this quarter: Micron’s blowout, the $2T combined gain in MU/INTC/AMD, and the semiconductor sector’s best quarter ever. Nvidia specifically lagged, which the data supports reading as idiosyncratic (custom-silicon/ASIC share shift, China zeroed, circular-financing web) rather than a demand crack, and GOOG’s homegrown-TPU advantage is the concrete competitive vector behind it. Meanwhile the megacap capex is being repriced on ROI doubt (AP-sourced, tier 1), and the FT study finds heavy AI-spenders are hiring faster than peers (cutting against broad AI-job-loss fears but not resolving the return question). The late-July NVDA/MSFT capex guide remains the decisive demand-economics arbiter. Hold NVDA/TSM/GOOG through the whipsaw per the do-not-flip-on-one-tape-day discipline; maintain ACN/WDAY/CRM/NOW short legs.
Oil: Floored Range Firms, Failure Tail More Violent
Crude posted its largest quarterly drop in six years (above $70) as Hormuz workarounds, weak Chinese imports, and record UAE post-OPEC-exit exports stacked on the supply side. UAE’s OPEC exit signals weakening cohesion, a structural ceiling-cap independent of Hormuz. The SPR at its lowest since 1983 removes the government buffer, so any Hormuz re-closure fires into depleted inventories, making the failure tail more violent even as spot eases. Iran refused to meet US envoys and canceled technical talks (Kushner/Witkoff to Doha with no confirmed meeting), the disconfirming counter-signal that, per the 0-for-N discipline, justifies not chasing the de-escalation. Refiners (MPC, VLO, PSX) benefit from tight domestic diesel/gasoline into summer; tankers (STNG, INSW) lean bearish on ton-mile normalization; LNG most insulated.
Private Credit: Stratification Is the Early Asset-Side Tell
The new data point is bond investors assigning wider spreads to smaller private-credit lenders than to larger funds. This is the beginning of asset-side differentiation in the liability→asset→spread sequence, and it favors scale/committed-capital managers (APO/ARES) over evergreen/redemption-exposed vehicles. But HY spread is at 2.80% (FRED, -0.03), near cyclical tights with IG access wide open, so there is still no conversion. HLNE’s committed-capital fee base remains insulated from the stratification repricing the complex. The first HYG spread move off 2.80% is the conversion tell; the contango term structure (5.2% near vs 10.3% far, OI P/C 3.91) says near-term calm with H2 stress priced.
Continuing Themes
Rates/duration: Higher-for-longer reinforced by the upward dot-plot revision (3.6-4.1% year-end) and the Williams/Hammack hawkish chorus. Do not pre-position into the reflexive, guidance-free Fed; let CME/CBOE/ICE carry the data-print vol. Bearish duration; TLT residual call OI is stranded dovish positioning.
Labor: Low-fire/low-hire mix intact — claims 215K (multi-year low, FRED) against continuing claims 1.821M (rising). Kalshi traders give under 60% odds of >100K on Friday’s payroll vs >118K consensus; a weak print would test the reflexive reaction function. Consumer job-finding confidence at a 2021 low despite the 7.6M JOLTS strength is the data-vs-sentiment gap.
Consumer cliff: Two-phase read intact. 30Y mortgage ~6.5% keeps housing subdued through 2027 (Reuters poll); Lennar selling homes at 2017-low prices confirms builder margin pressure; housing starts -8.7% YoY against existing-home-sales +3.2% YoY (the two-phase pattern). Nike’s low-quality tariff-refund beat and 12% China drop confirm discretionary caution. Consumer Discretionary BUY-prohibited. PGR over COF/SYF.
De-dollarization / China Shock 2.0: China’s manufacturing-over-consumption commitment intensifies global goods disinflation (Pimco export-glut EM-bond case), capping tradable-goods inflation against the energy impulse. World Bank China-lending phase-out escalates decoupling. Structural long-end pressure cluster intact.
Healthcare M&A / GLP-1: Patent-cliff consolidation continues; FDA PreCheck for LLY/REGN a marginal capacity tailwind; Anthropic Claude Science a long-term CRO/pharma-services disruption watch (IQV, ICLR, CRL). GLP-1 incumbents (LLY, NVO) get consistency support from the protein-demand-surge.
Defense: Multi-front demand structurally confirmed (US trillion-dollar budget, European rearmament sustaining 195,000 US jobs, UK plan). Munitions-replenishment favors RTX/LHX over the AVOID LMT — re-examine the LMT-vs-ACN pair, whose long leg conflicts with this read.
Crypto: CLARITY Act advanced 15-9; SEC/CFTC politicization marginally crypto-favorable; MiCAR pushing Binance and hundreds of platforms out of Europe is a consolidation catalyst. Bitcoin-ETF resilience being tested in the selloff. Low portfolio relevance.
The options tape confirms the regime rather than contradicting it: QQQ’s -4.3% 12-month put skew held essentially unchanged, keeping longer-dated downside protection into the late-July NVDA print intact — the cleanest signal the market treats AI-capex-ROI as its held tail. IWM retains the highest US-equity OI P/C at 2.34 on the small-cap-downside story, while HYG sits in contango (5.2% near vs 10.3% far, OI P/C 3.91) pricing near-term calm against H2 cascade stress. The premium section maps how to position around the late-July capex guide, where the contracted-power longs and application-layer shorts sit, and which of the seven risk scenarios — from the chip-layer ROI crack to a Hormuz re-closure into a 1983-low SPR — carry the most leverage. Full options positioning analysis, portfolio playbook, and risk scenario framework below for subscribers.
This publication is for informational and educational purposes only and does not constitute financial, investment, or trading advice. The analysis, opinions, and commentary presented here should not be interpreted as a recommendation to buy, sell, or hold any security. Always conduct your own research and consult a qualified financial advisor before making investment decisions. Past performance does not guarantee future results.
