AI Infrastructure Thesis Faces Historic Valuation Reckoning: $2.7 Trillion June Selloff Exposes Capex-Monetization Gap

A historic $2.7 trillion market-value erasure in June 2026 across the Magnificent Seven, Broadcom, and Oracle has crystallized a structural risk to the AI infrastructure thesis: hyperscalers are raising capex forecasts faster than they can demonstrate AI monetization, creating a valuation-reality mismatch that is now driving broad semiconductor and tech selloffs.

Published 2026-06-29 Updated 2026-06-29

What changed

The AI infrastructure thesis has encountered a sharp and broad-based valuation reset in late June 2026. The Magnificent Seven plus Broadcom and Oracle collectively lost $2.7 trillion in market value during the month, according to reporting on the historic de-risking event. This selloff was accompanied by a cascade of negative developments:

Goldman Sachs warnings on positioning and crowding: Goldman strategists flagged that AI trade positioning levels are "flashing red" and warned that investor assumptions about the AI trade are "starting to stretch reality." The firm noted that hyperscalers continue raising capex forecasts while AI monetization may not keep pace, describing the AI market as a "rubber band" stretched to its limits.
Oracle's capex shock and job cuts: Oracle cut approximately 21,000 jobs over the past year as AI reshaped its operations, signaling internal restructuring amid capex pressures. Oracle stock fell 34.2% over the 30 days ending June 29, 2026, the steepest decline among the thesis's anchor names.
Semiconductor supply-chain fragility: Beyond Nvidia, the selloff extended across the semiconductor ecosystem. Qualcomm fell 8% on news of ByteDance's custom ASIC deal, while Marvell fell 10%, indicating that custom-silicon competition is fragmenting the supply chain. Broadcom fell 12.59% post-earnings, signaling that AI semiconductor valuations were priced for perfection.
OpenAI's in-house chip development: Nvidia once explored a $100 billion chip deal with OpenAI, but OpenAI is now building its own chips from scratch, introducing direct competition to Nvidia's moat and raising questions about whether hyperscalers will internalize chip production rather than rely on external suppliers.
Broad tech selloff across geographies: Global chip stocks tumbled as the selloff swept from Asia (where South Korea's KOSPI fell 10%) through Europe toward Wall Street. Memory-chip makers including Micron, SK Hynix, and Samsung all declined sharply, signaling sector-wide reassessment.

Why it matters

Supply-chain fragmentation and internalization: OpenAI's shift to in-house chip development and ByteDance's custom ASIC deal signal that hyperscalers may be reducing reliance on external semiconductor suppliers rather than increasing it. If major AI workload owners build their own silicon, Nvidia's role shifts from sole supplier to one of many, and the entire semiconductor supply chain (Broadcom, Qualcomm, Marvell) faces demand destruction. This fractures the thesis's assumption that Nvidia and its partners will capture the full value of the AI infrastructure wave.

Grid and infrastructure constraints: Earlier warnings that America's electrical grid is "so far behind" that blackouts are coming even without AI infrastructure buildout introduce a hard constraint on the pace of data center expansion. If power availability becomes the limiting factor rather than capex appetite, the buildout thesis faces a structural ceiling that is independent of hyperscaler spending decisions.

Opposing sources and risks

Multiple sources directly contradict the thesis's upward direction:

OpenAI chip internalization (moderate certainty): The shift from a potential $100 billion Nvidia deal to in-house chip development directly undermines the thesis's assumption that external semiconductor suppliers will capture the full value of AI infrastructure. If major workload owners build their own silicon, the addressable market for Nvidia, Broadcom, and Qualcomm contracts.
Oracle's job cuts and restructuring (moderate certainty): The 21,000-person reduction signals that Oracle is not simply scaling to meet demand but is restructuring in response to AI-driven operational changes. This suggests internal uncertainty about the pace and profitability of the buildout.
Semiconductor supply-chain fragmentation (fairly high certainty): Qualcomm's 8% fall on ByteDance's custom ASIC deal and Marvell's 10% decline indicate that custom silicon is fragmenting the supply chain. If hyperscalers and large AI workload owners build their own chips, the thesis's assumption that a unified semiconductor supply chain will benefit from AI infrastructure buildout is invalidated.

What to watch

Hyperscaler Q3 2026 earnings and capex guidance: Microsoft, Amazon, Google, and Meta will report third-quarter results and update full-year capex forecasts. If guidance is flat or reduced from prior expectations, it will confirm that the capex-monetization mismatch is real and that the buildout is slowing. Conversely, if capex guidance is raised again despite June's selloff, it will signal that hyperscalers remain committed to the buildout despite valuation pressure.
Nvidia's next earnings and customer concentration: Nvidia will report Q3 results and provide guidance on customer concentration and custom-chip competition. If Nvidia signals that OpenAI, ByteDance, or other major customers are reducing orders in favor of in-house silicon, it will confirm that supply-chain fragmentation is accelerating and that Nvidia's moat is eroding.
Oracle's capex and debt sustainability: Oracle's next earnings report will clarify whether the 21,000 job cuts are a one-time restructuring or the beginning of a sustained capex pullback. If Oracle's capex guidance is reduced, it will signal that even the companies most bullish on AI infrastructure are reassessing the buildout's pace.
Broadcom and semiconductor supply-chain health: Broadcom's next earnings report will indicate whether the June decline was a temporary repricing or the start of a sustained contraction in AI semiconductor demand. If Broadcom guides lower on data center and AI infrastructure revenue, it will confirm that the buildout is slowing.
US electrical grid capacity and power availability: Ongoing reporting on grid constraints and power availability for data centers will clarify whether infrastructure buildout is limited by capex appetite or by hard constraints on power supply. If grid capacity becomes the binding constraint, the buildout thesis faces a structural ceiling.
Custom silicon adoption by hyperscalers: Continued reporting on OpenAI's chip development, ByteDance's ASIC strategy, and other hyperscalers' in-house silicon efforts will indicate whether supply-chain fragmentation is accelerating. If major workload owners shift to custom silicon, the thesis's assumption that external semiconductor suppliers will capture the full value of AI infrastructure is invalidated.

Related Arbora context

Megacap tech AI monetization and valuation divergence (concept-megacap-tech-ai-monetization): This thesis tracks divergence among megacap tech names on AI monetization credibility. The June selloff and Goldman's capex-monetization mismatch warnings directly support the view that AI monetization is uncertain and that valuations are diverging based on credibility of AI revenue realization.
AI model export controls and sovereign AI access risk (concept-ai-model-export-controls-sovereign-ai-access-risk): Export control risk on frontier AI models introduces a regulatory ceiling on AI-as-a-service revenue projections for cloud providers. If hyperscalers cannot monetize AI models internationally due to export restrictions, the capex-monetization mismatch widens further.
Micron memory chip supercycle — AI-driven DRAM demand (concept-micron-memory-chip-supercycle-ai-dram): Micron's sharp June decline alongside other memory-chip makers suggests that the memory supercycle thesis is also facing repricing. If AI infrastructure buildout slows, memory demand will contract in parallel.

Sources

This article is research notes and not financial advice.