Will aggregate Big 4 hyperscaler CapEx growth decelerate below 15% YoY by Q2 CY2026?

Base rate analysis: The Big 4 hyperscaler CapEx aggregate grew well above 15% YoY through CY2025, with all four companies announcing expanded AI spending plans. For aggregate growth to decelerate below 15% by Q2 CY2026 (April-June 2026), the combined CapEx would need to drop dramatically in the next ~4 months. The base rate for such a rapid reversal without any leading indicator is extremely low. Prior CapEx pauses (2022-2023) were preceded by quarters of slowing guidance and management commentary about 'digestion periods.' Currently, the opposite is happening — every hyperscaler is accelerating commitments. Meta raised its CY2025 CapEx guidance multiple times. Microsoft committed to $80B+ in AI infrastructure. Amazon is accelerating AWS CapEx. Google expanded AI infrastructure spending. None have signaled any pullback. The Marvell Q4 earnings call (today) confirms hyperscaler CapEx expectations have 'continued to increase' since December, and bookings are at record pace. The resolution window is now only 4 months away with zero deceleration signals.

Structural analysis of CapEx commitment mechanics: Hyperscaler CapEx is not discretionary spending that can be turned off quickly. Data center construction has 18-36 month lead times. Equipment orders (GPUs from NVIDIA, custom ASICs from Marvell/Broadcom, networking gear) are placed 6-12 months in advance. The supply chain is locked in — Marvell's COO confirmed supply secured 'this year, next year, and beyond,' which implies hyperscaler purchase commitments extending through CY2027. For Q1 or Q2 CY2026 CapEx to fall below 15% YoY growth, hyperscalers would need to have already canceled or deferred orders that are in the pipeline NOW. There is zero evidence of this. In fact, the structural commitments make a sub-15% scenario nearly impossible within the 4-month window. The only scenario would be a sudden macro crisis (severe recession, financial crisis) that forces emergency spending cuts — but even then, contractual commitments and construction-in-progress would sustain elevated CapEx for at least 2-3 quarters. I assign 4% to account for extreme tail risks like a major geopolitical event or financial crisis.

Scenario analysis across possible outcomes: Scenario A (70% likely): All Big 4 maintain 25-50% YoY CapEx growth through Q2 CY2026. This is the current trajectory and the base case. Scenario B (20% likely): One or two hyperscalers moderate growth to 15-25% range, but aggregate remains above 15% because others continue accelerating. This is the 'partial moderation' case. Scenario C (7% likely): Macro shock causes broad deceleration, but aggregate still stays above 15% due to contractual commitments and construction pipeline. Scenario D (3% likely): Extreme event causes aggregate to fall below 15% — would require simultaneous emergency cuts across all four, overriding existing contracts and construction-in-progress. Only Scenario D resolves YES. The new Q4 MRVL earnings evidence makes even Scenario B less likely — DCI modules shipping to ALL 5 major hyperscalers, bookings at record pace, CapEx expectations increased since December. The evidence strongly clusters around Scenario A.

The question asks about aggregate Big 4 CapEx growth falling below 15% YoY by Q2 CY2026. Since the previous prediction at 6%, we now have the Q4 FY2026 Marvell earnings providing a direct window into hyperscaler spending intentions. The evidence is unambiguous: CapEx expectations have increased, not decreased. Marvell raised its FY2027 revenue target by $1B in just 3 months — management would not do this if they saw any CapEx softness from their largest customers. The raise of DC growth guidance from >25% to +40% is particularly telling — this is a company with direct visibility into hyperscaler purchase orders for custom silicon. Furthermore, the resolution date is August 15, 2026, covering Q1 and Q2 CY2026. Q1 CY2026 is essentially already over (ends March 31). If there were a CapEx shortfall in Q1, we would expect some signal by now. For Q2 CY2026 (April-June), spending plans are already locked. I reduce from the prior 6% to 4%.

I want to steelman the YES case before assigning a final probability. What could cause Big 4 aggregate CapEx to fall below 15% YoY growth by Q2 CY2026? (1) A sudden AI winter triggered by a high-profile AI failure or regulatory shock — possible but would take quarters to flow through to CapEx. (2) A financial crisis or severe recession — the US economy shows no imminent signs. (3) Major geopolitical event disrupting supply chains or trade (e.g., Taiwan conflict) — this would actually cause CapEx to spike, not decline, as hyperscalers rush to secure inventory. (4) Coordinated pullback driven by disappointing AI monetization — possible, but hyperscaler commentary in Q4 CY2025 earnings (January 2026) showed the opposite: increased commitment. (5) Accounting reclassification — technically possible but would affect reporting, not actual spending. None of these scenarios are both likely AND fast enough to materialize in 4 months. However, I assign slightly higher probability (5%) than some models because I want to account for unknown unknowns — events we cannot foresee by definition. The resolution window is short but not zero.

Looking at the actual CapEx numbers: In CY2025, Big 4 aggregate CapEx was growing at approximately 40-60% YoY. For aggregate growth to decelerate below 15% by Q2 CY2026, total CapEx would need to be less than 115% of Q2 CY2025 levels. Given that Q2 CY2025 already represented significantly elevated spending, this would imply either flat or modestly growing absolute CapEx — essentially meaning hyperscalers would need to almost freeze their AI infrastructure buildout. This contradicts everything we know: (a) NVIDIA continues to report record data center revenue, (b) all hyperscalers raised CapEx guidance in their most recent earnings, (c) Marvell's record bookings and raised guidance reflect downstream effects of continued CapEx acceleration, (d) DCI modules shipping to all 5 major hyperscalers shows breadth of spending, not concentration. The prior 6% was already low; the new evidence from today's earnings reduces it further to approximately 4%.

All Big 4 hyperscalers accelerating CapEx. Marvell's Q4 earnings today confirm CapEx expectations increased further. Bookings at record pace. Revenue target raised $1B. Only 4 months left in resolution window. No leading indicators of deceleration. Prior prediction was 6% — new evidence pushes lower to 3%.

While the evidence strongly points to NO, I keep 5% for tail risk scenarios: unexpected macro shock, sudden AI sentiment reversal, or unknown unknowns. The 2022-2023 CapEx pause precedent shows these cycles CAN turn. However, that pause was preceded by quarters of decelerating guidance — there are no such signals now. The short 4-month window is the strongest constraint against YES. Even if sentiment shifted tomorrow, CapEx commitments and construction pipelines would sustain growth above 15% through Q2 CY2026.

Math check: Big 4 aggregate CapEx growing ~40-60% YoY. To hit <15%, CapEx would need to nearly freeze. All 5 major hyperscalers actively building DCI with Marvell. Supply chain locked multi-year. FY2027 target raised from $10B to $11B. DC growth guide raised to 40%. Interconnect guide raised to >50%. Every indicator points to acceleration, not deceleration. Resolution window closes in 4 months. This is about as close to a certainty-NO as markets get.