AI could take capex to $1Tn in CY27 By Investing.com

BofA sees global hyperscale AI capex reaching $1 trillion by 2027, with combined spending forecast above $800 billion in 2026, up 67% year over year. Microsoft lifted 2026 capex guidance to $190 billion, Alphabet to $185 billion, Meta to $135 billion, and Amazon kept $200 billion, underscoring accelerating AI demand and persistent hardware supply tightness. The report is constructive for Nvidia, memory, optics, semicaps, and power semiconductor suppliers as hyperscalers continue to absorb higher component costs.

Analysis

The main second-order effect is that hyperscaler capex is no longer just a semiconductor revenue story; it is becoming a balance-sheet and working-capital story for the entire AI stack. As the largest buyers absorb rising component costs, pricing power migrates upstream to memory, substrates, optics, and power-management vendors, while downstream cloud margins remain the pressure valve. That should keep the market rewarding the picks-and-shovels names, but it also raises the bar for “AI monetization” proof from the platforms because capex intensity is now scaling faster than visible cash conversion.

The more interesting read-through is that supply constraints through 2026 reduce the likelihood of a near-term capex air pocket, which means this is a multi-quarter earnings revision cycle rather than a one-quarter hype burst. The risk is not demand disappearing; it is a delayed ROI debate that hits valuation multiples first, especially for the hyperscalers with the richest expectations. If enterprise AI spend slows or incremental token growth starts cannibalizing lower-margin core products, the market can punish platform names even while the vendor ecosystem stays bid.

NVDA is still the clearest beneficiary, but the asymmetry is better in the enabling layers where sentiment is less crowded and pricing is stickier: memory, interconnect, photonics, and power semis. The biggest contrarian point is that consensus may be underestimating how much of this capex ends up being prepaid capacity rather than immediate revenue, which creates a 6-12 month digestion period after the initial order surge. That favors relative-value longs in the supply chain versus outright longs in the megacaps.

For the platforms, the hidden risk is that capex guidance itself becomes a ceiling on multiple expansion if investors decide that incremental AI dollars are earning sub-hurdle returns. In that scenario, the market can rotate from 'AI growth' to 'capital discipline' quickly, particularly if broader rates back up or cloud growth normalizes. The catalyst to watch is the next two earnings seasons: if revenue growth re-accelerates with capex, the trade extends; if not, the de-rating starts well before any actual demand rollback.

Key Decisions for Investors

• Long NVDA against a basket short of high-valuation megacap AI beneficiaries (e.g., GOOGL/MSFT/META) on a 3-6 month horizon; the pair captures continued capex upside while hedging multiple compression risk if ROI scrutiny rises.
• Add a tactical long in AI infrastructure suppliers with tighter supply curves than GPUs, funded by a short in the hyperscaler basket; best risk/reward is in names with pricing power in memory, optics, and power semis over the next 2 quarters.
• Use any post-earnings pullback in MSFT or GOOGL to buy 3-6 month call spreads rather than stock; upside remains intact, but the market is now more likely to demand proof of monetization before paying a higher multiple.
• Avoid chasing AMZN/META on capex optimism alone; instead, wait for evidence that AI workload growth is translating into operating leverage. If that does not appear by the next 1-2 prints, consider shorting the weakest-margin platform on rallies.
• For a cleaner relative-value expression, go long semiconductor equipment/power exposure versus the Nasdaq 100 over 1-2 quarters; the capex cycle should support the supply chain even if broad tech multiple expansion stalls.