High AI Spending is 'Not Temporary' Says Everpure CFO Robbiati

Everpure CFO Tarek Robbiati said AI capex spending is "not temporary," even if supply and demand rebalance as more AI capacity comes online. He added that the AI market may revert toward a mean, but not quickly and likely not to the same level. The comments suggest sustained AI infrastructure investment, but the piece is mostly qualitative and company-specific.

Analysis

The key read-through is that AI capex is shifting from a cyclical burst to a structural budget line item, which should support a longer runway for the entire compute stack. The market still tends to price AI spend as a near-term wave, but if hyperscalers and enterprise buyers keep treating it as strategic infrastructure, the beneficiaries are the “shovels” businesses with recurring demand, not the application names where monetization is still lagging.

Second-order, this is bullish for vendors with constrained supply and pricing power: advanced semicap equipment, networking, power, cooling, and high-bandwidth storage. The biggest risk is not demand disappearing, but supply normalizing faster than many expect, which would compress margins for lower-quality component suppliers even as unit volumes stay healthy. That creates a barbell outcome: durable winners at the top of the stack, weaker economics for commoditized hardware and any name relying on scarcity pricing.

The contrarian miss is that the debate may be moving away from “is AI spend temporary?” to “who captures the budget?” If capex persists but ROI remains uneven, boards will start demanding cost discipline and consolidation, which could slow spending growth by late 2025 even if the absolute dollar level stays elevated. That means the trade is not a blanket long AI; it is a selective long in infrastructure enablers and a short in names whose valuation already embeds perpetual acceleration.

Catalyst path: a few quarters of sustained hyperscaler capex and firm guidance from adjacent suppliers will validate the thesis, while any pause in cloud spending or evidence of underutilization would trigger a sharp factor rotation. The most vulnerable window for crowded AI longs is 3-9 months out, when investors realize revenue conversion is slower than capex growth and begin discriminating on free cash flow rather than narrative.

Key Decisions for Investors

• Overweight AI infrastructure beneficiaries versus application-layer AI names for the next 3-6 months: long NVDA/ANET/AVGO basket, funded by a short in a cash-burning AI software basket where revenue monetization is still uncertain. Risk/reward favors the infrastructure side if capex remains elevated, with 15-25% upside on a continued budget cycle and 10-15% downside if spending pauses.
• Use pullbacks to add long exposure to power/cooling data-center enablers, especially names with multi-year order visibility. The thesis is that rising AI compute intensity is a secular demand driver even if chip supply loosens; target positions with 12-month horizons and prefer companies with >50% recurring/service revenue.
• Avoid chasing the most crowded AI beta names after strength; instead sell call spreads or write covered calls on high-multiple AI software leaders into capex-confirmation rallies. If spending persists but monetization lags, these names are most exposed to multiple compression over the next 2-4 quarters.
• Pair trade idea: long semiconductor equipment/networking infrastructure vs. short lower-quality storage or component suppliers that depend on scarcity pricing. If the AI supply/demand balance reverts, the former can still grow while the latter faces margin normalization; target a 2:1 reward/risk over 6 months.