AI Has Become Major Capital Formation Cycle, Says Altimeter

Apoorv Agrawal said AI has become one of the largest capital formation cycles, underscoring continued investor appetite for the sector. He also discussed the IPO backdrop as SpaceX and Anthropic move closer to public listings. The article is largely commentary from the Bloomberg Tech event and does not include new deal terms, valuations, or timing specifics.

Analysis

The important signal is not that AI is large, but that it is becoming a capital allocation regime for private markets: once a narrative can absorb billions of growth equity with a credible IPO path, it starts to reprice adjacent assets across compute, networking, power, and data-center infrastructure. That creates a second-order winner set beyond the obvious model companies: picks-and-shovels providers with hard constraints on deployment capacity should see demand stay elevated even if headline AI software multiples compress.

If the IPO window reopens for category leaders, the market will likely reward revenue scale and strategic scarcity more than present-day profitability. That is bullish for late-stage private assets with dominant distribution, but it is also a warning sign for every near-peer competitor still funding at similar burn rates: the public markets tend to force a ranking exercise, and lower-quality names can be marked down sharply over 1-2 quarters once a few bellwethers price successfully.

The main risk is that the cycle remains sentiment-led longer than fundamentals justify. A rough patch in post-IPO trading for the first wave of AI listings would quickly tighten private capital, extend fundraising timelines, and force valuation resets across the venture stack; that reversal would likely show up first in secondary-market discounts and only later in primary rounds. The contrarian view is that the market may be underestimating how much of the AI spend is infrastructure-capex rather than software-margin creation, which means the near-term economic benefits accrue disproportionately to suppliers rather than to the highest-multiple application layer.

Key Decisions for Investors

• Bias long the AI infrastructure complex for the next 3-6 months: add exposure to NVDA, AVGO, ANET, and VRT on pullbacks, because capital formation in AI translates first into orders, not earnings, and the revenue lag is typically 2-4 quarters.
• Avoid chasing late-stage private AI app names into rumored IPO windows; use any public-market enthusiasm to fade weaker comps via basket shorts in unprofitable SaaS/software names with similar growth but lower differentiation.
• Pair trade: long AI infra / short high-multiple, low-cash-flow software (e.g., NVDA or ANET vs. a basket of unprofitable cloud/software names) to express the view that capex beneficiaries outperform if IPO appetite lifts the whole theme.
• If you want IPO optionality, prefer smaller starter positions in likely AI listing candidates only after first-day pricing and first 30-day lock-up dynamics are visible; the trade is better in weeks 2-8 than at the offer when retail enthusiasm is highest.
• Watch secondary and primary venture pricing for a 60-90 day tightening cycle; if discounts widen or late-stage rounds reprice lower, cut cyclical AI-beta exposure quickly because that is usually the first sign the capital formation cycle is cooling.