Upfront’s Aditi Maliwal makes 3 bets a year and ignores the hype cycle

Upfront Ventures GP Aditi Maliwal says she makes only 2-3 early-stage deals per year and prioritizes backing exceptional founders over writing more checks. Her portfolio spans fintech, AI applications, dev tools, and information markets, including Clair, General Translation, Arcade, and Neon Mobile. The piece is largely a profile of her investment philosophy rather than a market-moving event.

Analysis

The key signal here is not “venture is cautious,” but that capital is increasingly migrating toward businesses with some form of monetizable scarcity: proprietary data, workflow adjacency, and founders who can compress iteration cycles. That tends to favor infrastructure-enabling software and data-collection layers over flashy front-end AI apps, because the latter are the easiest to clone and the hardest to price on durable unit economics. In public markets, this is a subtle headwind for the highest-multiple application names and a relative tailwind for picks-and-shovels vendors with recurring revenue and switching costs.

The more interesting second-order effect is on AI model economics: if the market starts discounting demos and overvaluing data access, the bottleneck shifts from model capability to rights-cleared, time-bounded data pipelines. That can create a wedge for smaller data brokers, privacy-compliant voice/behavioral capture firms, and enterprise software vendors sitting between end users and model labs. For GOOGL, the implication is mixed: it benefits from a stronger data moat and cloud demand, but it also faces pressure to justify spend intensity if investors become more allergic to “AI narratives” without near-term monetization.

The contrarian take is that venture discipline can actually extend the lifespan of the AI bubble by reducing obvious waste and forcing capital into higher-quality assets rather than killing the theme outright. That means the next leg of outperformance is unlikely to be broad beta; it should be concentrated in companies that can prove either exclusive data rights or immediately positive contribution margins. In consumer fintech, the lesson is that distribution alone is no longer enough — lenders and payroll-adjacent products need clear repayment or retention economics, or they will be repriced quickly when funding markets tighten.

Key Decisions for Investors

• Long GOOGL vs a basket of high-multiple AI application names over 3-6 months: if the market rotates from narrative to economics, GOOGL's compute, distribution, and data advantages should outperform higher-burn peers; use a pair to reduce factor beta.
• Initiate a small long position in DB on a 3-6 month horizon if you want exposure to eventual private-market financing activity normalization; the risk/reward improves if capital scarcity forces more advisory, underwriting, and asset-management monetization, but sizing should stay modest given zero direct article linkage.
• Avoid chasing unprofitable AI app stocks for the next 1-2 quarters; instead favor profitable software with AI attach rates and >120% net retention, as the market is likely to penalize burn before it rewards TAM stories.
• Selective long on data-collection / privacy-compliant infrastructure names for 6-12 months; the trade works if AI labs keep paying for proprietary datasets, with upside from re-rating to data-monetization multiples rather than pure SaaS multiples.
• If positioning in fintech, prefer names with demonstrable take-rate/credit discipline over growth-only stories; over the next 2 quarters, weak unit economics should be the main multiple compression catalyst.