Descriptions:
Bloomberg Technology speaks with IVP venture partner Wilhelm about the current AI investment landscape and what patterns distinguish durable businesses from hype. Wilhelm’s thesis centers on usage-compounding companies — those whose revenue scales with consumption rather than seat count — naming Perplexity, Clickhouse, LangChain, Base10, and Cribble as portfolio examples that fit the pattern. He traces his interest in Perplexity back to September 2023, when it had $2 million ARR, and argues the mistake is treating it as only a search product rather than a platform that has already shipped a browser (Comet), deep research, and a computer agent.
The interview covers Anthropic’s acquisition of Descartes, which Wilhelm interprets primarily as an inference optimization play — software to extract more performance from existing chips — rather than a world-model capability move. He positions inference optimization as one of the hottest layers in the current AI stack, drawing an analogy to his early career selling Oracle databases across Chevron, Intuit, and Marmot, where infrastructure companies with broad TAMs compounded quietly over decades.
On the bubble question, Wilhelm distinguishes between a potential capital markets adjustment and the underlying technology trend. He argues both can be true simultaneously: we are unequivocally in a golden age of technology, but timing mismatches between investment and productivity realization are normal. He references Robert Solow’s 1987 productivity paradox observation and suggests we are still in an individual productivity phase of AI adoption, meaning enterprise-level GDP impact may not yet be visible in the numbers.
📺 Source: Bloomberg Tech · Published August 13, 2026
🏷️ Format: Interview
