For decades, the unwritten rule was sacred: a top-tier venture firm does not invest in two companies trying to kill each other. Benchmark, the partnership that backed eBay, Uber, and Instagram, built its reputation partly on this covenant — founders trusted that their board member wouldn't be whispering strategy to a rival down Sand Hill Road. That era, the firm now concedes, is functionally over.
In conversations with founders and limited partners over the past quarter, Benchmark's general partners have acknowledged what the AI startup ecosystem already suspected: the firm is actively backing companies with overlapping ambitions in foundation models, AI infrastructure, and applied intelligence. The justification is pragmatic rather than philosophical. When every promising company is building on the same transformer architectures, pursuing the same enterprise customers, and competing for the same GPU allocations, the traditional definition of "competitor" becomes meaningless.
The math that broke the covenant
Venture's non-compete norm emerged in an era when markets were large enough to support one dominant winner per category. Search had Google; social had Facebook; ride-hailing had Uber (eventually). A firm could pick its horse and ride. AI presents a different topology. The technology is horizontal, touching every industry simultaneously, and the winning architectures are not yet settled. A model company today might pivot to agents tomorrow; an infrastructure play might release its own foundation model next quarter. Benchmark's partners reportedly told one founder that refusing to back apparent competitors would mean sitting out the entire cycle.
What founders lose
The erosion carries real costs. Founders historically paid venture's steep economics — surrendering board seats, accepting valuation discipline — partly because they were buying exclusivity. A Benchmark partner on your board meant Benchmark's network, reputation, and deal flow were aligned with your success, not hedged against it. If that alignment is now partial, the implicit contract has changed. Some founders are responding by favouring less prestigious firms that will still offer genuine exclusivity, or by taking capital from strategics and sovereign funds that have no competing portfolio.
Our take
Benchmark is not villainous for adapting; it is simply being honest about incentives that other firms quietly face. But the admission matters because it names what the industry has been dancing around: AI has made venture capital more like public-market investing, where portfolio diversification trumps loyalty. Founders should price this into their cap tables accordingly. The old handshake meant something. The new one means less.
