Insight Partners' Deven Parekh makes the contrarian case for portfolio diversification at a moment when much of the VC industry is concentrating heavily into OpenAI and Anthropic. The interview surfaces real tension in the market about whether piling into a small number of frontier AI companies is prudent strategy or dangerous groupthink. It's a rare public signal from a major firm that concentration risk in AI is being taken seriously at the partnership level.

Sequoia is leading a deal that would value Mecka AI at nearly $500M, driven by surging investor demand for robot training data infrastructure. The round reflects how capital is flowing not just into frontier models but into the picks-and-shovels layer powering physical AI systems. It's another data point showing valuations remain aggressive in AI-adjacent infrastructure plays.

Mistral AI has closed a $3.5B round at a $24B valuation, setting another record for European AI fundraising. The deal underscores that the mega-round dynamic isn't limited to US frontier labs — global investors are willing to write enormous checks for credible open-weight model competitors. The valuation also signals continued investor appetite for alternatives to the OpenAI/Anthropic duopoly.

August saw 29 new unicorns minted, with AI software and semiconductor companies leading the cohort — a sign that private market valuations remain buoyant despite macro uncertainty. The breadth of sectors represented, including robotics, suggests the valuation boom is spreading beyond pure-play LLM companies. For investors, the data raises the perennial question of whether the pace of unicorn creation is sustainable.

This piece from Mighty Capital's Sophie Moatti argues that in the AI era, only two types of competitive moats reliably hold up for startups — a pointed counterargument to the assumption that any AI product can sustain defensibility. The framework is directly relevant to how investors are underwriting new deals, given how quickly AI capabilities commoditize. It's the kind of analytical lens VCs are increasingly applying as they sort winners from fast-followers.