Tuesday, 2 June 2026

The venture funding landscape is reaching a point where the numbers have stopped meaning much. Anthropic just raised $65 billion at a $965 billion valuation and filed confidentially for an IPO in the same breath. Google is hunting $80 billion for data centers while Berkshire Hathaway commits $10 billion to the same AI buildout race. These are sums that dwarf most national budgets, yet we're meant to parse them as routine quarterly milestones. I think what matters more than the headline figure is what it reveals about where the money actually flows.

The Anthropic filing is interesting precisely because it signals that the private market for frontier AI labs may be approaching a ceiling. When your valuation hits nearly a trillion dollars, there are fewer and fewer buyers willing to write the next check. Going public at this scale is less about capital raise and more about finding an exit mechanism and creating a liquid market for existing investors. By the way, this also marks a moment where AI research companies transition from being privately funded moonshots to public utilities subject to quarterly earnings pressure and shareholder scrutiny. That's a real shift in incentives.

But the more telling story might be what's happening in hardware. Nvidia entering the Windows laptop market directly challenges not just Intel's dominance but the entire ecosystem that kept AI capabilities locked behind expensive servers and data centers. Pairing that with the rush toward on-device AI agents—Amazon's AgentCore announcements and Nvidia's autonomous agent partnerships with major software makers—and you see the infrastructure moving from centralized to distributed. MiniMax's M3 model performing near GPT-level capability at a tenth of the cost adds another wrinkle: the competitive pressure on model pricing is real, and it's coming from unexpected geographies.

What strikes me is that all of this is happening simultaneously. Massive capital concentration at the frontier, decentralization of compute to the edge, cost compression in model performance, and a rush toward autonomous agent infrastructure. It suggests we're not in a single AI story anymore but several overlapping ones that sometimes support and sometimes undermine each other. The question worth asking isn't whether AI is overfunded—it clearly is—but whether that capital is chasing the right problems or simply flowing toward whatever promised returns fastest.