NEW YORK — The AI funding machine accelerated again this week, with five transactions totaling well over $700 million in fresh capital — and a valuation story at Cerebras that demands its own paragraph.
Start with the outlier. Benchmark Capital led a $225 million round into Cerebras, the AI chip designer, pushing its valuation to $23 billion. That figure represents a multiple expansion that would make most semiconductor analysts reach for a spreadsheet. Cerebras competes directly with Nvidia on AI inference hardware — a market where the incumbent's margins are essentially a standing invitation for challengers.
Benchmark did not stop there. The firm also co-led a $170 million Series A into Starcloud, a GPU cloud infrastructure provider, alongside EQT Ventures. Starcloud's post-money valuation: $1.1 billion. Two infrastructure bets from one firm in one week is a thesis, not a coincidence.
LMArena, the AI model evaluation platform that runs the widely-cited Chatbot Arena benchmarks, raised $150 million at a $1.7 billion valuation. The round reflects a market reality: as foundation model proliferation accelerates, independent evaluation infrastructure becomes a chokepoint asset. Enterprises need to know which model to buy. LMArena sells the answer.
The most strategically interesting transaction involves DeepSeek. The Chinese AI lab — which rattled Western markets in January with its cost-efficient R1 model — is reportedly pursuing outside capital despite holding adequate cash reserves. The South China Morning Post reports the move appears designed to build strategic relationships rather than address a capital gap — a playbook familiar to any observer of how Chinese tech firms have historically used funding rounds as alliance-building tools.
Rounding out the week: OpenAI, Google, and Anthropic announced a joint initiative targeting AI model theft, a rare display of coordination among competitors that signals the intellectual property stakes have grown large enough to override rivalry.
Collectively, the week's activity confirms that AI infrastructure — chips, cloud, evaluation, and model security — is where institutional capital is concentrating. The foundation model layer is increasingly assumed. What sits above and below it is where the next fortunes are being built.
