TLDR

A concise deal sheet for the busiest frontier-AI fortnight of 2026. Model launches, strategic acquisitions, and infrastructure megarounds reshaping corp-dev timelines.

Frontier AI capital deployed year-to-date reached an estimated $262.65B by early July. Roughly 5.1x the comparable 2024 period. Even as deal count fell and median round sizes climbed toward infrastructure-finance scale. Against that backdrop, the second week of July produced a cluster of events that corp-dev and growth teams should treat as a single market signal.

Models: Moonshot AI shipped Kimi K3 (2.8T MoE, #1 Frontend Code Arena, weights July 27). Meta released Muse Spark 1.1 with its first paid Model API. OpenAI's GPT-5.6 Sol/Terra/Luna tiering continued rolling out after government review. M&A: SAP closed Prior Labs (€1B+ investment commitment). Qualcomm's $3.9B Modular acquisition and the ongoing strategic tuck-in wave (HubSpot, Zoom, Salesforce) continued from earlier in the month. Infrastructure: Fireworks AI ($1.5B at $17.5B), Baseten ($1.5B Series F), Together AI ($800M Series C), and Physical Intelligence ($800M Series C) concentrated capital in serving and robotics.

The pattern: training labs raise balance-sheet-scale rounds. Strategics buy capability they cannot build (tabular AI, coding agents, inference software). Enterprises face a widening menu of frontier APIs at falling per-task cost. Due diligence and eval frameworks. Not model loyalty. Determine who captures value.

My View on This

if your corp-dev pipeline still categorizes AI as a single theme, split it into three workstreams: foundation models (partnership and routing), vertical capability (tuck-in targets like Prior Labs), and inference infrastructure (build vs buy for serving). The July deal sheet shows all three moving simultaneously. And acquirers who conflate them will overpay for the wrong asset class.

Sources