AI News: April 30, 2026

1. SoftBank Spins Up a Robotics Company to Build AI Data Centers, Eyes a $100B IPO

SoftBank. SoftBank is forming a robotics venture aimed specifically at constructing AI data centers, with early discussions targeting a roughly $100B IPO. The pitch is recursive: AI and robots are needed to build the infrastructure that powers AI and robots, and SoftBank wants to own that vertical. The move pairs with the conglomerate’s earlier compute and chip bets and reads as a bid to capture margin from the data-center construction boom rather than only the silicon and power layers. Source

2. Parallel Web Systems Hits a $2B Valuation Five Months After Its Last Round

Parallel Web Systems. Parag Agrawal’s AI-agent web infrastructure startup raised $100M led by Sequoia just five months after its last $100M round, doubling its valuation to $2B. The company’s APIs let AI models autonomously browse and verifiably interact with the web — a layer that has become a contested wedge as agent harnesses move from “screenshot the page” to structured, programmable browsing primitives. Source

3. Firestorm Labs Raises $82M for Container-Sized Drone Factories

Firestorm Labs. The San Diego defense startup raised an $82M Series B (led by Washington Harbour Partners) to put drone manufacturing inside shipping containers and deploy it to front lines, bringing total funding to $153M. The pitch matches a growing pattern in defense-AI funding: distributed manufacturing co-located with the autonomy stack, rather than centralized factories feeding centrally commanded fleets. Source

4. Scout AI Raises $100M to Train AI Agents for the Battlefield

Scout AI. Colby Adcock’s defense-AI startup raised $100M to develop AI agents that let individual soldiers command fleets of autonomous vehicles. TechCrunch profiled the bootcamp where the models are being trained, with reporters noting the unusually tight loop between operators and model engineers. The round is another data point that defense AI is no longer a niche — it is a primary destination for top-tier capital chasing reinforcement-learning and tool-use research. Source

5. Aidoc Closes a $150M Series E Led by Goldman Sachs for Clinical Radiology AI

Aidoc. The clinical AI radiology company closed a $150M Series E led by Goldman Sachs Growth Equity, fueling expansion of its FDA-cleared imaging triage platform across hospital systems. Aidoc remains one of the few AI-in-healthcare names with both regulatory coverage and meaningful deployed footprint, making this round a useful late-stage benchmark for what regulated medical-AI exits actually price at. Source

6. Rogo Crosses $300M Total Funding for Investment-Banking AI Agents

Rogo. Sequoia, Thrive, and Khosla led a new round taking Rogo’s total funding past $300M for an AI platform that automates deal workflows and financial modeling, now used daily by 35,000+ bankers across more than 250 institutions. The metric to watch is the daily-active count: vertical AI in regulated workflows tends to plateau quickly when agents are evaluated against real artifacts, and Rogo’s penetration suggests its harness has cleared that bar. Source

7. Pure Data Centre Pauses All Middle East Projects After Iran-War Strike Damage

Pure Data Centre Group. The London-based developer (which operates more than 1 GW across Europe, the Middle East, and Asia) halted all Gulf-region investment after one of its facilities took damage from an Iranian missile/drone strike. War damage is uninsurable under standard data-center policies, throwing into question a multi-hundred-billion-dollar regional AI buildout that hyperscalers and sovereign funds have been racing to underwrite. Source

8. Runway CEO Argues World Models Are the Real Frontier After AI Video

Runway. CEO Cristobal Valenzuela told TechCrunch’s Equity podcast that AI video was a prequel and that world models — not generation — are the actual frontier. The New York-based company has raised roughly $860M at a $5.3B valuation and is pushing into simulation-style training environments where video output is one observable of an underlying physics-and-causality model rather than the product itself. Source

9. Shapes Launches a Group-Chat App That Mixes Humans With AI Characters

Shapes. A new social app puts customizable AI personalities directly into group chats alongside human users, attempting a Discord-style experience where bots are first-class participants rather than slash-command sidekicks. The product is a useful test of whether persistent, addressable AI characters can sustain engagement in mixed groups, or whether they collapse to pet-bot novelty once the launch buzz fades. Source