BP sells VC assets, ex-DeepMind raises $55M seed

BP is off‑loading over ten climate‑tech startups to Nordic PE firm Verdane, effectively winding down its BP Ventures unit launched in 2007. The move signals growing skepticism about corporate VC returns in the energy transition, and leaves BP retaining only a handful of bets that could directly benefit its core oil business.

Carta’s latest Startup Ecosystem Leaderboard (covering $124 B of venture capital from July 2025‑June 2026) shows the Bay Area pocketing 51 % of every AI dollar and 53 % of every B2B dollar. That concentration means founders raising AI or B2B rounds still gain a tangible edge by locating in San Francisco, where investors and talent are densely clustered.

Andrew Dai left Google DeepMind and raised a $55 million seed round at a $300 million pre‑seed valuation to build Elorian, a visual‑AI platform. He chose strategic investors like Nvidia and Menlo Ventures over higher offers, showing that deep technical focus and speed matter more than headline valuations. The deal signals that visual intelligence is the next hot frontier for AI capital.

First Round Review maps four levers, problem, persona, promise, product, that startups like Plaid, Clay and Lattice adjusted to escape pre‑product‑market‑fit limbo. The piece dissects each pivot, showing the exact change that sparked traction, and gives founders a playbook for diagnosing which lever to turn next.

Udemy co‑founder Gagan Biyani says the biggest product‑market‑fit lesson is to stop forcing a vision and let user behavior guide you. He recounts how Udemy’s live‑class model flopped, and a swift pivot to on‑demand courses drove the $1B business, showing founders that tracking real usage beats vanity metrics.

Alexandre LeBrun says AGI and “superintelligence” are meaningless labels and refuses to use them at AMI Labs. Instead he’s betting on world‑model AI that predicts physical states, a capability he argues is crucial for safe robotics in factories, homes, and streets. The stance redirects hype toward tangible, real‑world impact.

After a year‑and‑a‑half of rewriting 300 K lines, the Roc compiler now matches the original Rust version feature‑for‑feature, enabling a Wasm‑packed demo that’s under half the size. The post reveals the technical trade‑offs, team coordination, and why Zig proved faster and more compact for this project. The team targets a 0.1.0 release later this year.

When a founder’s day‑to‑day workload becomes the bottleneck, a COO can institutionalize processes, free the CEO to focus on strategy, and keep the company on an exit‑ready trajectory. Edison Partners shows the warning signs and how hiring the right operational partner prevents burnout and accelerates scaling.

After a YC startup’s co‑founder forced him out, Pranav Piyush faced a 60‑day deadline on his immigrant visa to secure a new role. Rather than panic, he leveraged the pressure to pivot, land a new job, and seed his next venture, showing how a career‑crushing split can become a launchpad.

Gurley argues that blanket tariffs are eroding American competitiveness in AI and semiconductors, and that policy should target strategic "pinch points" while letting private firms race ahead. The piece maps his insights to concrete industry examples, urging firms to outpace rivals instead of relying on state protection.

In 2025, half of all tech deals over $100 million involved companies with only about 100 staff. The smallest, Voyage AI, had 19 employees yet fetched $220 million from MongoDB. The data shows lean, bootstrapped startups are now prime M&A targets.

In a month, Salesforce, Autodesk, and Schneider Electric each paid roughly $3 billion for Fin (Intercom), MaintainX, and Cognite. The common thread isn’t revenue, it's the firms’ domain‑specific data that powers AI products. For builders, owning defensible data is now the premium acquisition asset.

The first quarter of 2025 saw eight cloud‑software deals total over $45 billion, highlighted by Google’s $32 billion purchase of Wiz and ServiceNow’s $2.9 billion acquisition of Moveworks. The burst signals a swing back to inorganic growth for mature platform players, reshaping the exit landscape away from IPOs.

Startups are shifting AI from generic chat to four concrete patterns, enhanced “Chat++” tools, invisible helpers that embed in existing workflows, hybrid workspaces that co‑author with users, and agent control centers that orchestrate multiple AI workers. These models promise tighter integration, higher productivity, and a more natural AI experience.