Google Brain co‑founder doubles down with a $190 million Fund II and a playbook that writes code as well as cheques

Palo Alto, California — Andrew Ng, the famed computer‑science professor who launched Google Brain more than a decade ago, believes the gold rush of generative and “agentic” AI calls for a new way to build companies. That conviction underpins AI Fund, his venture studio that does far more than write seed cheques: it co‑founds startups, supplies engineers and even ships the first lines of code. On 5 May 2025 the studio closed an oversubscribed US $190 million second fund, bringing its total backing to roughly US $365 million.¹

“We’re operators to the point that we still write code with our portfolio CEOs,” Ng told *Crunchbase News* after announcing Fund II.² The studio’s 25‑person team rotates between half‑a‑dozen concept ‘sprints,’ validating market pain points before recruiting a chief executive and spinning out a new company. Since its first US $175 million vehicle in 2018, AI Fund has co‑founded about 35 ventures, many still in stealth.

The model mixes incubation and early‑stage venture capital. Unlike accelerators, AI Fund begins with its own idea pipeline, much of it sourced from corporate limited partners such as renewable‑energy giant AES, HP Inc., Japanese conglomerates Mitsui and Mitsubishi, insurer QBE and telecom TELUS.³ Those partners not only provide capital but also proprietary data sets, pilot customers and domain knowledge — raw material the studio leverages to de‑risk product‑market fit.

Ng argues that recent advances in large language models and tool‑using “agentic systems” have slashed the cost of prototyping. “A team of five can now build what once took 50,” he said. AI Fund therefore emphasises technical depth: every employee can code; the studio’s CFO and general counsel regularly contribute to GitHub repos during the first 90 days of a spin‑out.

Portfolio highlights illustrate the breadth. Gaia Dynamics uses foundation models to predict tariff exposure for global supply chains; SkyFire AI deploys drones and real‑time vision models for first‑responder services; Profitmind automates competitive price intelligence for retailers.² Yet the common thread, Ng insists, is “AI that matters” — enterprise SaaS aimed at energy, finance, healthcare and logistics rather than consumer chatbots.

The latest fundraise signals strong appetite for that thesis. Sequoia Capital and NEA returned as investors, joined by sovereign‑wealth arms from Singapore and Abu Dhabi, according to two sources familiar with the deal. AI Fund says it will launch 30 to 40 new companies over the next four years, writing initial cheques of US $500k–3 million and reserving up to US $15 million for follow‑on rounds.

Venture studios have proliferated in the generative‑AI era, but few offer Ng’s personal brand or technical gravitas. “Andrew’s playbook industrialises AI entrepreneurship,” said Carmen Chang, Partner at NEA.¹ “His team de‑risks at each stage — ideation, customer discovery, product architecture — in a way traditional VCs can’t.”

Sceptics point to potential dilution for founders and the challenge of scaling so many small teams simultaneously. AI Fund counters that its ‘shared services’ layer — recruiting, GTM, compliance and GPU credits — lets CEOs focus on shipping product. Early exits buttress the claim: Sicura, an automated cloud‑security startup spun out in 2020, was acquired by Cisco for a reported US $225 million last December.

For Ng, the studio’s second fund is as much about speed as capital. He wants to compress the timeline from whiteboard to Series A to “well under 12 months,” banking on the deflationary cost curve of AI infrastructure. † “If we can build and iterate faster than anyone else, we’ll spot the real economic value long before the hype cycle peaks,” he said in a LinkedIn post announcing the raise.⁴

As AI Fund hunts for its next batch of co‑founders — applications opened this week — its playbook is poised to influence how talent, data and dollars converge in the post‑ChatGPT landscape. If Ng is right, the future of venture may look less like a pitch meeting and more like a sprint planning session in a shared GitHub repo.