Google is channeling resources toward a new wave of AI startups, including those founded by former employees, through a support structure that provides up to $350K in credits alongside technical mentorship and infrastructure access.

The backing comes via Google’s existing startup programs rather than a purpose-built fund, but the practical effect is the same: former Googlers building AI companies now have a credible path back to the mothership’s resources without giving up equity to get there.

What the support actually looks like

Google for Startups and Google Cloud serve as the primary vehicles here, offering eligible early-stage companies credits, compute access, and hands-on technical guidance.

The programs are not exclusively for ex-employees, but the growing number of former Google and DeepMind alumni founding companies means the overlap is significant. Nearly 200 former DeepMind employees have founded or joined AI startups, making them a natural constituency for any program Google runs in this space.

The equity-free structure is worth noting. Startups receive assistance without surrendering ownership, which is a different model from traditional venture-backed accelerators where funding comes attached to a stake in the company.

Google’s 2025 AI First accelerator in India offers a useful illustration of how competitive these programs have become. This cycle selected 20 startups from a pool of more than 1,600 applicants.

The Area 120 factor

Google’s internal incubator, Area 120, historically gave employees a structured way to build experimental projects inside the company. That changed in 2022, when Area 120 went through significant restructuring and cuts that reduced its scope considerably.

When a project is internal, Google owns the output. When former employees build independently with $350K in cloud credits and mentorship, Google gets proximity to the innovation without the liability of ownership, and the founders keep their upside.

What this means for investors watching the AI space

Programs that provide meaningful compute credits matter more in AI than in most other sectors. Training and running models is expensive, and $350K in cloud credits can extend a startup’s runway in ways that a cash equivalent might not, since the credits go directly toward the infrastructure cost that would otherwise drain the bank account fastest.

For investors evaluating early-stage AI companies, participation in Google’s accelerator programs functions as a useful filter. It does not guarantee a good business, but it signals that the team cleared a competitive selection process and has access to Google’s technical mentorship network.

Companies coming out of these programs have not already sold a chunk of ownership to the accelerator, which leaves more room for early-stage investors and keeps founder incentives intact.

Microsoft has deepened its relationship with OpenAI, Amazon has backed Anthropic, and Meta has taken an open-source route. Google’s approach, cultivating a distributed network of AI startups that depend on its infrastructure, is a different kind of ecosystem play, less about owning the winner and more about making sure the winners run on Google Cloud.