Google's AI Futures Fund bets on consumer-facing media apps, not infrastructure

What the fund actually bundles together

On May 12, 2025, Google introduced the AI Futures Fund, described by co-founder and director Jonathan Silber as an initiative that 'invests and works with startups to accelerate their ambitious ideas.' The offer has four components, and it's worth separating them because they don't all carry equal weight.

Startups get early access to select Google DeepMind models — the announcement names Gemini, Imagen for image generation, and Veo for video generation. They also get hands-on collaboration with Google researchers, engineers, product managers, designers, and go-to-market specialists; Google Cloud credits with dedicated support; and, for select companies, direct equity investment from Google.

The equity is gated ('select startups get the opportunity to seek direct investment'), and Cloud credits are standard for any cloud vendor courting builders. The genuinely scarce item here is pre-release access to Imagen and Veo alongside people who built them.

The three named startups are all frontend media products

Google chose to highlight three early participants, and the through-line is striking. Toonsutra is an Indian digital comic and webtoon app using Gemini's translation to reach India's linguistic diversity. Viggle is an AI meme-making platform exploring video creation with Gemini, Imagen, and Veo. Rooms lets users create, play, and share interactive 3D spaces, prototyping content with Gemini.

None of these is a developer tool, an agent framework, or a backend infrastructure play. All three are consumer-facing apps where the model output is the surface the user touches directly — a translated comic panel, a generated meme video, an interactive 3D scene. The interface is the product.

That fits the models on offer. Imagen and Veo produce images and video — outputs that only matter when rendered in front of a person. A fund leading with those two models is implicitly recruiting teams whose hard problem is presentation, latency, and interaction, not just inference.

Why early Veo and Imagen access shapes what founders can build

For a team building on generative image or video, the model's release timing dictates the product roadmap more than almost anything else. Getting Veo before it's broadly available means Viggle can design meme-video interactions around capabilities competitors can't yet call. That's a real frontend advantage: you can commit interface decisions to a generation quality your rivals haven't seen.

Google frames this plainly: these collaborations show how founders can 'test bolder ideas and build products that weren't possible before.' Read against the named apps, 'weren't possible' largely means output quality — translation fidelity for Toonsutra, video coherence for Viggle — reaching a bar where a consumer app becomes worth shipping.

The tradeoff for founders is dependency. Building your app's core experience on pre-release Imagen or Veo means your product's ceiling, pricing, and stability are tied to Google's model cadence and, for equity recipients, to Google's cap table.

The implication: this fund seeds the demand-side of Google's own models

The AI Futures Fund is open anywhere Gemini is available, which is a wide net. But the showcased picks reveal intent more than the eligibility rules do. Google is cultivating a class of consumer apps that make Imagen and Veo visible, useful, and habit-forming to end users.

For frontend-focused startups, the practical read is that the sharpest value in this program is model access plus the designers and product people Google is offering — not the credits or the funding, which you can source elsewhere. If your differentiation lives in the interface and depends on generation quality, being early on Veo is the lever. If your product is infrastructure or a backend agent, the three named participants suggest you're not the profile Google is currently spotlighting.