Eleven years after co-founding Sound Ventures with Guy Oseary, Ashton Kutcher is turning the page. According to TechCrunch, the actor-turned-investor is joining forces with Morgan Beller – co-creator of Libra (later Diem) at Meta – to launch a new venture capital firm. The stated goal: back early-stage startups operating in AI infrastructure, energy, and deep tech. Kutcher believes these are exactly the areas where the next truly transformative companies will emerge.

Betting on AI’s invisible layer

AI infrastructure is the least flashy but most decisive component of the ongoing revolution. It’s not about consumer models or generative apps, but everything that makes those models work: specialized chips, memory bandwidth, distributed training networks, cooling systems, and increasingly, a stable energy supply. Kutcher and Beller’s entrance into this space signals that venture capital is recognizing a bottleneck: without a leap forward in hardware and energy resources, the LLM race risks stalling. For those running on-premise workloads, this attention could translate into a broader ecosystem of suppliers, with more chip options (from GPUs to alternative architectures) and integration services designed for local, hybrid, or air-gapped environments.

Energy: the real enabler for on-premise deployment

The fund explicitly targets energy, and that’s no small detail. A GPU cluster for inference or training consumes hundreds of kilowatts, and private data center operators know it well. The on-premise deployment game is also fought on the ability to cool and power nodes without prohibitive costs. Startups developing on-site generation, energy storage, or liquid cooling systems could lower the TCO of a self-hosted installation, making it competitive against the cloud in scenarios where data sovereignty is non-negotiable (healthcare, defense, finance). Interest from a fund like the one announced could accelerate the commercialization of solutions that so far have been confined to labs or supercomputing centers.

Deep tech: beyond models, toward autonomy

When deep tech meets AI, the thought often goes to new computing architectures, photonic networks, or neuromorphic approaches. These are technologies that take years to mature, but promise to drastically reduce VRAM requirements, boost per-token efficiency, and enable fine-tuning of large models on mid-range hardware. From an on-premise perspective, this means running LLMs with less extreme hardware demands, widening the pool of organizations that can afford a local installation. A fund focused on early-stage deep tech adds competitive pressure and could shorten the time-to-market for such breakthroughs.

A shifting risk geometry

Kutcher’s departure from Sound Ventures to dedicate himself to this new vehicle is not just a personal repositioning. It points to a deeper transformation in venture capital priorities: from investing in application layers (apps, cloud services) to the physical and logistical building blocks of AI. For IT decision-makers evaluating deployment strategies, the message is clear: the future of AI infrastructure will not be a monopoly of a few vendors. If capital follows this path, the natural consequence will be more competition, more open standards, and more accessible hardware. And for those choosing the on-premise route for control, latency, or regulatory compliance, that is anything but a minor detail.

AI-RADAR will continue tracking these investments, focusing on their concrete impact on deployment architectures and the quantization and serving choices of teams operating away from the public cloud. For analytical frameworks on the trade-offs between cloud and on-premise, the /llm-onpremise section offers dedicated insights.