Ashton Kutcher and Morgan Beller Bet on AI Infrastructure in New VC Fund

Ashton Kutcher is leaving Sound Ventures, the firm he co-founded in 2015, to launch a new venture capital firm with Morgan Beller, the former Facebook executive known for co-creating the Libra stablecoin project. The move signals a strategic pivot from backing category-leading AI labs to investing in the infrastructure and energy layers that power them.

• Kutcher’s new fund shifts focus from high-conviction AI lab bets to the underlying infrastructure and energy supply chains.
• Morgan Beller, a veteran of crypto and Web3, brings experience in decentralized systems and early-stage platform bets.
• The move reflects a broader VC trend: as AI models commoditize, the value is migrating to compute, data centers, and power generation.
• Energy constraints are becoming a bottleneck for AI scaling, making infrastructure investments increasingly strategic.
• This transition may signal a maturation of the AI investment cycle, similar to the shift from internet companies to telecom and cloud providers in the 2000s.
• Founders building in AI should watch for capital flowing into energy and hardware, which could reshape startup funding dynamics.

Why Is the AI Investment Cycle Shifting From Labs to Infrastructure?

The departure from Sound Ventures is not just a personnel change; it represents a calculated bet on where the next wave of AI value creation will occur. Sound built its reputation on concentrated, high-conviction bets in labs like OpenAI and Anthropic. But as large language models become more accessible and their capabilities converge, the competitive advantage is moving downstream. The new fund appears to be chasing the layer underneath those companies: the data centers, the specialized chips, and the energy grids that make AI training and inference possible. This mirrors historical patterns in technology investing. In the late 1990s, capital flooded into internet companies. A few years later, the smart money moved to the infrastructure providers: Cisco for networking, Equinix for data centers, and utilities for power. AI is following a similar trajectory. The cost of training a frontier model now runs into the hundreds of millions of dollars, and inference at scale requires massive energy consumption. Investors are realizing that the real bottleneck is not algorithmic innovation but physical constraints.

The shift from AI labs to infrastructure mirrors the 2000s transition from internet companies to telecom and cloud providers. The winners in that era were often the picks-and-shovels suppliers, not the dot-coms themselves.

How Will Morgan Beller’s Background Shape the New Fund’s Strategy?

Morgan Beller is best known for her role at Facebook (now Meta), where she co-created the Libra stablecoin project, a bold attempt to build a global digital currency. She later became a general partner at NFX, a venture firm focused on early-stage startups. Her expertise spans decentralized systems, platform economics, and early-stage company building. Beller’s experience in crypto and Web3 is particularly relevant to the new fund’s focus on infrastructure. Decentralized networks, token-based incentive models, and distributed computing are all concepts that overlap with the challenges of scaling AI. For example, energy markets are increasingly decentralized, with renewable sources, battery storage, and grid management becoming software-defined. Beller’s background could help the fund navigate these complex, multi-sided markets. The partnership also brings a complementary skill set: Kutcher’s star power and network in Hollywood and tech, combined with Beller’s deep operational and investment experience. This could give the new firm access to deal flow that other infrastructure-focused VCs might miss.

Investment Layer	Traditional VC Focus	New Fund Focus	Key Implication
AI Model Labs	High-conviction bets on OpenAI, Anthropic	Minimal direct lab investment	Models commoditize; value moves upstream
Compute Hardware	GPU manufacturers (Nvidia)	Chip design, cooling, networking	Specialized silicon becomes critical
Data Centers	Colocation providers	Energy-efficient design, location strategy	Power availability drives site selection
Energy Generation	Utility-scale renewables	Grid management, battery storage, nuclear	AI’s energy appetite reshapes power markets

Which Infrastructure Bottlenecks Will the New Fund Target?

The most immediate bottleneck is energy. Training a single large language model can consume as much electricity as a small town over several months. As AI inference becomes ubiquitous, that demand will only grow. The new fund is likely to focus on solutions that address this constraint: advanced nuclear reactors, long-duration battery storage, and software-defined grid management. Another critical area is data center design. Not all data centers are created equal. AI workloads require high-density racks, advanced liquid cooling, and low-latency interconnects. Companies that build next-generation data centers optimized for AI could see explosive growth. Finally, the fund may target the supply chain for specialized chips. While Nvidia dominates today, the market is ripe for disruption from custom ASICs, optical interconnects, and chiplet architectures. These are capital-intensive bets, but the potential returns are enormous.

What Does This Mean for AI Startup Founders?

Founders building AI-native companies should pay close attention to this shift. The availability and cost of compute and energy will increasingly determine which startups succeed. Those that can optimize their models for efficiency, or that build on infrastructure that is itself efficient, will have a competitive advantage. The new fund’s focus may also signal a change in fundraising dynamics. Startups in the AI infrastructure space, such as those building data center cooling solutions or energy management software, may find a more receptive audience. Conversely, founders building yet another LLM wrapper may face a tougher fundraising environment as VCs chase deeper tech moats.

• Compute efficiency: Startups that reduce the cost of inference through model pruning, quantization, or specialized hardware will be prized.
• Energy innovation: Companies developing modular nuclear reactors, grid-scale batteries, or hydrogen fuel cells for data centers could attract significant capital.
• Data center design: Next-generation cooling, modular construction, and software-defined networking are areas where startups can differentiate.
• Supply chain resilience: The chip shortage of 2020-2023 showed that hardware supply chains are fragile. Startups that diversify or localize production will be valued.

Infrastructure investing is capital-intensive and requires long time horizons. VCs expecting quick exits may be disappointed. The energy and hardware sectors are also subject to regulatory and geopolitical risks that software startups rarely face.

Which Warning Signs Predict Problems Ahead for This Strategy?

While the thesis is compelling, there are risks. Infrastructure investments are notoriously slow to mature. Building a new nuclear reactor or a chip fabrication plant takes years, not quarters. VCs accustomed to software-like returns may find the timeline frustrating. Another risk is technological obsolescence. The AI hardware landscape is evolving rapidly. What looks like a smart bet today could be rendered obsolete by a breakthrough in photonic computing or quantum processing. The fund must be careful not to over-invest in a single approach. Finally, there is the risk of overcapacity. If every major VC firm piles into AI infrastructure, the market could become saturated, driving down returns. The key will be identifying bottlenecks that are genuinely hard to solve, not just trendy.

The formation of this new fund is a bellwether for the AI industry. As the low-hanging fruit of model improvements is harvested, the real work of building the physical and digital scaffolding for an AI-powered world begins. Kutcher and Beller are betting that the infrastructure layer will generate the next generation of outsized returns. Whether they are right will depend on their ability to navigate the complex intersection of technology, energy, and regulation.

Source: TechCrunch AI