Imagine a world where AI doesn’t rely on distant data centers or shaky cloud agreements. That future is inching closer, and it’s reshaping how businesses think about technology, risk, and innovation. As financial instability rattles the AI supply chain – with private company defaults hitting over 9.2%, the highest in years – venture capital firm Lux Capital recently urged companies to secure compute capacity commitments in writing. But what if the solution isn’t just better contracts, but bypassing external infrastructure altogether?
The Rise of Compressed AI Models
Enter Multiverse Computing, a Spanish startup pushing compressed AI models into the mainstream. By shrinking models from giants like OpenAI and Meta, Multiverse enables AI to run locally on devices, eliminating cloud dependency and counterparty risk. Its CompactifAI app, featuring the tiny Gilda model, offers a glimpse of AI on the edge – processing data offline without sending it to remote servers. However, there’s a catch: older devices may lack the RAM and storage, forcing a fallback to cloud-based models and compromising privacy.
For enterprises, the real game-changer is Multiverse’s new self-serve API portal, providing direct access to compressed models for production use. CEO Enrique Lizaso highlights real-time usage monitoring as a key feature, addressing the core appeal: lower compute costs. In a market where efficiency is paramount, smaller models are becoming viable alternatives to large language models (LLMs), especially as their capabilities narrow the gap. Multiverse claims its HyperNova 60B 2602 model delivers faster responses at lower costs than its OpenAI-derived predecessor, a critical advantage for agentic coding workflows.
Broader Industry Shifts and Tensions
This move toward localized AI isn’t happening in a vacuum. French AI startup Mistral recently launched Mistral Forge, a platform allowing enterprises to build custom models from scratch using their own data. Timoth�e Lacroix, Mistral’s co-founder, notes that smaller models trade off breadth for customization, letting companies emphasize what matters most. Meanwhile, the AI investment landscape is booming, with GV’s managing partner Tom Hulme revealing that 80% of their investments are in AI or AI-native companies, driven by unprecedented growth rates from mobile distribution and natural language interfaces.
Yet, as AI becomes more embedded in critical operations, tensions are flaring. The U.S. Department of Defense has labeled Anthropic an ‘unacceptable risk to national security’ after the company refused to allow its AI for mass surveillance or lethal targeting, citing ethical red lines. This dispute, part of a broader legal battle, underscores a pivotal question: Who controls AI’s use in high-stakes scenarios? Similarly, Microsoft is considering legal action over a $50 billion cloud deal between Amazon and OpenAI, alleging it breaches exclusive partnership terms – a conflict highlighting the fierce competition and regulatory scrutiny shaping the AI ecosystem.
Practical Implications and Future Outlook
For businesses, compressed models offer more than cost savings. They enable AI deployment in environments where connectivity is unreliable, such as drones or satellites, and enhance privacy for sensitive fields like finance or healthcare. Multiverse already serves over 100 global clients, including the Bank of Canada and Bosch, and is rumored to be raising �500 million at a valuation exceeding �1.5 billion. But challenges remain: device limitations and the need for robust local processing power could slow mass adoption.
Looking ahead, the trend toward edge AI reflects a broader shift toward sovereignty and control. As Hulme points out, energy costs – a major variable in AI intelligence – raise questions about national competitiveness, with the UK facing rates around $350 per megawatt hour compared to sub-$100 in Norway. This isn’t just about technology; it’s about strategic advantage in an increasingly fragmented global landscape. For professionals, the message is clear: embracing compressed models could mitigate supply chain risks, but it requires balancing innovation with practical constraints and ethical considerations.