AI is transitioning to local, compressed models on devices instead of relying solely on cloud-based systems
The path for AI is shifting. Until now, we relied heavily on the cloud to process complex tasks, because in most cases, we had to. The models were too big, the hardware too limited. But that’s changing quickly. Breakthroughs in model compression are putting serious AI power directly into consumer-grade devices, laptops, smartphones, even industrial equipment in the field. We’re no longer bound by remote servers. Intelligence can now live at the edge.
Multiverse Computing has made this clear. Roman Orús, the company’s Co-founder and Chief Scientific Officer, said compressed models are enabling high-performance language models to run directly on devices. What this means is we’re going from AI in the cloud to AI in your hand, locally processed, fast, and untethered. And yes, it’s reliable.
This isn’t a race to cut cloud costs. It’s about giving users more speed, control, and independence. In many environments, especially those where connectivity is poor or intermittent, embedded AI becomes essential. You get latency-free interaction blended with hyper-personalized functionality. No network? No problem. The AI still works. The rise of Edge AI is one of the clearest signs that centralization in technology is breaking down, and in a good way.
For executives, this shift is worth watching closely. Moving processing to the edge decentralizes system risk and strengthens resilience. It also opens the door to new kinds of products, faster, safer, more autonomous, and that means opportunities. Any organization that leverages this shift intelligently can gain a competitive edge, not just in speed, but in how flexibly and securely they operate.
User data privacy and ownership are key drivers behind the adoption of local AI processing
Here’s something most users care about, even if they don’t always say it directly: trusting who holds their data. Right now, people are increasingly skeptical about cloud services storing sensitive information. That’s a problem for businesses who rely on personal data to deliver service, not to mention regulators and policymakers who are stepping in with stricter standards.
This is where local AI completely changes the game. If a device processes information on-site, whether that’s a conversation, a schedule, a document, then the data doesn’t leave that device. There’s nothing for third-party providers or networks to intercept. This restores control to the user. And control is the foundation of trust.
As Roman Orús pointed out, privacy is the lead driver for consumers choosing local AI. You get the same functionality, or better, and you don’t have to give up your sensitive information. That’s a good trade. He also mentioned something important, Router AI. It’s a small coordinating model that helps decide if a task gets handled locally or switched to the cloud, optimizing for both capability and control. Smart orchestration like this is how you maintain a seamless user experience while staying privacy-compliant.
For the C-suite, here’s the bottom line: privacy-centric solutions are more than just good ethics, they’re good business. In regions with heightened regulatory pressure, like the EU, local AI offers a way to stay compliant while still scaling AI-driven services. It also builds credibility with customers. If you’re offering a product that protects privacy by design, you can lead in markets where trust has become a key differentiator.
Edge AI is particularly transformative in sectors like defense and healthcare
Some environments can’t afford delay or disconnection. Defense and healthcare top that list. In these sectors, fast execution and absolute data control aren’t just preferences, they’re requirements. That’s where Edge AI is taking hold. When AI models are compressed and embedded directly into local devices, platforms in the field or inside hospitals don’t need to reach out to a data center to make decisions. They can process information instantly, right where it’s collected.
Defense is a clear example. Roman Orús, Co-founder and Chief Scientific Officer at Multiverse Computing, outlined how drones, vehicles, naval systems, and wearable soldier gear can all be equipped with onboard AI. In unstable or hostile environments, with limited or contested connectivity, centralized cloud processing simply doesn’t work. Instead, the AI needs to operate in real time, detecting threats, navigating autonomously, and analyzing sensor data without delay.
In healthcare, the logic is just as strong. Orús explained how hospitals are using ultra-compressed AI models on secure workstations and private clouds to handle diagnostics and patient summaries without transmitting any sensitive data outside the network. That matters in regulated health systems, where patient privacy isn’t optional. These models also reduce the need for costly infrastructure, making advanced intelligence affordable for more organizations.
For leaders in defense or healthcare, the main takeaway is strategic independence. With AI embedded locally, operations no longer hinge on backhaul connectivity or third-party systems. What you get is speed, security, and the freedom to act immediately, without compromising on regulatory or mission-critical constraints. For businesses supplying technologies to these sectors, offering trusted, independent systems is now a growth market.
Governments are harnessing self-hosted AI models to enforce digital sovereignty and achieve strict security standards
Governments are moving deliberately toward AI that’s not only efficient, but sovereign. That means building systems that run internally, without relying on external platforms or vendors. When national agencies need to process sensitive, strategic, or classified data, hosting AI models on internal infrastructure is no longer optional. It’s foundational. The goal is to maintain full control of critical information systems end to end.
Roman Orús of Multiverse Computing made this clear. He pointed out that the public sector is widely adopting self-hosted, compressed large language models. Use cases include automating bureaucratic workflows, summarizing lengthy policy documents, and extracting insights from public consultations. These tools don’t just speed up administration, they can support more informed, data-driven decision-making at the highest levels.
But speed isn’t the key driver. Security is. Orús explained that local deployment ensures data is kept within secure, internal networks, isolated from the public internet, accessible only to authorized personnel. This approach aligns with rising concerns about digital sovereignty across Europe and other advanced economies, particularly as threats to data infrastructure intensify.
For C-suite executives in government tech or vendors working with the public sector, supporting that sovereignty matters. Cloud-based AI can be fast to deploy, but it can’t always meet strict national compliance requirements. Offering models that deliver high performance offline, while giving agencies full control over access and data flow, is quickly becoming a baseline expectation. If you’re building or enabling AI in the public sector, prioritize transparency, containment, and internal control. That’s where public trust and future procurements are headed.
Escalating energy demands from AI operations are driving the design and regulatory focus
AI is pushing boundaries, but energy use is pushing back. Training and running large-scale models demands enormous compute power. That translates into high energy costs, increased carbon emissions, and escalating pressure from regulators. As AI enters more sectors and scales up, governments and enterprises don’t just want performance, they expect efficiency. Models that consume less power while delivering comparable output are moving to the front of the line.
Goldman Sachs estimates that AI will drive a 165% increase in data center power consumption by 2030. This isn’t hypothetical. It’s becoming a regulatory and environmental issue. The EU AI Act now requires general-purpose model developers to document energy usage. On top of that, the 2023 EU Energy Efficiency Directive mandates detailed reporting on energy, water, and sustainability metrics in data centers. And more is coming, between now and 2026, multiple initiatives will imply new compliance thresholds.
Roman Orús, Co-founder and Chief Scientific Officer at Multiverse Computing, said model efficiency is no longer just an engineering problem, it’s becoming a condition for market viability. He emphasized that compressed AI models running at the edge aren’t just faster and more resilient, they fundamentally consume less energy. That’s good for enterprise costs and good for meeting ESG expectations.
For C-suite leaders, here’s what this means: the environmental footprint of your technology choices is becoming part of your competitive strategy. Regulators are paying attention. Customers are watching. Investors are asking questions. To stay ahead, your AI deployments need to be regulatory-ready, energy-aware, and performance-optimized. That doesn’t mean sacrificing capability, it means optimizing design, choosing better architectures, and aligning with where the market, and legislation, are going.
Key takeaways for leaders
- Edge AI is gaining momentum: Executives should anticipate a shift from cloud AI to on-device models as compression technology advances. This transition enables faster, offline capabilities with reduced latency and infrastructure dependence.
- Privacy drives user demand: Business leaders should prioritize local processing to meet growing consumer expectations around data ownership and privacy. Local AI boosts trust by eliminating exposure to third-party providers.
- Sector-specific impact is accelerating: Leaders in defense and healthcare should invest in Edge AI to improve resilience, security, and real-time decision-making in constrained or high-stakes environments.
- Governments are building AI independence: Vendors and public sector strategists should align with the rising demand for self-hosted models that support national security, efficiency, and digital sovereignty mandates.
- Energy efficiency is becoming a competitive edge: Executives must choose AI architectures that deliver high performance with low energy use to meet sustainability targets and comply with tightening EU energy regulations.
