The AI revolution is in full swing — and for good reason. From a technological perspective, we’ve already sprinted through two intense laps of this digital race. But here’s the twist: crossing the “finish line” reveals it was merely a warm-up. The real race lies ahead. Welcome to The Last Mile.
The First Mile: The Silicon Titans
The first lap was all about silicon — the race to build faster, more powerful chips. Taiwan Semiconductor Manufacturing Company (TSMC) solidified its dominance, producing roughly 60% of the world’s semiconductors and over 90% of advanced chips. NVIDIA surged ahead with a market cap exceeding $4 trillion, driven by its AI-focused roadmap, including the announcements of Blackwell, Rubin, and Feynman GPUs through 2028. AMD retooled its offerings with the MI300 series, while Intel faced challenges but remains a key player with its upcoming 14A process and Gaudi AI accelerators. Newcomers like Groq also entered the fray, pushing innovation forward.
The stakes are immense. A modern chip fabrication plant costs over $20 billion to build, creating sky-high barriers to entry.
I’m cautiously optimistic for Intel to regain its mojo and reignite its foundry and competitive GPU/NPU business. It’s like betting on a 12-to-1 underdog — but who doesn’t love a good long shot? With the majority of the world’s semiconductors and advanced chips made in Taiwan, it’s a matter of national interest — especially for the U.S. — to have alternative chip-scale options. Call it Mission: IntelPossible.
Quantum computing is now coming up around the bend — a complex challenge and opportunity for a future lap.
The Second Mile: The AI Model Arms Race
The second lap saw the rise of platform companies — creators of Large Language Models (LLMs). This is the Belmont Stakes of AI, with thoroughbreds like OpenAI, Anthropic, Google, Meta, xAI, and China’s DeepSeek vying for dominance. These firms are backed by massive investments — Google, Microsoft, and Meta are projected to spend over $200 billion on AI-related R&D and infrastructure in 2025.
For a physical footprint perspective, Meta’s planned Hyperion data center in Louisiana will span an area comparable to Manhattan.
OpenAI is reportedly designing its own AI chips with TSMC for production by 2026, aiming to reduce reliance on NVIDIA. However, even the most advanced LLMs and GPUs are only as valuable as their real-world applications.
This is “build it, and they will come” on an industrial scale. The hard work now begins. Modernization of the Planet.
The Last Mile: Where Value Takes Shape
The Last Mile is where trillions in foundational investments translate into tangible value for businesses and society. This phase will greatly eclipse the first two laps in impact.
Human history’s greatest engineering feats — the Giza Pyramids, the Great Wall of China, the Moon Landing — may soon be rivaled by three AI-driven Last Mile engineering transformations:
1. Digital Modernization : Systems of Record
Much of the global economy runs on legacy code — COBOL, Assembler, Fortran, Powerbuilder — written when the Beatles and Stones topped the charts. Banking, healthcare, manufacturing, telecom, and other industries rely on “Systems of Record” so fragile that updates and rewrites risk catastrophic failure and personal job loss. We have been living in an endless era of kicking the technical debt can down the road.
Multi-modal Generative AI offers a lifeline, capable of analyzing vast codebases, generating modernization blueprints, and enabling rebuilds without prohibitive costs or risks. This isn’t just an upgrade — it’s a digital renaissance, allowing the big tech firms to evolve into AI-driven “Service as Software” models to help with this mass refactoring of code or face obsolescence.
2. Agentics: Systems of Work
Imagine an AI with unparalleled reasoning and IQ, integrated with Systems of Record and top LLMs, running on secure data ontologies performing complex tasks under human oversight. These “Systems of Work” could revolutionize industries — driving revenue, optimizing supply chains, accelerating scientific discovery, and enhancing education. AI agents could deliver healthcare via smartphones or education through smart devices, boosting global GDPs.
The important point about Agentics is that it can be implemented now on existing platforms to rapidly drive value. Start by connecting to your LLM(s) of choice, integrating your existing data sources, and building secure knowledge graphs. Once you build your super and service agents, you’re off to the races. Imagine a super sales agent watching your CRM database 24/7, automatically producing proactive proposals for your sales reps based on current sales trends. Or consider an AI agent acting as a sales advisor on large deals, knowing every detail about your competitors to help you form creative, winning strategies.
3. Physical AI: The Rise of Robotics
The physical realm of AI — autonomous vehicles, robotaxis, and humanoid robots — is taking shape. Companies like Tesla (Optimus) and Figure AI are leading early efforts. Picture robots managing factories, transporting goods, or handling household tasks (with clear instructions to avoid the Porsche).
The compute and energy demands for Robotics and these other two areas will be immense, potentially requiring advanced energy solutions like Small Modular Reactors (SMRs), though their widespread deployment remains speculative and likely decades away.
The True Value of the Last Mile
The Last Mile is where AI’s promise becomes reality. Without modernizing legacy systems, creating intelligent workflows, and transforming labor, AI data centers risk becoming costly relics. Innovation must accelerate — slow progress, armies of consultants, or building without AI integration are no longer viable.
Companies like Palantir are already leading in this new last mile playbook, with startups like Cursor, Perplexity, Blitzy, and ConceptVines quickly gaining momentum.
By 2035, we may look back and see this headline in The New York Times: “The World Transformed — All Because of the Successful Building of Last AI Mile.”