Memo to the President: How AI, Compute, and Connectivity Shape The Future of American Power
Recapping the AI+ Compute & Connectivity Summit
Hello, I'm Ylli Bajraktari, CEO of the Special Competitive Studies Project. In this edition of SCSP's newsletter, we distill key takeaways from our AI+ Compute & Connectivity Summit held on March 6, 2025. This was our third event in our AI+ Summit Series, following our previous summits on AI+ Energy and AI+ Robotics. We are grateful to everyone who participated in this critical conversation about America's technological future.
In today’s episode of the #MemosToThePresident series, I’m sharing my conversation with Dylan Patel, Co-Founder of SemiAnalysis. We discuss the Future of AI Supercomputing Clusters at the AI+ Compute & Connectivity Summit. I hope you enjoy the conversation!
America’s lead in AI, compute, and connectivity will define this century—but continued leadership is not guaranteed. China has made AI and microelectronics a national priority, deploying state-directed resources, a comprehensive national strategy, intellectual property theft, and aggressive industrial scaling to close the gap.
The AI+ Compute & Connectivity Summit brought together leaders from government, industry, startups, and academia to tackle one fundamental question: What actions must the United States take to secure global leadership in AI, semiconductors, and advanced networks? The answer is clear: removing regulatory restrictions, rebuilding domestic manufacturing, securing critical infrastructure, and incentivizing private-sector investment in the next generation of computing paradigms and advanced networks.
America’s AI Infrastructure: A Race Against Time
Rapid progress in AI continues to drive a frenzied infrastructure buildout, but the U.S. power grid isn’t keeping up. The nation is facing a gap in power production for AI infrastructure in the tens of gigawatts. Data centers are being planned at an unprecedented scale, but permitting delays, grid limitations, and regulatory choke points are slowing the rollout. Meanwhile, Beijing is actively improving compute efficiency to reduce infrastructure demands, as seen in its DeepSeek model, which suggests a new wave of efficiency gains that could change the competitive landscape. If the United States does not act now, it will find itself boxed in—by technology limitations and by its failure to scale domestic infrastructure.
Export Controls: Adapting to the Pace of AI Innovation
Current export controls are having an impact, though challenges persist. One of the primary issues has been the time it has taken to update and revise these controls, a process that has spanned approximately 12 months. Export controls are an important tool for the U.S. in addressing China’s accelerating progress in AI, but the pace of technological advancements often outstrips the timeline for implementing these measures. While the United States is focused on blocking the most advanced chips, scaling has shifted towards inference-based compute, where many chips remain unrestricted. If export controls don't evolve at the speed of shifting paradigms in AI progress, they will fail to have full effect. Future restrictions must anticipate shifts in AI architecture, targeting inference chips alongside training chips. They must also be developed in coordination with industry, where the highest situational awareness exists on cutting-edge advancements.
Semiconductors: The CHIPS & Science Act Is Just the Beginning
The CHIPS & Science Act has catalyzed hundreds of billions of dollars of investment in the domestic semiconductor industry, but investments to-date account for a fraction of the global semiconductor supply chain. If the United States wants to fundamentally change the game, it must do more. The global semiconductor supply chain was built over decades to optimize for cost and efficiency. Cost differences with Asia remain too large, and without additional policy levers like tariffs, tax incentives, and strategic procurement, chip production will remain offshored.
Workforce development is another critical challenge. The United States must scale training programs and create immigration pathways for top engineering talent to fill key roles in fabrication plants. Beyond workforce issues, the biggest vulnerabilities exist in the supply chain itself. Right now, America still lacks capacity in critical areas such as advanced packaging—a bottleneck that will hinder domestic semiconductor production if not addressed.
In addition, developing energy-efficient chips for AI has emerged as a critical battleground. Near-zero energy computing through reversible computing techniques could potentially deliver a 1,000x increase in efficiency using existing manufacturing processes. Promising alternative computing paradigms include superconducting computing that operates with almost no energy consumption. If these paradigms can be successfully commercialized and scaled, they can alleviate the AI energy bottleneck and power the next generation of large-scale AI systems.
The Quantum Race: A “Silent Mushroom Cloud” Moment
America still leads in quantum computing—but for how much longer? China has invested eight times more than the United States in this domain, and meaningful quantum computing capabilities could emerge by 2030 - considerably sooner than previous projections. Quantum computing will not just enhance existing capabilities—it will fundamentally rewrite the rules of computing.
Quantum computing will not replace classical systems in the future compute stack, but rather complement them as a specialized "QPU" alongside CPUs, GPUs, and AI accelerators. By exploiting unique quantum properties such as entanglement and superposition, quantum systems offer exponential scaling capabilities that classical architectures simply cannot match. Unlike classical computing's gradual progress, quantum's exponential scaling means breakthroughs could arrive suddenly, which could take the form of a "silent mushroom cloud" rather than a visible Sputnik moment—with implications that could shape the next century of global competition.
Connectivity: The Hidden Front in the AI War
The United States fell behind in 5G, and the consequences were immediate. Huawei built global dominance. Global networks became dependent on untrusted infrastructure. China gained leverage in global telecom standards. Slow execution of the 5G “rip and replace” program has left U.S. networks exposed to cyber threats from Chinese actors. Moreover, as the Salt Typhoon attacks underscored, adversaries are actively pre-positioning cyber capabilities to disrupt critical infrastructure in times of crisis.
Without a secure, high-performance network foundation, American innovation in AI and advanced computing will be constrained by logjams in data transmission, latency, and cybersecurity threats. Furthermore, spectrum allocation remains a major logjam, with regulatory delays slowing commercial network deployment while China rapidly expands its next-generation networks. With 6G and optical networks on the horizon, the United States must lead in connectivity outright, rather than risk another costly catch-up effort.
America Cannot Afford to Hesitate
Securing America's technological leadership in the AI era requires more than incremental improvements—it demands bold, coordinated action. The challenge before us is significant: closing a massive power gap for AI infrastructure, developing computing systems that are orders of magnitude more efficient, securing our networks and critical infrastructure against increasingly sophisticated attacks, and outcompeting China's state-directed approach without compromising American values. By creating market demand through strategic procurement, streamlining regulations that impede innovation, fostering deeper public-private partnerships, and investing in the nation’s unique research and startup ecosystems, the United States can maintain its competitive edge.
Missed the AI+ Compute and Connectivity Summit or interested in rewatching our sessions? Enjoy the entire event on SCSP’s YouTube Channel.