The alternative investment world converged at the Miami Beach Convention Center for iConnections’ Global Alts Miami 2025, where industry visionaries Antonio Gracias (Founder, CEO & CIO of Valor Equity Partners) and Gavin Baker (Managing Partner & CIO of Atreides Management) took the stage with iConnections CEO Ron Biscardi. Their talk ranged from the explosive breakthroughs in artificial intelligence to broader implications for U.S. economic policy and global competition. Here’s a look at the highlights from their conversation.
Baker kicked off the discussion by referencing “Deep Seek,” a recent development that showcased how quickly AI models can be replicated and open-sourced. Just as Antonio Gracias and Gavin Baker predicted a year earlier, any large-language model (LLM) without proprietary data is at risk of becoming commoditized.
“Everybody was certain they could replicate OpenAI’s accomplishments,” Baker noted, “but nobody had done it—until now.” Deep Seek’s success in mirroring the performance of established foundation models, and then open-sourcing it, has rung alarm bells for companies lacking distinctive data advantages. Baker emphasized that if a firm cannot feed its AI engine with unique information (or highly efficient architectures), it will struggle to differentiate itself in this increasingly crowded space.
Both Gracias and Baker see enormous growth potential in AI, but they stressed that the real gains will come from inference—using trained models in live applications—rather than simply training bigger models. Baker pointed out that the cost of deploying AI has the potential to drop drastically, with new breakthroughs reducing the time and computational expense per query. That drop in cost, he believes, will spur demand: historically, when a key technology becomes cheaper, its usage soars.
Gracias added that companies must focus on “high growth and high return on capital,” especially in data-center spending. He believes that xAI, a venture in which both he and Baker are invested, will have “the highest return on capital in the U.S.” thanks to more efficient deployments that can adapt to cheaper or differently configured GPUs.
A notable concern raised by both speakers: the possibility that China’s cutting-edge AI may come bundled with certain ideological constraints. Gracias pointed out that models developed under the Chinese government’s purview might embed the values of the Chinese Communist Party, distributing those views to global users, especially in developing regions. These influences, he warned, need careful scrutiny, particularly if they arrive “for free” to countries that lack resources to develop homegrown models.
Steering away from purely technical AI topics, Biscardi asked about “Doge”—shorthand for a rumored initiative involving Elon Musk and a host of top technologists to help the U.S. government address the federal budget deficit and bureaucratic inefficiencies. Gracias likened it to a “Manhattan Project” for streamlining government operations, clarifying that the team aims to tackle America’s debt crisis by cutting trillions in unnecessary spending and unleashing productivity. “We basically spend $6.5 trillion and take in $4.5 trillion. The math doesn’t square,” he said. “We can’t keep running these deficits.”
Both Gracias and Baker stressed that reining in excessive regulation could unlock faster growth—potentially lifting annual GDP increases to the 3–5% range instead of the modest 1–2% norm. Baker recounted anecdotes about how regulations can hinder innovation, even forcing SpaceX to conduct over-the-top environmental studies (like putting headphones on baby seals to test rocket noise). These stories, though humorous, illustrate how an entrenched bureaucracy can choke potentially transformative technologies.
Looking ahead, Gracias and Baker predicted 2025–2026 as a key inflection point for robotics, especially humanoid robots. They referenced Tesla’s Optimus robot as an example of how AI can merge with robotics to handle tasks in everyday life—everything from taking out the trash to washing dishes. This “unleashing” of physical automation could mirror the way ChatGPT and other LLMs accelerated widespread adoption of AI-driven software.
“We’re entering the year of the robot,” Gracias said, “and it will fundamentally change the productivity equation in the U.S. economy, igniting growth.” Baker concurred, suggesting that next year’s conference might even feature a humanoid robot on stage.
The conversation made clear that while AI model development continues at a breakneck pace, the real transformation lies in applications—where AI meets real-world usage. From the deployment of large-scale language models in enterprise and government workflows, to the rise of humanoid robots for everyday tasks, this new wave of intelligence promises a dramatic boost in productivity.
Yet the conversation also carried a note of caution: as AI technology becomes cheaper to develop and more ubiquitous, questions around global competition, data sovereignty, and social governance gain urgency. Gracias and Baker’s optimism for American leadership in AI is grounded in the assumption that deregulation and strategic national policies can foster both innovation and a balanced budget.
“There’s a 90% chance this is awesome for humanity,” Baker remarked, hinting at the Star Trek-like future AI could unlock. For Antonio Gracias, these next years of AI progress will coincide with a renewed American drive to cut waste, spark enterprise, and ensure the country’s global influence endures.
Their predictions may be bold, but if the last few years of breakthroughs in AI are any indication, the future—complete with humanoid robots and streamlined government—may be closer than we think.
