AI’s future hinges on hardware, not just models: expert
NEW JERSEY, UNITED STATES — As artificial intelligence (AI) investment continues, most of the spotlight remains on foundation models and flashy demos. But behind every breakthrough is hardware, the physical systems that power intelligence.
In 2024, less than 10% of AI capital went into infrastructure like chips, sensors, and embedded compute, according to Bain & Company. The imbalance, experts warn, could become AI’s Achilles’ heel.
“We’re getting too obsessed with the models,” Zhou Shaofeng, founder and chairman of Shenzhen-based Xinghan Laser, told Fortune. His deep tech company develops semiconductor laser chips, industrial LiDAR systems, and other high-precision hardware built to integrate intelligence directly into machines.
“Real intelligence isn’t just about prediction,” he said. “It’s about perception, interaction and action. And that all starts at the hardware level.”
Funding gaps, real-world bottlenecks
Shaofeng believes the tech world is ignoring the harder, more expensive piece of the AI puzzle. “The real bottlenecks aren’t technical,” he noted. “They’re economic.” Hardware systems require long development cycles, high integration costs, and compliance testing — challenges most VCs avoid.
While software products can be built, demoed, and deployed in months, hardware demands patience and scale. But the return, Shaofeng argues, is bigger. “Embedding intelligence into physical systems may cost more upfront, but the competitive moat it creates is much deeper.”
His sentiments echo broader concerns raised by industry leaders. OpenAI CEO Sam Altman recently urged U.S. lawmakers to back AI infrastructure. At Princeton, professor Naveen Verma is developing new energy-efficient AI chips, highlighting the limits of current architectures.
Real intelligence, real impact
Xinghan Laser is one of the few firms pushing AI into real-world systems — from factories to autonomous machines. “We’re building systems that can learn from the process itself and adjust in real time,” Shaofeng said. This shift, he adds, is crucial for industries like manufacturing and logistics, where AI is already driving the most ROI, according to McKinsey.
“Large models and hardware innovation aren’t opposing forces,” he added. “They’re mutually reinforcing. One pushes boundaries; the other brings that intelligence to life.”
