AI data center push collides with rising public resistance

Eclipse Ventures turned a long-held focus on physical-world technology into a $2.5 billion return this week after Cerebras Systems went public. The firm first put $6.5 million into the semiconductor company in 2016 and later increased its total stake to $147 million. At the IPO price of $185 per share, that position delivered a 17-fold gain. Founder Lior Susan started Eclipse in 2015 with the view that 85 percent of global economic output remains tied to tangible goods and processes rather than pure code. Early on, that stance drew little interest in a venture scene dominated by software-as-a-service deals.

Susan pointed to recent market signals that align with the original thesis. Shares of Taiwan Semiconductor Manufacturing and Micron Technology reached record highs, while top founders now pursue companies that combine hardware and software. He noted that software advantages have narrowed because large language models can generate much of the code that once required teams of engineers. Enterprises, he added, can now build custom tools without relying on off-the-shelf SaaS platforms.

The same week brought fresh evidence that public sentiment toward artificial intelligence is cooling. A Gallup survey found only 18 percent of people ages 14 to 29 feel hopeful about AI. An Economist/YouGov poll showed more than 70 percent of Americans believe the technology is moving too fast, with majorities in both parties sharing that view. Negative opinions have risen from 34 percent three years ago to just over 50 percent today. Concerns center on job displacement, higher electricity demand, environmental effects, and further concentration of gains among already wealthy firms.

Several AI company leaders have expressed surprise at these reactions. Some described the technology as simply the next stage after the internet, yet polling indicates many workers and consumers see concrete trade-offs rather than automatic progress. Commencement speeches praising AI as the next Industrial Revolution have drawn audible pushback from audiences.

Eclipse’s payout from Cerebras illustrates how returns have flowed to investors who placed capital in physical infrastructure instead of betting solely on software layers. Semiconductor production requires large-scale facilities, supply-chain coordination, and engineering talent that cannot be replicated by prompting a model. Public markets appear to be pricing those constraints into higher valuations for chip makers even as pure software valuations face pressure.

The pattern fits a broader record in which technological advances have delivered gains when they lowered the cost of producing real goods and services. Investments that ignore those production realities have repeatedly produced shorter-lived results. Eclipse’s experience shows that founders and investors who stayed focused on the physical constraints of energy, materials, and manufacturing captured outsized value once those constraints became binding.