AI data center push collides with rising public resistance

Eclipse Ventures captured a striking payout this week when Cerebras Systems went public, turning the firm's early bets into a reported $2.5 billion return. The firm first put $6.5 million into the semiconductor startup in 2016 and followed with additional rounds that brought its total investment to $147 million. At the IPO price of $185 a share, that stake produced roughly a seventeenfold gain. For Eclipse founder Lior Susan, the outcome validates a thesis he has held since starting the firm in 2015: most global economic activity remains tied to the physical world, and durable value will come from companies that combine hardware and software rather than pure digital platforms.

Susan has argued for years that software-only advantages are eroding. In recent remarks, he noted that new AI tools can generate code quickly enough to reduce the protective barriers once enjoyed by enterprise software companies. Public-market investors appear to be reaching a similar conclusion. Shares of chip manufacturers such as TSMC and Micron have reached record levels, reflecting demand for the physical infrastructure that underpins advanced computing. A growing number of technical founders are now directing attention toward systems that link sensors, power systems, and manufacturing processes with algorithmic advances.

At the same time, surveys indicate rising public skepticism toward AI more broadly. Recent polling from Gallup found that only 18 percent of Americans aged 14 to 29 express hopefulness about the technology. An Economist/YouGov survey released this week showed more than 70 percent of respondents believe AI development is proceeding too rapidly, with majorities in both parties sharing that view. Negative sentiment toward AI has climbed from 34 percent three years ago to just over 50 percent in the latest YouGov data. Concerns center on employment displacement, higher electricity costs, and environmental effects from expanded data-center construction.

These attitudes create a practical constraint for companies racing to deploy large-scale models. Executives at several frontier laboratories have expressed surprise at the depth of opposition, viewing the technology's progress as comparable to earlier infrastructure buildouts such as the internet. Yet the physical requirements of scaling AI—specialized chips, reliable power supply, and extensive cooling—make the industry's footprint visible in ways that earlier software products rarely were. States and localities are already weighing permitting rules and utility-rate structures that could slow project timelines.

The Eclipse outcome illustrates one path through this tension. By concentrating on hardware that improves the efficiency of model training and inference, investors can capture returns even as software margins compress. Cerebras and similar firms must still navigate supply-chain limits and energy demands, however, which keeps their success linked to the same public worries showing up in polls. Whether those worries translate into durable regulatory or market pressure remains unclear, but the divergence between rapid technical gains and measured public confidence is now a measurable factor in capital allocation decisions.