The Lonely Years Are Over
A decade ago, Lior Susan was pitching a contrarian idea: that the most important technology investments weren't in software, social media, or apps — they were in the physical world. In factories, in hardware, in the atoms-and-bolts infrastructure that software alone couldn't replace.
Most of the venture capital world wasn't buying it.
Fast forward to 2026, and Eclipse Ventures — the firm Susan built around that thesis — finds itself at the center of one of the most talked-about stories in tech. The firm's bet on Cerebras Systems, the AI chip startup that designs and manufactures some of the largest silicon wafers ever built, has become a landmark moment: a $2.5 billion outcome that validates years of patient, unfashionable investing.
What Eclipse Actually Bets On
Eclipse's portfolio reads like a supply chain wish list: industrial automation, advanced manufacturing, energy hardware, robotics. The through-line is companies that build things — companies that require not just code but deep engineering, physical infrastructure, and long development cycles that make most VC funds nervous.
Cerebras fit that profile perfectly. The company's signature product, the Wafer Scale Engine, is essentially a single chip the size of an iPad. Building it required solving manufacturing problems that had stumped the industry for years. It's the kind of hard, capital-intensive work that Eclipse was specifically designed to back.
"Investing in the real world was lonely," Susan has said of the firm's early years. "Nobody wanted to talk about it."
The Cerebras Moment
The $2.5 billion milestone — whether viewed as a valuation, an exit, or a funding round — represents more than a single win. For Eclipse, it's proof of concept. It demonstrates that patient capital applied to hard physical problems can produce returns that rival, and sometimes exceed, the software bets that have dominated venture returns for the past 20 years.
Cerebras has found a market in AI inference and training workloads where its architecture offers genuine advantages over conventional GPU clusters. As demand for AI compute continues to surge, the company's unusual approach — fewer, larger chips rather than more, smaller ones — is attracting serious enterprise interest.
A Shift in the Venture Landscape
What makes Eclipse's story notable is the broader shift it reflects. The era of cheap capital and software-only bets has given way to a renewed interest in deep tech: semiconductors, defense technology, energy infrastructure, manufacturing. Eclipse was early to this pivot by years.
Firms that once dismissed hardware as too slow and too capital-hungry are now scrambling to build physical-world portfolios. Eclipse already has one — seasoned, tested, and now validated at scale.
For Susan and his team, the Cerebras win isn't a destination. It's a datapoint in a longer argument: that the real economy, the physical world, is where the most durable technology companies get built. The lonely years, it turns out, were just the prologue.
Source: TechCrunch
