Benchmark's Eric Vishria on the decade-long Cerebras bet: naivete, relentless grind, and a $23B SPV

Benchmark's Eric Vishria on the decade-long Cerebras bet

Eric Vishria joined Benchmark in 2014 and has spent the last twelve years doing the same thing: finding early-stage founders, backing them, and staying close for a decade or more. Cerebras is the clearest expression of that model.

Vishria describes his role across a long-hold investment as fluid. Early on it's fundraising and team-building; later it's being a sounding board when the highs and lows of building a company become hard to carry alone. He says that founder relationship is the part of the job he loves most, and that Cerebras is the fourth company he's worked with for more than ten years.

“In 2016, slide three was 'GPUs actually suck for deep learning — they just happen to be 100 times better than CPUs.' As soon as he said it, it's just like a light bulb went off. Fast forward six, seven years, we're still slogging it out and have raised so much money and have very little revenue. Then inference exploded. In February, we raised an SPV — which we've never really done before — to invest in Cerebras.”

The February SPV

In February 2026, Benchmark raised an SPV to invest in Cerebras at a $23 billion valuation — something Vishria describes as unusual for a firm that has stayed resolutely early-stage. He frames it as a special-opportunity exception, not a strategy shift, and draws a parallel to a brief stretch of public-market investing Benchmark did when the Nasdaq fell during COVID and early-stage deal flow dried up. The throughline is the same: when the firm sees something compelling outside its normal lane, it moves.

Fund track record

Vishria walks through three fund vintages that illustrate how Benchmark's early-stage discipline plays out across market cycles.

Fund VII (pre-2014, before Vishria joined) includes Uber, Snapchat, Elastic, Stitch Fix, Discord, and WeWork. He credits the team for riding the consumer-mobile wave with enough winners to deliver fund-level multiples from several positions.

Fund VIII, the 2014 vintage Vishria helped build, swings entirely the other direction — enterprise software. Confluence and Amplitude returned capital; Cerebras and Chainalysis are the larger bets still playing out.

The 2022 vintage holds Sierra, Fireworks AI, LangChain, Legora, and Merkor — all early AI-infrastructure and application plays that Vishria says look promising.

The pattern across three consecutive funds covering consumer, enterprise, and AI is that the strategy stayed fixed while the opportunity set rotated. Vishria argues that's the point: staying early-stage while remaining open to what's actually ripening is harder to execute than it looks, especially when peers are expanding fund sizes and strategies.

Portfolio construction

Vishria splits his current portfolio into three buckets. AI application companies like Sierra and Legora are scaling now. Infrastructure plays like Fireworks AI are riding inference demand. And two longer-dated bets — Star Cloud (space data centers, led by Benchmark partner Chase) and Sunday Robotics (home robotics) — are not expected to generate meaningful revenue in the near term. He's explicit that those investments are priced for a longer ripening window, the same logic he applied to an AI hardware investment in 2016.

Benchmark's structure reinforces this: each partner gravitates toward different sectors and founder types, and the portfolio reflects that range rather than a single firm-wide thesis.