Cerebras systems (NSDQ:CBRS) Dont buy this IPO just yet

AI chip hype, OpenAI backlog risk and IPO supply pressure collide

When Cerebras systems went public last week, we noted that it was one of the largest tech IPOs in recent memory. The stock priced at US$185 per share before opening at US$350, which tells you how intense the demand was from day one.

This was always a euphoria trade. Cerebras sits right in the middle of the AI inference boom, and investors were willing to pay a steep premium for one of the few pure-play AI chip names coming to market.

But the stock is already down more than 20%, and the risks should not be ignored.

The biggest one, as we flagged in our analysis last week, is the US$24.5bn revenue backlog. More than 80% of that backlog is tied to OpenAI.

That creates a clear concentration risk. OpenAI is committing to enormous infrastructure spending while competition is rising quickly, particularly from Anthropic. The concern is not that OpenAI lacks ambition. The concern is whether every major compute commitment it signs today will prove economically durable over time , because the one with NVDA didn’t !

For Cerebras, that makes the backlog less clean than the headline number suggests. The market may love the AI exposure, but investors still need to ask how much of that demand converts into long-term, high-quality revenue.

Raises US$3.5bn as AI Chip Euphoria Builds

CBRS is one of the more concentrated plays on the AI inference boom. That creates strong upside when sentiment is hot, especially after a major IPO, but it also makes the stock highly sensitive to pressure when the market starts questioning the quality of that growth.

Cerebras went public on 14 May, raising up to US$3.5bn by selling 28m shares at an estimated range of US$115 to US$125 per share. That gives investors a sense of the valuation large institutions were willing to pay for direct exposure to AI infrastructure.

The chip euphoria is tied to the market narrative that processors used to run AI models are becoming a critical part of the next growth cycle. Cerebras has built an unconventional chip architecture designed to improve the speed and latency of running AI workloads.

Its core product is a wafer-scale processor that is around 58x larger than Nvidia’s B200 chip. That scale is the point of difference. Instead of stitching together thousands of smaller chips, Cerebras is trying to process more AI compute on a single, much larger wafer-scale engine.

Smarter IPO Lock-Up, But Supply Risk Remains

Many investors focus on the technology and growth profile, but in the near term, the IPO capital structure may be just as important.

In a normal IPO, insiders, including founders, employees and early backers, are usually locked up for a set period. Once that lock-up expires, often after 6 months, the market can suddenly face a wave of new supply. If too many insiders sell at once, the stock can come under heavy pressure.

Cerebras is taking a different approach. Instead of releasing insider shares all at once, the company is staggering the supply in tranches.

That is a cleaner structure than a standard IPO. It should reduce the risk of one large supply shock and may also lower the incentive for insiders to rush for the exit, especially after such a hot listing.

But it does not remove the dilution risk. It simply spreads it out.

Investors should mark the Q1 and Q2 prints carefully, because that is when the next waves of supply start to hit the market.

We wouldn’t be buying this stock just yet

Cerebras has an interesting growth story and it sits perfectly inside the AI euphoria, particularly around processor design and AI infrastructure. The IPO timing was almost perfect.

The issue is valuation. The stock is priced for a lot to go right, which is why investors are likely to see sharp swings like the one we saw today.

On that basis, we rate CBRS a sell.

The company still needs to prove it can compete at scale, sell its processors at attractive margins and convert its massive contracted backlog into real revenue. The current multiple already assumes much of that happens.

For now, the risk profile is too high. We would want to see more diversified demand, stronger proof of margin durability and clearer evidence that the backlog can translate into high-quality, repeatable revenue before taking a more constructive view.

Investors can find more coverage of ASX-listed technology names and global AI stocks here at Stocks Down Under.