TLDR
- Cerebras launched the CS-4, a new rack-scale AI inference system built around three WSE-3 Turbo wafer-scale processors
- The company claims CS-4 delivers 30x more tokens per second per user compared to GPU-based systems
- CBRS stock fell over 35% from its IPO price of $185, trading around $218 midday Tuesday, and was down 12.69% on the day
- Q2 earnings showed a loss per share of -$2.98, compared to a profit of $1.91 in the same quarter last year
- Cerebras targets 600 megawatts of computing capacity by end of 2027 and claims 20x throughput improvement is planned
Cerebras Systems (CBRS) dropped 12.69% on Tuesday as the AI chipmaker unveiled its newest server system while reporting a steep quarterly loss that rattled investors.
The stock was trading around $218 midday Tuesday, well below its IPO price of $185 when it began trading at $350 in May. That puts it more than 35% below its opening day price.
The Q2 results were tough. Cerebras posted a loss per share of -$2.98, a sharp reversal from the $1.91 profit it reported in the same quarter last year. Better-than-expected Q3 guidance wasn’t enough to reassure Wall Street.
The launch of the CS-4 came on the same day. The system is a rack-scale unit powered by three WSE-3 Turbo processors, which Cerebras calls the largest AI semiconductors ever built, each packing 4 trillion transistors.
Cerebras says the CS-4 delivers 30x more tokens per second per user compared to GPU-based alternatives. That’s the headline number the company is leaning on hard.
The performance claim comes down to chip design. Cerebras uses static random-access memory (SRAM) rather than the dynamic RAM (DRAM) found in most competitor chips. SRAM is faster but more expensive and complex, which is why it only works on Cerebras’ dinner-plate-sized wafers.
Because the processor is one large wafer rather than multiple linked chips, data travels shorter distances. That cuts latency compared to systems from Nvidia or AMD, where data has to move between separate chips.
CS-4: What’s New
The CS-4 is built on TSMC’s 5-nanometer manufacturing process. It includes new networking components designed to speed data movement and ships with 50% fewer components than previous systems, which the company says simplifies data center setup.
Chief Technology Officer Sean Lie said the reduced component count would speed up data center construction. The system is available in Q3 2026.
CEO Andrew Feldman laid out the roadmap at a media briefing in San Francisco. He said the company expects to deliver 600 megawatts of computing power by end of 2027 and targets a 20x increase in throughput alongside a 4x speed improvement by that same date.
“Every aspect of the design has been optimized to deliver the highest speeds with massive throughput,” Feldman said.
What This Means for the Nvidia Fight
Cerebras is squarely targeting Nvidia’s dominance in AI inference workloads. Nvidia (NVDA) also fell on the news, dropping 2.34% on Tuesday.
Cerebras also runs its own AI cloud service, renting access to its chips to customers rather than selling hardware alone.
Another generation of the chip and server system is planned for 2027.
Stop guessing and start investing with confidence. KnockoutStocks gives you the AI insights, market intelligence, and stock research you need to spot opportunities, cut through the noise, and make smarter investment decisions — all in one powerful platform.
Sign up today and get 50% OFF full access to our premium stock picks.
Simply use coupon code SPECIAL50 at checkout to claim your exclusive discount.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
