Cerebras Systems: The AI Chip Company That Made More Money But Saw Its Stock Drop

What Happened? (The Big Picture)

Imagine you run a lemonade stand. You sell way more lemonade than you thought you would. You even tell everyone, "Hey, next year I’m going to sell THREE TIMES as much!" But then… your friends who invested in your stand get worried and sell their shares.

That’s basically what happened to Cerebras Systems (a company that makes special computer chips for AI) in their second earnings report since going public.

Quick Timeline:

• May 2026: Cerebras had its IPO (went public on the stock market) at $185/share
• July 2026: Reported Q2 earnings → Stock dropped ~14% after hours
• Current price: $262.06 (still 42% higher than IPO price)

The Numbers Made Simple

Here’s the "report card" for the last quarter (April–June 2026):

The Good News (Raised Guidance!)

Cerebras raised its full-year forecast:

• Old target: $855–865 million core revenue
• New target: $880–890 million core revenue
• Next quarter margin: 38–40% (getting more profitable per chip!)

Big Partnerships Announced

1. AMD (Nvidia’s rival) – products coming later this year
2. OpenAI – will use Cerebras chips for GPT 5.6-Sol
3. Cloud business – made $126 million last quarter alone

Why Did the Stock Drop If Things Are Good?

This confuses grown-ups too! Here are the main reasons:

1. The "Paper Loss" Scared People

That $450.5 million loss looks terrifying. But $386.6 million of it was just stock compensation – giving employees shares instead of cash. No actual money left the bank! It’s like promising your lemonade stand workers "future lemonade credits" instead of paying them today.

2. Investors Expected Even More

When a hot AI company goes public, people expect perfect results. Beating estimates wasn’t enough – they wanted a blowout.

3. "Show Me the Profits" Mood

Wall Street is getting picky. They want to see real profits (GAAP), not just "adjusted" numbers. Cerebras is still investing heavily to grow.

4. Nvidia’s Shadow is Huge

Nvidia controls ~80% of the AI chip market. Investors worry: Can Cerebras really steal enough customers?

What Makes Cerebras Special? "Fast Inference"

Most AI chips (like Nvidia’s) are great at training – teaching AI models by feeding them massive data. That takes weeks/months.

Cerebras focuses on INFERENCE – actually using the trained model to answer questions, generate text, etc.

The "Low Latency" Advantage

• Latency = delay between asking and getting an answer
• Cerebras chips = super fast answers (milliseconds!)
• Why it matters: Chatbots, real-time translation, gaming AI, trading bots – all need instant responses

CEO Andrew Feldman says: "AI demand is through the roof. Companies are paying a PREMIUM for fast inference."

Premium Pricing = Better Margins

Because their chips are specialized for speed, Cerebras charges more → gross margins improving to 38–40%.

IMPORTANT POINTS TO REMEMBER

1. Revenue is growing fast – $210M this quarter, targeting ~$885M for the year
2. The big "loss" is mostly accounting – stock compensation, not cash burning
3. $25.4 BILLION in future orders (Remaining Performance Obligations) – massive backlog!
4. Revenue expected to TRIPLE next fiscal year – management is very confident
5. Partnerships with AMD & OpenAI validate their technology
6. Stock still UP 42% since IPO – despite the recent drop

Looking Ahead: What’s Next for Cerebras?

1. Manufacturing Scale Benefits

CEO Feldman explains: "We’ll manufacture more efficiently. Better pricing on parts. Spread factory costs over more chips." → Costs per chip should drop.

2. Cloud Business Growth

Their "rent our chips by the hour" cloud made $126M last quarter. As more companies need fast AI without buying hardware, this could explode.

3. OpenAI Partnership = Huge Validation

If OpenAI (maker of ChatGPT) uses Cerebras for GPT 5.6-Sol, other big AI labs will follow.

4. The Nvidia Battle

Cerebras doesn’t need to beat Nvidia everywhere – just own the "fast inference" niche. That could be a massive, profitable market.

Summary

Bottom line: Cerebras is executing well on a promising niche (fast AI inference). The stock drop reflects short-term sentiment, not fundamental breakdown. But it’s still a risky, early-stage company in a fiercely competitive market.

FAQ

1. What is "inference" in AI? Is it different from "training"?

Yes! Think of it like a student:

• Training = Studying for the exam (reading textbooks, practicing problems) – takes a long time, needs massive compute
• Inference = Taking the exam (answering questions using what you learned) – needs to be fast and efficient
  Cerebras chips are built for super-fast exam-taking.

2. Why does stock compensation count as a "loss" if no cash is paid?

Accounting rules (GAAP) say: When you promise shares to employees, that’s an expense. It reduces reported profit on paper, but cash stays in the bank. It’s like a lemonade stand promising "free lemonade for life" to workers – costs future revenue, not today’s cash.

3. What are "Remaining Performance Obligations" ($25.4B)?

Fancy term for signed contracts for future delivery. Customers have legally committed to buy $25.4B worth of chips/services. It’s like having a waiting list with deposits for 5+ years of lemonade.

4. Can Cerebras really compete with Nvidia?

Not head-on everywhere. Nvidia dominates training and general AI. But Cerebras targets a specific niche: ultra-low-latency inference. If they own that niche (like a specialty tool vs. a Swiss Army knife), they can be very profitable without beating Nvidia overall.

5. Should I buy Cerebras stock now?

Not financial advice! Consider:

• Pros: Huge backlog, rising margins, unique tech, big partnerships, revenue tripling forecast
• Risks: Still unprofitable (GAAP), Nvidia competition, stock-based dilution, volatile new public stock
• Best for: Long-term, risk-tolerant investors who believe in the "fast inference" thesis

Article based on CNBC reporting of Cerebras Systems Q2 FY2026 earnings (released July 2026). All figures in USD.