CPO (Co-Packaged Optics) and high-speed optical modules are hot right now. Everyone talks about AI driving demand. But many stocks in this space trade on hype, not real business strength. Here is how to tell the difference.
Real CPO demand comes from confirmed data center orders. Hype comes from press releases with no revenue yet.
Start with the basics. A company can talk about future CPO products, but what matters is money in the bank. Check if revenue is growing from actual shipments, not just pilot programs.
| Red Flag (Avoid) | Green Flag (Consider) | How to Check |
|---|---|---|
| No revenue from CPO/optical products yet | Revenue from 800G or higher modules growing | Read quarterly earnings reports |
| CEO talks only about "AI opportunity" | CEO breaks down customer names and order sizes | Listen to earnings call transcripts |
| Valuation based on 2027 or 2028 forecasts | Profitable today or clear path to profit | Check trailing and forward P/E ratios |
| Single customer is 80%+ of sales | Diversified customer base across cloud providers | Review 10-K/20-F filings |
| Announces "partnerships" with no dollar values | Disclosed supply contracts with confirmed volumes | Search SEC filings and press releases |
Many investors get burned buying on "partnership" news. A real supplier deal has numbers attached. If you cannot find dollar amounts or unit volumes, be careful.
Company A announced a "strategic partnership" with a major cloud provider for CPO technology in March 2024. The stock jumped 40%. Three quarters later, revenue was flat. There was no follow-up deal. The partnership had no minimum purchase commitment.
Company B, meanwhile, quietly reported $200 million in 800G module sales to two named hyperscale customers. Its stock rose slower but held gains because the money was real.
Next, look at who the real buyers are. The optical module market is driven by a small group: Amazon, Google, Microsoft, Meta, and some large Chinese cloud firms. If a supplier names these as customers with multi-year deals, that is stronger than vague claims.
| Indicator | Speculative (Weak) | Confirmed (Strong) |
|---|---|---|
| Customer disclosure | "A leading AI company" | Named in supplier filings or confirmed by buyer |
| Product generation | Sampling 1.6T modules with no date | Shipping 800G at scale, 1.6T qualified |
| Capacity plans | Announces factory with no timeline | Factory operational, capex (Capital Expenditure) guided up |
| Competitive position | Claims "industry leading" without data | Market share numbers from third-party trackers |
| Backlog visibility | "Strong interest" or "pipeline" | Disclosed backlog or lead times extending |
Note: Third-party trackers include LightCounting, Omdia, and Yole Développement. They publish market share data for optical transceivers.
Not every CPO player is equal. Some make the optical engine, some assemble modules, and some just design chips. The closer to the core IP (Intellectual Property), the more defensible the business.
Module assemblers without proprietary tech face price pressure and lower margins.
Understanding where a company sits in the value chain helps you avoid middlemen. Middlemen get squeezed when supply is plentiful or when customers build in-house.
| Value Chain Layer | Examples | Typical Gross Margin | Hype Risk Level |
|---|---|---|---|
| Optical chips / DSP (Digital Signal Processor) | Marvell, Broadcom, Inphi | 60-70% | Lower |
| Laser sources (EML, VCSEL) | Coherent, Lumentum | 40-55% | Medium |
| Module assembly (narrow product set) | Some tier-2 Chinese suppliers | 25-35% | Higher |
| Module assembly (broad portfolio, scale) | Coherent, Accelink, Fabrinet | 30-45% | Medium |
| Subsystem / CPO integration | Arista, Cisco (in-house); start-ups | Varies widely | High for start-ups |
If a company claims CPO leadership but mainly does assembly, dig deeper. Ask: what do they own that a competitor cannot copy in six months?
Start-up C raised $100 million for CPO technology. It had no products in production. Its pitch relied on prototypes tested in labs. The technology worked, but scaling to millions of units was unproven. The company burned cash for three years and never reached profitability.
Established player D spent ten years building relationships with hyperscalers. Its CPO products were delayed, but its base business in 400G and 800G modules kept revenue stable. When CPO orders finally came, it captured them because it was a trusted supplier already.
Timing matters in tech investing. Being right too early is the same as being wrong for your returns.
| Metric | Weak Signal | Strong Signal |
|---|---|---|
| Revenue growth (year-over-year) | <10% or declining | >20% with accelerating trend |
| Operating margin | Negative with no clear path | Positive or improving sequentially |
| R&D spend ratio | Very low (not investing) or very high (desperate) | 15-25% of revenue, steady |
| Free cash flow | Consistently negative | Positive or turning positive |
| Stock-based compensation | Growing faster than revenue | Stable or shrinking as % of revenue |
| Debt / Cash position | High debt, burning cash fast | Net cash or manageable leverage |
One simple test: is the company getting better as it gets bigger? If revenue grows but margins shrink every quarter, something is wrong. Either competition is fierce, or the "growth" comes from low-margin business.
Another trap is confusing total addressable market (TAM) with actual sales. A TAM of $50 billion means nothing if the company has 0.1% market share and no plan to win more.
Company E presented a slide showing the CPO market reaching $12 billion by 2028. Its own projected revenue for that year was $500 million. But it had no products shipping in 2024 and no firm orders. The TAM looked huge, but its path to capture it was unclear.
Company F showed a smaller TARX of $4 billion, but it already had $300 million in revenue and named contracts for $200 million more. The smaller market was real. Its slice was growing.
Even good companies become bad investments if you pay too much. Set a price target based on realistic earnings, not best-case scenarios.ocoaot
If a stock trades at 50x sales with no profit, ask what has to go right for you to make money. Usually, everything.
When evaluating any CPO or optical module stock, run a stress test. Cut the 2025 revenue forecast by 30%. Cut the margin by 5 percentage points. Does the stock still make sense? If not, the hype is priced in. You have no room for error.
Key Takeaways
| Key Point | What It Means | Action Item |
|---|---|---|
| Revenue beats promises | Real money today beats big promises tomorrow | Read last 4 quarters of earnings; ignore forward-only pitches |
| Name the customers | Vague "AI demand" hides weak traction | Find named customers and contract sizes in filings |
| Know your place in the chain | Assembly without IP gets crushed | Map the company's role; favor chip and engine makers |
| Check the cash | Profits and cash flow reveal true health | Review free cash flow and operating margin trends |
| Price for imperfection | Hype stocks assume best case | Stress-test your valuation; only buy with a margin of safety |