Our guides recommend boring index funds inside an ISA. That remains our starting point for UK beginners. But there is a technology theme we have been researching that we think is genuinely interesting to learn about — even if most people should never invest in it directly. It is called co-packaged optics (CPO).
What Is CPO in Plain English?
Today, data centres use separate plug-in modules called optical transceivers to convert electrical signals into light (and back again) for high-speed communication between servers. These transceivers sit in the front panel of networking switches — you can physically pull them out and replace them.
Co-packaged optics takes a different approach: instead of separate pluggable modules, the optical components are integrated directly onto the chip package itself. The light source, modulator, and detector are moved as close to the silicon as possible — sometimes onto the same substrate.
Think of it like the difference between plugging a USB Wi-Fi dongle into your laptop versus having Wi-Fi built into the motherboard. Same function, but the integrated version is more efficient, more compact, and uses less power.
Why Does CPO Matter?
AI data centres are growing at an extraordinary pace. Each new generation of AI model requires more GPUs, and those GPUs need to communicate faster. The current approach — pluggable optical transceivers — is hitting physical limits:
- Power consumption: Pluggable transceivers waste energy converting signals between electrical and optical domains multiple times. CPO reduces this by shortening the electrical path.
- Bandwidth density: There is only so much physical space on the front panel of a switch for pluggable modules. CPO frees up that space constraint by moving optics onto the chip.
- Signal integrity: The longer an electrical signal travels before being converted to light, the more it degrades. CPO minimises that distance.
- Thermal efficiency: Less wasted energy means less heat to manage — critical in data centres already struggling with cooling.
In short: as AI clusters scale to hundreds of thousands of GPUs, the networking becomes the bottleneck. CPO is one of the leading candidates to solve that bottleneck.
The Timeline: When Is CPO Coming?
CPO is not a 2026 technology. The industry consensus timeline looks roughly like this:
- 2025-2026: Lab testing, prototype demonstrations, early samples from major vendors
- 2027-2028: First commercial deployments in hyperscaler data centres (Google, Meta, Microsoft)
- 2029-2030: Broader adoption across enterprise data centres
This matters for investors because it means CPO is still a forward-looking theme. Companies positioning for CPO are investing heavily now in R&D, but the revenue will not arrive for 1-3 years. That creates uncertainty — and therefore risk.
Who Benefits from CPO?
Several categories of company stand to benefit if CPO is adopted at scale. These are mentioned for educational purposes — they are not recommendations:
Silicon Photonics Companies
Broadcom is widely considered the CPO leader. Its silicon photonics programme has been running for over a decade, and it supplies switch ASICs to hyperscalers who are actively planning CPO deployments. Broadcom's approach integrates photonic components onto silicon using existing semiconductor manufacturing processes.
Marvell designs custom silicon for hyperscalers and has its own optical DSP capabilities. It is positioned to offer CPO-ready solutions alongside its networking silicon.
Laser Source Companies
Even with CPO, you still need a light source. External laser sources (ELS) are the approach most CPO designs use — the laser sits outside the chip but close to it. Companies in this space include:
- Coherent Corp — laser sources, indium phosphide expertise, strong data centre relationships
- Lumentum — laser chips used in high-speed optical systems
- Sivers Semiconductors — Swedish company working on silicon photonics and laser integration
What Happens to Pluggable Transceiver Companies?
This is the risk side of the CPO story. If CPO replaces pluggable transceivers over time, companies whose revenue depends on selling those pluggables could face declining demand. However, many are adapting:
- Pivoting to supply components for CPO systems (lasers, photodetectors, modulators)
- Pluggable demand will likely remain strong through 2027-2028 regardless, due to 800G and 1.6T upgrade cycles
- The transition will be gradual — pluggables and CPO will likely coexist for years
The "Wait for Picks and Shovels" Analogy Applied to CPO
Some investors apply the picks-and-shovels framework to CPO: rather than guessing which company will "win" CPO, invest in the companies that supply components to all CPO implementations — laser sources, packaging materials, test equipment.
The logic: no matter whose CPO design wins, everyone needs external laser sources. Therefore, laser source companies might be the picks-and-shovels play.
The counter-argument: laser source companies are often small-cap, thinly traded, and burning cash while they wait for CPO revenue to arrive. If CPO is delayed or a different architecture wins, these stocks could collapse.
Why Most UK Beginners Should Not Invest in CPO Directly
We want to be transparent about this. CPO-related stocks carry multiple layers of risk:
- Technology risk: CPO might not be adopted as fast as expected, or a competing approach might win
- Timing risk: Revenue is 1-3 years away for most companies; meanwhile, you are holding speculative positions
- Size risk: Many CPO-focused companies are small or mid-cap with high volatility and thin liquidity
- Complexity risk: Understanding whether a specific laser source company will benefit from CPO requires deep technical knowledge that most retail investors do not have
- Concentration risk: Putting money into a single theme (CPO) within a single sector (optical networking) within a single trend (AI) is extremely concentrated
FAQ
What are co-packaged optics (CPO)?
Co-packaged optics is a technology that moves optical connections directly onto the chip package instead of using separate pluggable transceiver modules. This reduces power consumption, increases bandwidth density, and shortens the physical distance data needs to travel — all critical for next-generation AI data centres.
When will co-packaged optics be widely adopted?
Industry consensus points to 2027-2028 for initial commercial deployments at scale, with broader adoption following through 2029-2030. Some early CPO implementations are being tested in 2025-2026 in hyperscaler labs, but mass production is still ahead.
Which companies benefit from co-packaged optics?
Companies in the CPO supply chain include Broadcom (silicon photonics, switch ASICs), Marvell (custom silicon, optical DSPs), Coherent (laser sources, optical components), Lumentum (laser chips), and Sivers Semiconductors (silicon photonics). These are educational mentions, not investment recommendations.
Is CPO a threat to existing optical transceiver companies?
Potentially yes. CPO could reduce demand for traditional pluggable transceivers over time. However, many existing optical companies are pivoting to supply components for CPO systems — laser sources, modulators, and photodetectors are still needed, just in a different form factor. The transition will be gradual.
Should UK beginners invest in CPO-related stocks?
For most UK beginners, no. CPO-related companies are highly speculative, often small or mid-cap, and the technology timeline is uncertain. A global index fund inside a Stocks and Shares ISA is far more appropriate for beginners and provides exposure to the broader tech sector without single-stock risk.