Nomura: 1.6T Optical Transceiver DSP Supply Gap to Persist Through H2 2027
nashnova research
A Nomura expert call reveals the 1.6T DSP shortage will widen from 3–4 million units in 2026 to over 20 million in 2027, driven by 3nm capacity constraints rather than demand swings — handing incumbent suppliers sustained pricing power.
How large is the 1.6T DSP gap?
2026 demand is projected above 20 million units, but actual shipments will reach only 16–17 million — a gap of roughly 3–4 million.
In 2027 the mismatch widens: demand climbs to 60–70 million units, supply caps at 40–50 million, leaving a shortfall of over 20 million.
This means → the gap is not closing — it is doubling. Meaningful capacity relief arrives only in H2 2027.
Why is 1.6T short while 800G is not?
1.6T DSP requires 3nm process technology, where fab capacity is tight and yields remain low — a structural bottleneck, not a scheduling issue.
800G DSP uses mature 7nm and 5nm nodes with ample supply: shipments run ~25–26 million in 2025, ~50 million in 2026, and leap to 120–150 million in 2027.
In plain terms = the more advanced the process node, the scarcer the fab line — and 1.6T sits on the tightest one.
Why can't new entrants break in?
DSP barriers span three layers: advanced process (sub-7nm, with prohibitive tapeout costs), signal-processing algorithms (forward error correction, bit-error-rate control, equalization, dispersion compensation), and power and thermal management.
Customer qualification takes roughly two years; switching costs are high, so end-users stick with incumbents.
This means → even second-tier suppliers that can produce 800G parts struggle to win sockets without an established customer ecosystem. Chinese players such as Sitrus and Joywell have yet to ship at scale.
Where are prices heading?
Current contract prices: 800G DSP at $45–70, 1.6T DSP at $120–150; spot prices run 10–20% above contract.
By 2027, as capacity ramps, 800G contract prices are expected to fall ~10%; 1.6T prices will decline less, given the wider gap.
This reflects a direct link between supply tightness and pricing power — the bigger the shortage, the smaller the room to negotiate.
Could NPO/CPO upend DSP's role?
Under NPO/CPO — co-packaged optics that embed optical modules directly into switch chips — the DSP is eliminated. Its functions shift to TIAs (transimpedance amplifiers), driver chips, and the switch ASIC itself.
That raises the bar for TIAs and drivers: they must deliver high linearity, low noise, and low power simultaneously. Leading suppliers are shipping 200G per lane; 400G is still in development.
Chinese vendors are volume-shipping 100G parts and sampling 200G, but a clear gap remains in product stability and customer qualification versus overseas peers.
What would invalidate the shortage thesis?
Experts flag three downside triggers: ① 2027 1.6T supply materially exceeds the 40–50 million assumption; ② DSP price declines significantly surpass the ~10% baseline; ③ NPO/CPO adoption accelerates, compressing DSP demand earlier than expected.
The critical verification window is H1 2027 — the period before major capacity comes online. Quarterly earnings and optical-module shipment data from leading suppliers will be the key checkpoints.
In plain terms = the current thesis rests on "capacity doesn't arrive until H2" — if that premise loosens, the entire gap logic needs reassessing.
市场有风险，内容仅供研究参考，不构成投资建议。