Theses›Datacentre & AI compute›Photodetection

Photodetection

Datacom photodetection (Ge-on-Si photodiodes, III-V APDs, the receiver/TIA front-end) is real and necessary but not standalone-investible.

  • Bear
  • Shut
  • $9.9B by 2035

The callThe detector is a free device in every silicon-photonics foundry kit, high-volume and earning nothing on its own; only an optical engine around it keeps a margin.

D1D2R1R2R3C1foragainst D1D2R1R2R3C1foragainst
Fig. 1The balance. Drivers and catalysts against restraints, each block sized by its weight; catalysts hang dashed because they have not happened yet, sized by their odds. It tips toward the restraints.
  1. D1Optics replaces copper inside the rack at 200G and then 400G per lane, and every new link needs a receiver.
  2. D2At 400G per lane a standard germanium photodiode plus a generic amplifier runs out of bandwidth and sensitivity, so the receiver becomes the limit.
  3. R1The detector is grown on the same silicon-photonics die as everything else, so value goes to the optical engine and package owners.
  4. R2Of about 70 funded optical-component startups, none is a pure datacom detector company, and NVIDIA's $4B capacity lock-up went to lasers, not detectors.
  5. R3The ceiling comp, Hamamatsu at about $5B, is a broad photonics company, and a bottom-up sizing reads about 50x short of a venture-scale outcome.
  6. C1A hyperscaler pulling a merchant receiver supply chain out of the package, the way external lasers came out; nothing dated.end-2027
Rests onthe beliefs underneath
Supported by · confidence
What would prove itdated tests, open and resolved
20272028202920302031today12345predictionswatch 20272028202920302031today12345predictionswatch
Fig. 2Each ring is a dated test: dashed while open, struck through when overdue, ticked or crossed once resolved.
  1. 1Through end-2027 no pure-play datacom photodetector/receiver startup founded 2023+ raises at >=$200M valuation; receivers keep integrating into SiPh platforms.Dec 2027 · open
  2. 2Any datacom detector or receiver startup founded 2023 or later raising at $200M or more, which would reopen it.Dec 2027 · open
  3. 3The receive device becoming the speed limit at faster lane rates, or a detector design becoming a system in quantum or LiDAR.Dec 2028 · open
  4. 4No pure-play datacom photodetector/receiver startup (founded 2023-2026) reaches a >$500M primary-funding valuation by 2030; outcomes are integration into a SiPh transceiver, strategic acquisition by an incumbent, or sub-scale.Dec 2030 · open
  5. 5Whether any such company reaches a $500M valuation.Dec 2030 · open
Expressionslisted, private, technologies
  • 6965.THamamatsu Photonics
    • full
     Photomultipliers + SPADs + scientific sensors.
  • COHRCoherent Corp
    • cheap
     InP laser-chip in-house + transceivers.
  • LITELumentum
    • cheap
     200G/lane EML lasers — only volume shipper.
  • SMTCSemtech
    • fair
     Sells TIAs, CDRs and drivers for 800G and 1.6T optical transceivers to module makers; data-centre segment ~$100M a quarter (not receive-only).
  • MTSIMACOM Technology Solutions
    • full
     Sells TIAs, photodiodes and other receive-chain analog chips for data-centre optical modules.
  • AVGOBroadcom
    • cheap
     Sells photodiodes, TIAs and DSPs for optical transceivers and its own optical engines, to module makers and hyperscalers.
Private · 2 private companies.

Moon Photonics · Aeluma

1 more private company tracked, not shown

Technologies
Who has already bet on itprivate rounds, from public sources

$26M raised by 1 company, $12M of it in the last three years. Early: Little money in so far; a lead investor would set the pace.

What changedthe record, newest first
  • 23 Jun 2026When a hyperscaler pays billions to lock optical-component supply, it pays for the lasers, not the detectors. The detector never left the chip, which is exactly why it is not a separately-investable layer.
  • 17 Jun 2026A bottleneck only becomes investable when it de-integrates into a separately-sourced layer (lasers) or becomes a system in another market (single-photon sensing) — datacom photodetection does neither, so it stays a commodity inside the silicon-photonics receiver.
Reading
Show the reading