Aeva ($AEVA) launched an Optical Connectivity segment in Aug 2026 — a hyperscaler JDA worth hundreds of millions a year from 2027. The lidar business, the SOA tech, the $SIVE.ST supply chain, and the risks.
More Macroplane supply-chain guides · Baskets · Investment theses · Macro trends
Aeva Technologies ($AEVA) is a 2021 SPAC that spent five years being valued as a lidar company and, in August 2026, started being valued as something else: a supplier of high-power optical sources into AI data center interconnect. The Q2 2026 results named a new "Optical Connectivity" segment, a joint development agreement with an optical engine maker serving a top-four hyperscaler, and a revenue opportunity management sized at multiple hundreds of millions of dollars a year. The lidar business is still the revenue base. The optical business is why the stock re-rated.
👉 See it in context: Aeva now sits upstream of the same buildout as the Photonics / CPO basket and the Silicon Photonics & Optical Interconnects trend.
Two things now, and they share a component.
The original business is 4D lidar. Aeva uses frequency-modulated continuous wave (FMCW) sensing rather than the time-of-flight approach most lidar vendors use. FMCW sends out a continuous, frequency-swept laser beam and reads the beat frequency of the return, which gives you per-point velocity as well as range. That is the "4D" part. It costs more optical engineering than time-of-flight, and it needs a much better laser.
That laser requirement is the bridge. To make FMCW work at automotive range, Aeva had to build high-power, narrow-linewidth optical sources and semiconductor optical amplifiers (SOAs) in volume, at automotive temperature and reliability grades. In January 2026 the company published the specs: output above 28 dBm with wall-plug efficiency above 20% at temperatures up to 50 °C, fabricated in commercial semiconductor fabs rather than a specialist III-V line.
Those are data center numbers. Wall-plug efficiency at elevated temperature is precisely the constraint that matters when the laser sits next to a switch ASIC instead of in a plug-in module on the faceplate.
On the Q2 2026 call (5 August 2026), management laid out the structure:
| Item | Disclosure |
|---|---|
| Counterparty | A "leading provider of high speed optical engines" (unnamed) |
| End customer | A "major hyperscaler", described as top three or four (unnamed) |
| Aeva's content | High-power optical sources and silicon photonics for near-packaged optics |
| Revenue at scale | "Multiple hundreds of millions of dollars" annually |
| Unit volume | "Multiple millions of units" annually |
| Initial volumes | Second half of 2027 |
| Production ramp | 2028 |
Two details matter more than the headline number.
First, this is near-packaged optics (NPO), not co-packaged optics. NPO puts the optical engine on the same board as the switch ASIC but not on the same package. It is the intermediate step between today's pluggable transceivers and full CPO, and it is easier to manufacture and service. Aeva is selling into the transition, not betting on the endpoint.
Second, Aeva is a component supplier two tiers down. It sells the light source to an optical engine maker, which sells the engine to the hyperscaler. That is a durable position if the engine maker wins, and an invisible one if it does not. Neither party is named, which is the single largest unverifiable claim in the story.
The reason a lidar company can walk into this market is that co-packaged and near-packaged optics moved the bottleneck from the switch silicon to the light.
A pluggable 1.6T transceiver carries its own laser inside the module, on the faceplate, in relatively cool air. Move the optics next to a switch ASIC dissipating a kilowatt and the laser now runs hot, which drops its efficiency and its lifetime at the same time. Broadcom's Tomahawk 6 "Davisson" claims a 70% reduction in optical interconnect power against pluggables, and NVIDIA's Quantum-X Photonics switch pulls 3.95 kW against roughly 7 kW for the pluggable-based equivalent. Those savings only hold if the light source survives the thermal environment.
That is a laser and amplifier problem, and there are not many suppliers who can do it at volume. The names that come up are Lumentum ($LITE), Coherent ($COHR), and a short tail of specialists.
Aeva's claim to a seat at that table got independent support in November 2026, when the semiconductor analyst behind Irrational Analysis published a technical teardown of Intel's CPO work. The piece is harsh on Intel, arguing that four-wave mixing in its two-stage amplifier design wrecks bit-error-rate performance. In passing it grades the SOA suppliers, and Aeva's design comes out well on failure-in-time reliability. Read the whole thing rather than the pull-quote, but the relevant point for $AEVA holders is that a hostile technical reviewer with no position disclosed rated the amplifier as competitive.
Aeva does not make everything itself, and the dependency runs somewhere useful.
Sivers Semiconductors ($SIVE.ST) supplies high-power DFB lasers and optical amplifiers into Aeva's FMCW systems. Macroplane tracks this as an active supplier relationship with roughly $3M of disclosed value. If Aeva's optical connectivity volumes arrive on the 2027-2028 schedule, Sivers is the second-order beneficiary, and it is a far smaller company than Aeva. We wrote about the Sivers position in more detail in the Sivers and POET risk teardown.
Jabil ($JBL) runs the automated assembly line that builds the Atlas sensor for the Daimler Truck and Bendix programs. Cadence ($CDNS) signed a licensing deal with Aeva in May 2026.
You can walk the full graph on the Aeva company page.
It would be easy to read the optical pivot as a distressed company changing the subject. The Q2 2026 operating updates argue against that:
Q2 2026 revenue was $6.1M, split $2.5M product and $3.6M professional services and NRE. The NRE growth is what you would expect from a company doing paid engineering work for a new segment before any of it ships.
This is a pre-revenue-scale company by any reasonable definition, and the numbers should be read that way.
| Metric | Value |
|---|---|
| Market cap | ~$990M |
| Q2 2026 revenue | $6.1M |
| Q2 2026 GAAP net loss | $79.6M |
| Q2 2026 non-GAAP operating loss | $26.0M |
| Non-GAAP loss per share | $0.41 (vs $0.43 expected) |
| Total liquidity | $302.9M ($177.9M cash and securities + $125M facility) |
| FY2025 R&D | $85.4M |
| FY2025 cash used in operations | $115.1M |
The liquidity position is the thing that changed. A $115M equity raise in June 2026 (on top of a $100M follow-on launched that month) took available liquidity to $302.9M against a roughly $100-120M annual operating burn. That buys the runway to reach the 2027 initial volumes without another raise, which was not true a year ago. It also means holders have been diluted repeatedly, and will be again if the ramp slips.
Note the shape of the loss: the $79.6M GAAP number is much larger than the $26.0M non-GAAP operating loss, so most of it is non-cash. Judge the burn from the cash flow statement, not the net income line.
Four things, in rough order of how much they matter:
The CFO transition adds noise: Saurabh Sinha departs in September 2026, with the VP Corporate Controller taking the role on an interim basis. That is not a thesis-breaker, but a company guiding to a 2028 ramp with an interim CFO is worth watching.
| Company | Role in optical interconnect | Market cap | Data center revenue today |
|---|---|---|---|
| Coherent ($COHR) | Transceivers, lasers, materials | Large cap | Substantial |
| Lumentum ($LITE) | Lasers, DFB arrays, transceivers | Mid cap | Substantial |
| Fabrinet ($FN) | Contract optical assembly | Mid cap | Substantial |
| POET Technologies ($POET) | Optical interposer platform | Small cap | Pre-revenue |
| Aeva ($AEVA) | High-power optical sources, SOAs | ~$990M | Zero until 2027 |
Aeva is priced closer to the option-value end of that table than the revenue end. That is the trade: a real, funded lidar business underwriting a call option on the interconnect market, with the first exercise date in late 2027.
Aeva ($AEVA) designs and manufactures 4D lidar sensors using frequency-modulated continuous wave (FMCW) technology, which measures velocity as well as range for each point. Since August 2026 it also sells high-power optical sources and semiconductor optical amplifiers into AI data center optical interconnect through a new Optical Connectivity segment.
Today it is a lidar company by revenue: all of the $6.1M of Q2 2026 revenue came from lidar products and engineering services. The data center business generates no product revenue until initial volumes in the second half of 2027. The market is pricing the optical opportunity ahead of the cash flows.
Aeva has not named it. Management described the end customer as a "major hyperscaler" in the top three or four, and the direct counterparty as a leading provider of high-speed optical engines. Neither is disclosed, which is the main unverifiable element of the story.
No. Aeva posted a $79.6M GAAP net loss in Q2 2026 and used $115.1M of cash in operations across FY2025. It ended Q2 2026 with $302.9M of total liquidity after a $115M equity raise in June 2026.
Co-packaged optics (CPO) puts the optical engine on the same package as the switch ASIC. Near-packaged optics (NPO) puts it on the same board but in a separate package, which is easier to build and service. Aeva's disclosed opportunity is in NPO. Our co-packaged optics explainer covers the full architecture.
The Photonics / CPO basket tracks the optical interconnect supply chain, including $COHR, $LITE, $FN, $POET and $AAOI. Sivers Semiconductors ($SIVE.ST) is directly exposed as an Aeva supplier.
This is research and education, not financial advice. Aeva is a loss-making small cap whose central thesis depends on undisclosed counterparties and a 2027-2028 schedule. Do your own work.
Aeva ($AEVA) designs and manufactures 4D lidar sensors using frequency-modulated continuous wave (FMCW) technology, which measures velocity as well as range for each point. Since August 2026 it also sells high-power optical sources and semiconductor optical amplifiers into AI data center optical interconnect through a new Optical Connectivity segment.
Today it is a lidar company by revenue: all of the $6.1M of Q2 2026 revenue came from lidar products and engineering services. The data center business generates no product revenue until initial volumes in the second half of 2027. The market is pricing the optical opportunity ahead of the cash flows.
Aeva has not named it. Management described the end customer as a "major hyperscaler" in the top three or four, and the direct counterparty as a leading provider of high-speed optical engines. Neither is disclosed, which is the main unverifiable element of the story.
No. Aeva posted a $79.6M GAAP net loss in Q2 2026 and used $115.1M of cash in operations across FY2025. It ended Q2 2026 with $302.9M of total liquidity after a $115M equity raise in June 2026.
Co-packaged optics (CPO) puts the optical engine on the same package as the switch ASIC. Near-packaged optics (NPO) puts it on the same board but in a separate package, which is easier to build and service. Aeva's disclosed opportunity is in NPO. Our co-packaged optics explainer covers the full architecture.
The Photonics / CPO basket tracks the optical interconnect supply chain, including $COHR, $LITE, $FN, $POET and $AAOI. Sivers Semiconductors ($SIVE.ST) is directly exposed as an Aeva supplier. This is research and education, not financial advice. Aeva is a loss-making small cap whose central thesis depends on undisclosed counterparties and a 2027-2028 schedule. Do your own work.