POET
POET Technologies
Summary
What they do:
Develop co-packaged optical (CPO) engines using a proprietary optical interposer platform that integrates lasers, modulators, and detectors on a single chip — a pre-revenue technology company sitting at Layer 11 as the most speculative name in the optical stack, targeting the elimination of discrete pluggable transceivers by putting optics directly on the switch package.
Why they matter:
If co-packaged optics work at scale, they could fundamentally restructure the optical transceiver market by eliminating pluggable modules — reducing power consumption by 50%+ and enabling the bandwidth density required for million-GPU AI clusters — and POET's optical interposer is the most advanced CPO platform with production orders in hand.
Recent performance:
TTM revenue ~$1.07M (essentially pre-revenue). $5.6M in initial production orders from two customers. $430M cash raised. 30,000+ optical engine shipments targeted for 2026. Stock at ~$6.90, market cap ~$1.1B.
Our Verdict
The most advanced co-packaged optics platform with $430M in cash, production orders, and a 30,000-unit 2026 shipment target — but with only $1M in TTM revenue, unproven manufacturing scale, and 5+ years of execution risk, this is a venture-capital-style bet on whether CPO replaces pluggable transceivers.
Structural trends
Structural
55
/ 100
Moat
3/10
Optical interposer IP (unproven)
AI Exp.AI Exposure
Pure Play~100% AI
Play Type
SpeculativeAI Growth
N/A
Rel. Value
36
FAIRPriceLIVE
$6.71
-8.08%
Live via Yahoo Finance · refreshes every 5 min
Market Cap
$1.0B
P/E Ratio
N/A
P/S Ratio
953.3x
52W High
$9.41
52W Low
$3.58
52W Chg
87.4%
Beta
0.37
POET Technologies is developing co-packaged optical engines built on a proprietary optical interposer — a photonic integrated circuit platform that combines lasers, modulators, photodetectors, and electronic driver/receiver chips on a single substrate. The vision is to eliminate the pluggable transceiver module entirely, embedding optical connectivity directly onto the switch ASIC package. If it works at scale, CPO could reduce per-port power consumption by 50%+, increase bandwidth density, and dramatically lower the cost of optical interconnects.
The company is headquartered in Toronto with manufacturing ramping in Malaysia. POET's Infinity optical engine platform is the current product family, with 800G engines in initial production and 3.2 Tbps engines in co-development with partners. The company has received $5.6M in initial production orders from two customers and targets shipping 30,000+ optical engines in 2026. Revenue is negligible — $1.07M in all of 2025 and $341K in Q4 2025.
POET raised $430M+ in capital, providing substantial runway for a pre-revenue company. The strategic thesis is that as AI clusters scale from 100,000 to 1 million GPUs, the power and density constraints of pluggable transceivers become untenable — each pluggable module consumes 15–25 watts, and a million-GPU cluster would require millions of modules. CPO eliminates the pluggable form factor entirely, integrating optics at the package level and reducing per-port power by half or more.
The competitive landscape for CPO includes Broadcom (which has its own CPO development program), Intel (silicon photonics integration), Ayar Labs (photonic I/O tiles), and Ranovus (quantum dot laser CPO). POET's differentiation claim is its optical interposer, which uses a passive alignment approach that avoids the active alignment precision required by competitors — potentially enabling lower manufacturing costs at scale. But this claim is unproven at commercial volumes. The 30,000-unit 2026 target is tiny relative to the 24M+ pluggable transceiver units shipping in the 800G+ market, and manufacturing yield at scale is the critical unknown.
Supply Chain Dependencies
The Catch
POET is a pre-revenue company valued at $1.1B on the promise of a technology that has never been manufactured at commercial scale. The optical interposer concept is elegant — integrating lasers, modulators, and detectors on a single platform to eliminate pluggable transceivers — but elegant concepts and commercial products are separated by manufacturing yield, thermal management, reliability testing, and cost parity that can take 5–10 years to bridge. POET's 30,000-unit 2026 target is less than 0.1% of the pluggable transceiver market. Broadcom, Intel, and Ayar Labs are all pursuing CPO with deeper engineering teams and larger budgets. If any of them achieves CPO first, POET's optical interposer becomes one of many approaches rather than the winning platform. And the fundamental timing question remains: pluggable transceivers may prove sufficient through 2030 at 3.2T speeds, in which case CPO adoption is pushed beyond POET's cash runway. The company has $430M — enough for 3–4 years. If CPO adoption slips to 2030+, POET faces dilutive financing or shutdown before the market arrives.
If They Win
If co-packaged optics becomes the standard for AI clusters by 2028 — if NVIDIA mandates CPO for its next-generation platform, if hyperscalers conclude that pluggable transceivers cannot scale to million-GPU clusters, and if POET's optical interposer proves to be the lowest-cost, highest-yield CPO manufacturing approach — then POET becomes the TSMC of photonic integration. The optical interposer becomes the standard platform on which all CPO engines are built. Every switch ASIC in every AI data center contains a POET engine. Revenue reaches $500M+ by 2029. The company is acquired by NVIDIA, Broadcom, or a hyperscaler at $10–15B valuation — a 10–15x return from current levels. POET transforms from a speculative technology bet into the foundational photonic platform that enabled the transition from electrical to optical computing interconnects. But this scenario requires multiple independent low-probability events to all occur: POET manufacturing works, CPO adoption accelerates, POET wins over competitors, and timing aligns with POET's cash runway.
Others in Transmit the Data
Not financial advice. All scores generated via AI algorithms using public data.