POET Technologies posts 112% revenue rise on photonics-for-AI push

A $400m raise and a $50m Lumilens order position POET's optical interposer at AI infrastructure's next bottleneck: the photon.

By Disrupts Editorial Team Wed, 2 September 2026

An expansive campus features several reflective spherical buildings, rectangular structures with green roofs, ground-mounted solar panels, a prominent power pylon, roads, and parking lots, all set within a scrubland environment under bright

POET Technologies, the Toronto-headquartered photonic integrated circuit (PIC) designer listed on NASDAQ, has reported its sixth consecutive quarter of sequential revenue growth, with second-quarter 2026 revenue reaching $569,925, up 112% year on year and 13% on the prior quarter. The figures are modest in absolute terms for a NASDAQ-listed company, but the direction of travel and the capital the company has assembled around it tell a more consequential story about where AI infrastructure spending is heading next.

The headline number for capital markets is not the revenue line but the balance sheet. POET closed a $400 million registered direct offering in May, pricing units at $21, a premium to market at the time, and ended the quarter with $796.3 million in cash and short-term investments. For a company whose quarterly revenue still sits below $600,000, that war chest is extraordinary, and it signals that institutional investors are funding a position in photonics-for-AI well ahead of commercial scale.

The Lumilens deal and what it signals

The strategic substance lies in POET's supply agreement with Lumilens, a wafer-level photonic integration partnership that carries an initial purchase order of $50 million for POET Optical Interposer-based engines. The company says the arrangement could scale to more than $500 million in cumulative purchases over five years, a figure it flags as forward-looking and subject to significant uncertainty, as the release's cautionary language makes clear.

The POET Optical Interposer is a platform that co-integrates electronic and photonic devices on a single chip using wafer-level semiconductor manufacturing. In practical terms, the company's products, running at 800G and 1.6T speeds, are aimed at the optical interconnect layer inside AI clusters and hyperscale data centres, and at the chip-to-chip communication links within AI servers. That last application is the more strategically novel: as GPU clusters scale and the latency cost of electrical interconnects compounds, optical chip-to-chip links become a credible architectural answer. POET is positioning the Blazar hybrid laser, enhanced post-quarter by an exclusive component supply deal that the company says significantly improves output power, as a candidate for that emerging market.

The net loss narrowed to $11.3 million in Q2 from $12.3 million in Q1, though R&D expenditure jumped to $5.8 million from $3.1 million a year earlier, reflecting what management describes as a shift from technology development to product development. The company expects to begin shipping substantial numbers of production units for qualification in the second half of 2026.

Convergence read-across: photonics as the AI infrastructure chokepoint

The POET story sits at the intersection of two capital flows that Disrupts readers will recognise from adjacent sectors. First, the AI compute buildout: the GPU-cluster capex wave funded by hyperscalers and sovereign AI programmes has largely been about silicon and power. The next constraint is bandwidth and latency inside and between those clusters, a problem that is fundamentally optical, not silicon. That is drawing PIC specialists, fibre manufacturers, and laser companies into the AI supply chain, and institutional capital is pricing that transition early, as POET's premium-to-market offering demonstrates.

Second, the semiconductor geography angle. POET operates across Toronto, Singapore, Penang, and Shenzhen, a supply chain footprint that spans the same friend-shoring fault lines visible in the broader chip industry. The company's decision, announced in the same results release, not to redomicile to the United States despite its NASDAQ listing is a quiet but notable data point: the regulatory and tax calculus of US incorporation does not yet outweigh operational flexibility for a company with manufacturing nodes in Southeast Asia.

For cross-sector investors watching the AI infrastructure spend cascade outward from GPU silicon into optics, power, and cooling, POET's trajectory offers an early-stage read on where photonics capital is concentrating and on what terms.