Bullish Sentiment
Share this article
Spread the word on social media
The thesis in one paragraph
$POET Technologies has spent more than a decade building a single, contrarian idea: that photonics can be manufactured the way semiconductors are. In 2026 that idea is colliding with the largest infrastructure build-out in technology history. Artificial intelligence has turned optical interconnect from a niche telecom concern into the central bottleneck of the data center, and POET sits at the exact intersection of the two trends the industry now cares about most: higher-speed optical engines (800G to 1.6T to 3.2T and beyond) and the high-power light sources that feed them. The company is still pre-scale, and the risks are real and named below. But the gap between what POET is building and what the AI complex needs has rarely been narrower, and a fortified balance sheet now gives management the runway to close it.
Why optics is suddenly the most important problem in the data center
For most of computing history, the hard problem was the processor. In the AI era, the hard problem is moving data between processors. Training and serving frontier models means lashing tens of thousands of accelerators together into a single fabric, and at those scales copper runs out of reach, bandwidth, and power budget. The answer is light, and the industry is migrating there at remarkable speed.
The clearest signal came from the top of the market. At GTC 2025, NVIDIA introduced its Spectrum-X and Quantum-X silicon-photonics switches, a watershed moment that put the dominant AI infrastructure provider squarely behind co-packaged optics (CPO), the approach of integrating optics directly alongside the switch or accelerator die to shorten electrical paths and slash power. Broadcom, the leading switch-ASIC supplier, has pursued a complementary CPO path with its Bailly platform. When the two companies that define AI networking both commit to optical integration, the rest of the ecosystem follows.
The market math is staggering. Independent research pegs the co-packaged optics market for AI data centers at roughly $9 billion in 2025, with projections of well over $100 billion by the mid-2030s at compound growth rates near 37 percent. Inside that shift, the 1.6T data-rate tier is forecast to be the fastest-growing segment as next-generation training clusters move into volume deployment from 2026 onward. One widely cited LightCounting forecast sees more than 125 million 1.6T DR8-class units shipping between 2027 and 2031. This is not a story about whether optical interconnect grows. It is a story about who supplies it.
What POET actually makes: the Optical Interposer
POET designs and develops photonic integrated circuits, light sources, and optical modules for AI systems and hyperscale data centers. Its differentiator is the patented POET Optical Interposer, a platform that co-integrates electronic and photonic components onto a single chip using conventional, wafer-scale semiconductor techniques. Management calls the approach the “semiconductorization of photonics,” and the engineering implications are the heart of the bull case.
Hybrid integration with no wire-bonds. Electrical and optical components sit together on one interposer, eliminating the wire-bonds that introduce parasitics, bandwidth limits, and reliability risk in conventional photonic assemblies.
Monolithic passives. Multiplexers, demultiplexers, waveguides, micro-mirrors, and V-grooves are integrated directly into the platform, moving photonics toward a true system-on-chip architecture.
Wafer-scale assembly. By using semiconductor-style wafer processing rather than one-device-at-a-time active alignment, the platform targets higher yield, lower cost, and genuine volume manufacturability — the qualities the AI build-out demands and that hand-assembled optics struggle to deliver.
Form-factor and speed agnostic. The same integration approach scales across 800G, 1.6T, and 3.2T-plus, and across pluggable, near-packaged optics, and co-packaged optics. That breadth lets POET serve customers wherever they are on the migration curve instead of betting the company on one form factor.
From slideware to a booked pipeline: the commercial inflection
For years the knock on POET was that it had impressive technology and not enough customers. 2026 is the year that critique started to break down.
Lumilens — the marquee deal. On May 14, 2026, POET announced a strategic supply and joint-development agreement with Lumilens to advance wafer-level photonic integration for frontier AI infrastructure. The agreement includes an initial $50 million purchase order for Electrical-Optical Interposer-based engines, within a framework management believes could exceed $500 million in cumulative purchases over five years. For a company whose entire 2025 revenue was roughly $1 million, a single initial order at this scale is a step-change in commercial visibility.
Lessengers — a 1.6T product on the clock. POET is jointly developing a 1.6T 2×DR4 optical transceiver module for AI clusters and hyperscale networks, combining its Optical Interposer engines with Lessengers’ Direct Optical Wiring. Sample units were targeted for the second quarter of 2026, putting POET directly into the fastest-growing speed tier just as it ramps.
LITEON — scale and credibility. In March 2026 POET entered a strategic collaboration with LITEON Technology, a global leader in optoelectronic components, to co-develop next-generation optical modules on the Interposer platform, with prototypes targeted for late 2026 and high-volume production anticipated in 2027.
A widening roster. Those headline deals sit on top of a 1.6T receiver launch with Semtech, a 100G bidirectional engine with NTT Innovative Devices aimed at mobile front-haul and 5G, and a 3.2Tbps co-development with Quantum Computing Inc. The pattern matters: POET is positioning the Optical Interposer as a common platform that drops into many partners’ product families rather than a single point product.
Underpinning the partnerships is a concrete output target. Management has guided to shipping more than 30,000 optical engine modules in 2026, with 800G and 1.6T engineering samples due by year-end and volume production targeted for 2027. The unit numbers are still small against the eventual opportunity, but they convert the narrative into something measurable.
A pipeline that now has a number on it.
At its Annual General Meeting on June 26, 2026, with results released June 30, POET disclosed that it has more than ten active customer engagements that combined are expected to exceed $100 million in future annual revenue. That roster includes the Lumilens agreement, a previously announced production order for 800G optical engines, and a significant external-light-source NRE agreement from a new customer. After years of being a technology story in search of customers, POET has now attached an aggregate revenue figure to its commercial pipeline — a meaningful shift in how the company can be evaluated.
Blazar and the light-source opportunity
If the Optical Interposer is POET’s platform, Blazar is its most provocative product. Blazar is a highly integrated hybrid external-cavity laser designed to power co-packaged optics and high-bandwidth, chip-to-chip optical links. At OFC 2026 in Los Angeles — from POET’s largest booth in its history — the company performed its first public live demonstration of Blazar, alongside its next-generation Starlight light-source engine. Just twelve months earlier, Blazar had been shown only privately. That move from closed-door prototype to public live demo in a single year is a real maturity milestone, not a press-release flourish. POET’s 1.6T Teralight engine separately earned a 4.5 score from the Lightwave Innovation Reviews, among the highest the program has awarded.
The strategic logic behind Blazar is supply. AI networks need enormous quantities of high-power, narrow-linewidth lasers, and that supply is concentrated and constrained. The dominant indium-phosphide laser suppliers are capacity-limited, and AI vendors have been pre-paying to lock in allocation. Conventional distributed-feedback laser arrays also fight physics as power and channel counts rise: yields fall, packaging complexity grows, and tight wavelength control gets harder. Blazar’s external-cavity architecture, with wavelength selection handled on the Interposer rather than the gain chip, is designed to attack exactly those constraints — a wafer-scale, multi-wavelength, high-power light source for a market that is short of light sources. Blazar’s revenue contribution is a later-decade story, but it is the kind of optionality that does not show up in a near-term model. At its June 2026 AGM, management reiterated that Blazar remains on schedule for deployment at scale in 2028.
Credibility: a balance sheet built for the ramp
Vision and supply constraints mean little without the capital to build capacity. Here POET’s 2026 looks very different from its past. By the company’s own account at its June 2026 Annual General Meeting, POET raised $830 million in equity capital over the trailing twelve months, with the potential for up to an additional $661 million if its outstanding warrants are exercised. The most recent tranche was a US$400 million non-brokered registered direct offering that closed on May 18, 2026 with a single institutional investor — 19,047,620 shares plus an equal number of three-year warrants struck at a 25 percent premium, priced at a slight premium to market rather than the discount distressed issuers accept. Capital raised over a year is not the same as cash in the bank today, but it has transformed a company that spent much of its history capital-constrained into one funded to build.
Equally important is what that capital is for. Management has guided to an initial deployment of roughly $50 million in the second half of 2026 to purchase manufacturing equipment, and is targeting capacity of up to one million units per month by the end of 2027 — a more than tenfold increase from current production. The balance sheet also lets POET pursue long-term supply agreements for critical components at a time when the industry faces severe shortages, turning a potential bottleneck into a relative advantage. A single large investor writing a nine-figure check at a premium, with warrants struck well above market, is a meaningful third-party vote of confidence in that plan.
The team is being built to match. In May 2026 POET appointed Dr. Sandeep Kumar as Chief Operating Officer, bringing more than eighteen years at Silicon Labs, most recently as Senior Vice President of Worldwide Operations. His mandate is squarely operational: stand up high-volume manufacturing in Penang, Malaysia, and turn design wins into shipped product. For a company whose entire investment case now rests on execution, hiring a career operations leader is the right signal. POET also told shareholders it intends to add roughly fifty employees to its global workforce of about 115 in the coming months, scaling the organization to match the production ramp.
Removing a structural overhang: the U.S. redomicile
One reason institutional U.S. capital historically kept POET at arm’s length was its Canadian domicile and the associated Passive Foreign Investment Company (PFIC) complications for U.S. holders. POET has moved to eliminate that friction directly, announcing its intention to redomicile in the United States so it will no longer be treated as a foreign corporation, a step that would remove future PFIC classification risk. The plan, approved by the board and put to shareholders in mid-2026, also pairs with QEF-election paperwork the company provided for the 2025 tax year. If approved, the change lowers a real barrier between POET and the U.S. institutional investors closest to the AI capital flows it is chasing.
The other side of the ledger: risks investors must weigh
A credible bull case is honest about what can go wrong, and POET carries a meaningful risk profile that every reader should weigh independently.
It is still pre-scale. First-quarter 2026 revenue was about $0.5 million and full-year 2025 revenue roughly $1 million, against a valuation that prices in years of future growth. Reported net losses have been large — about $63 million in 2025 — though a substantial portion is non-cash, including warrant remeasurement; first-quarter 2026 operating cash outflow was a more modest $8.8 million. The story depends on contracts converting into recognized revenue.
Dilution. POET reported roughly 172.6 million shares outstanding at the time of its June 2026 AGM, and the capital raised over the past year came with warrants that could bring in up to an additional $661 million if exercised — implying further share issuance ahead. The capital strengthens the balance sheet, but existing holders bear the dilution.
Execution and concentration. Most of the pipeline is partnership and development work that still has to clear sampling, qualification, and volume production — the steps where photonics companies most often stumble. Historically, several of POET’s partnership announcements drew muted or negative one-day market reactions, a reminder that investors want shipped revenue, not just signed deals.
Active litigation. POET faces a securities class action covering purchases between April 1 and April 27, 2026, with allegations relating to PFIC disclosures and an executive’s confidentiality obligations; CEO Suresh Venkatesan and CFO Thomas Mika are named as individual defendants, and the lead-plaintiff deadline is June 29, 2026. To be fair the success rate of these types of lawsuits are below 5%, but investors should always be aware of these issues. The odds on this one have dropped significantly as a lead plaintiff has yet to be found before the cutoff date of June 29th to our knowledge.
Formidable competition. POET is competing against far larger, well-capitalized incumbents in lasers and switching, plus the in-house optical efforts of the very hyperscalers it hopes to supply.
Bottom line: a high-conviction, high-variance way to play AI optics
$POET is not a story for capital that cannot tolerate volatility. It is a pre-revenue platform company whose value will be decided by execution over the next several quarters. But the setup is rare: a differentiated, patented manufacturing approach; a public roadmap spanning 800G to 3.2T-plus and every relevant form factor; a marquee supply agreement with a framework that could exceed half a billion dollars; a light-source product attacking a genuine industry shortage; $830 million in equity capital raised over the past year to fund the build-out; an operations-first leadership upgrade; and a corporate-structure change designed to open the door to U.S. institutional capital — all pointed at the single largest infrastructure build-out in technology.
The catalysts to watch are concrete and near-term: the outcome of the redomicile vote; delivery of 800G and 1.6T samples by year-end; the 2026 production ramp in Malaysia; conversion of the Lumilens order and the broader $100 million-plus engagement pipeline into recognized revenue; and the trajectory of cash burn against the capital raised. Clear those bars, and the gap between POET’s market value and its opportunity could close quickly. The thesis is straightforward: the AI build-out runs on light, light is supply-constrained, and POET has spent a decade learning to manufacture it like a chip.