Alpha BreakingAre These the Best AI Infrastructure Stocks Right Now? The Optical Stocks To Watch Today
Share this article
Spread the word on social media
In the stock market today, a familiar pattern is reasserting itself: optical interconnect names are outperforming the broader tape without a single headline to explain it. Applied Optoelectronics (NASDAQ: AAOI), Coherent Corp (NYSE: COHR), Marvell Technology (NASDAQ: MRVL), and POET Technologies (NASDAQ: POET) are all moving with the kind of quiet conviction that tends to precede a proper institutional rotation. Traders scanning for stocks to watch today found the optical sector sitting at the top of their momentum screens — and looking closer, the reasons aren't hard to find.
Before digging into the individual names, some macro context: futures edged higher this morning on reports that the U.S., Iran, and a group of regional mediators are discussing the terms of a potential 45-day ceasefire. Oil is down on the news. For a sector that has struggled under the weight of geopolitical risk premiums and energy cost uncertainty, easing tension in the Middle East could be read as a quiet tailwind for capital expenditure planning at the hyperscalers. When the fog of geopolitics clears, data center infrastructure budgets tend to get signed.
But this rally started long before this morning. The optical interconnect complex has been building toward something structural since the Optical Fiber Communication Conference in mid-March, and anyone running a list of AI infrastructure stocks to buy has had these names in front of them for weeks. The macro news today may have provided the opening — what's underneath it is a supercycle that is still early in its adoption curve.
The 1.6T supercycle: why 2026 is different from 2000
The comparison to the fiber-optic bubble of 2000 is the first objection every skeptic reaches for. It's worth addressing directly. The late 1990s boom was built on projected demand that didn't materialize for another decade. The demand driving optical stocks today is real, contracted, and already showing up in quarterly results.
Hyperscaler capital expenditure — Alphabet, Meta, Microsoft, and the cloud infrastructure arms of Amazon and Oracle — has moved past the GPU procurement race and is now attacking the bandwidth problem. NVIDIA's Blackwell and Rubin platforms require exponentially higher interconnect density than their predecessors. At 1.6 terabits per second, copper cable loses so much signal integrity and generates so much heat that it cannot reliably bridge GPU racks — what some in the industry have called the "2-meter limit." What replaces it is fiber and photonics — and the ramp is happening now, not in five years.
Global shipments of 800G-and-above transceivers are forecast to jump from roughly 24 million units in 2025 to nearly 63 million in 2026. Volume 1.6T shipments are expected to ramp through the second half of 2026. Indium phosphide lasers — the critical component inside these transceivers — are reportedly sold out through 2027 and into 2028. When you're sold out two years forward, you don't need a catalyst. You need capacity.
For investors running screens of AI stocks to buy in 2026, this structural shift is the context that makes the optical sector interesting as a multi-year theme rather than a single-quarter setup. The GPU is no longer the central bottleneck in this analysis — the question shifts to what connects them, and which companies sit at that layer of the stack. That's the analytical case for looking at this sector beyond a simple momentum trade.
Breaking down the names moving today
AAOI stock — the hyperscaler bet with domestic manufacturing leverage
Applied Optoelectronics has been the loudest mover in the optical complex since February. The company reported record Q4 2025 revenue and issued guidance projecting a roughly 3.5x jump in data center revenues for calendar 2026. Management has set a target of more than $1 billion in revenue for the year, supported by a wave of contracted orders that have provided unusual visibility for a company of its size.
The most significant recent development was a volume order exceeding $200 million for 1.6T data center transceivers from a long-term hyperscale customer, with shipments scheduled for the third and fourth quarters of this year. Additional 800G orders totaling north of $53 million from major customers are expected to begin shipping in Q2. AAOI is also ramping its Sugar Land, Texas facility — a 210,000-square-foot production site targeting 500,000 combined 800G and 1.6T units per month by year-end. In an era of tariff scrutiny and supply chain sovereignty concerns, the domestic manufacturing angle is not incidental. It appears to be a strategic asset that hyperscalers are actively rewarding.
OFC Declares an Optical Supercycle — Photonics Repricing Mee
The stock is not cheap on traditional metrics, and EBIT margins remain negative as the company invests heavily in its capacity ramp. But the revenue multiple looks compressible if execution meets guidance, and the production ramp timeline keeps the company in active coverage across the optical sector.
COHR stock — pricing power in a sold-out market
Coherent holds roughly 25% market share in optical transceivers and reported fiscal year 2025 revenue of $5.81 billion, up 23% year-over-year. It is NVIDIA's named collaborator on silicon photonics for the Spectrum-X Ethernet platform, and a major supplier of continuous-wave lasers. In any supply-constrained market, the incumbent with the deepest customer relationships and the most production capacity tends to exercise outsized pricing power. Coherent is that incumbent.
The Rosenblatt analyst covering the sector noted at OFC that indium phosphide lasers are sold out through 2027 and into 2028. Coherent is one of a small number of companies with the fab capacity and process expertise to build new supply at scale. It is investing in capacity for scale-up co-packaged optics, the technology expected to succeed pluggable transceivers as the dominant architecture for AI networking — potentially through the end of this decade.
COHR stock is not a discovery — it's an established institutional name. What may be underappreciated is how durable the pricing environment could remain for the next two to three years given the structural laser shortage. At OFC, the message from suppliers was consistent: demand is pulling harder than supply can follow.
MRVL stock — the DSP toll booth
Marvell Technology is the connective tissue between optical hardware and the compute layer. Its digital signal processors manage the optical signals traveling through fiber-based data center fabrics. As the industry moves from 400G to 800G and then to 1.6T, every upgrade cycle generates incremental DSP content — more lanes, higher speeds, more complexity for Marvell to solve.
The most strategically significant recent development for MRVL is its agreement to acquire Celestial AI for up to $5.5 billion, announced in December 2025. Celestial AI's technology targets the co-packaged optics architecture — integrating photonic connectivity directly onto the processor package rather than plugging in from the front of a switch. If CPO becomes standard, as the industry broadly expects by the late 2020s, Marvell's Celestial acquisition could look prescient. It's the kind of move that positions a company not just for the current transceiver cycle but for the generation after it.
MRVL's positioning reflects scale across every networking upgrade cycle and a balance sheet commitment to the next architecture through the Celestial AI acquisition. For those analyzing the AI infrastructure stack as a multi-year theme, Marvell represents the DSP layer — recurring content growth with large-cap liquidity and an embedded bet on where the architecture is going.
The platform layer: where the valuation debate is still open
The names above represent the established positions in this trade. AAOI, COHR, and MRVL have been on institutional radars for at least a year. Earnings beats, order announcements, and conference presentations have all served to validate the thesis that is now broadly reflected in their valuations.
The more interesting question — the one that tends to generate the most durable returns in a secular infrastructure cycle — is whether there is a layer of the stack that provides similar exposure but has not yet been priced for the cycle it's entering.
MRVL: AI & Networking Growth vs Valuation
That question tends to lead back to the platform layer. Transceivers are a component. What underlies the transceiver is the optical engine — the integrated assembly of lasers, modulators, and photodetectors that converts electrical signals to light and back again. The optical engine is where the real integration challenge lives, and where the cost structure of the industry is still being worked out. Pluggable transceivers are a transitional technology. Co-packaged optics will eventually bring the photonic layer closer to the processor. The companies building the platforms that make that integration possible are competing for a position that could be worth substantially more than the current component market.
Anyone who has been tracking the optical interconnect space over the past eighteen months has encountered POET Technologies in that context.
POET stock — the platform name that keeps coming up
POET Technologies is not a new idea to anyone who has done serious work in this sector. The company's Optical Interposer platform — which integrates lasers, modulators, drive electronics, and coupling structures onto a single chip using a wafer-scale process — has been a consistent point of discussion at industry conferences and in analyst coverage for the better part of two years. Its Teralight optical engine line won Lightwave's Innovation Award in February 2026, scoring a 4.5, among the highest in the competition's history. At OFC in March, POET demonstrated both its Blazar external light source and its Starlight eight-channel compact engine, and accepted the award on the conference floor. These are not the actions of a company still in research mode.
The commercialization timeline has been advancing. POET has announced a 1.6T 2×DR4 optical transceiver module in joint development with Lessengers targeting AI clusters and hyperscale data centers, with samples targeted for Q2 2026. A strategic collaboration with LITEON Technology to co-develop next-generation optical communication modules is underway, with prototype targets in late 2026 and high-volume production targeted for 2027. The company has also partnered with Sivers Semiconductors on combined DFB laser plus Optical Interposer light engine sub-systems, targeting production readiness by end of 2026.
Q4 2025 results showed EPS of $(0.32), a meaningful improvement from $(0.50) the prior year. The company completed a $150 million registered direct offering in January 2026, which leaves it with significant capital to execute on its commercialization roadmap. For a company at this stage of the product cycle, the balance sheet matters as much as the technology.
The bull thesis on POET has always rested on the Optical Interposer's ability to reduce the cost and complexity of optical engine assembly relative to conventional manufacturing. The technology has been demonstrated across multiple configurations at the bench level. The open question is whether POET can execute at the volumes and timelines the partnerships are implying, and whether the competitive landscape allows it to convert demonstrated capability into contracted revenue at the scale the market would need to see to re-rate the stock.
For analysts tracking the optical interconnect space, POET sits in a category that invites scrutiny on both sides: credible enough within the photonics community to be taken seriously, early-stage enough in its commercial ramp that the valuation debate remains genuinely open, with execution on current timelines representing the central variable.
The risks are material and worth stating plainly. POET has a history of equity issuance to fund operations — the January 2026 offering was its most recent, and dilution remains an ongoing consideration for existing shareholders. Revenue is minimal and concentrated in early-stage production orders, meaning a single customer delay or partnership setback has an outsized impact on the commercialization timeline. Every production target cited above — Q2 2026 samples, late 2026 prototypes, 2027 high-volume production — represents company-stated guidance, not contracted delivery obligations. And the competitive landscape is not benign: every major optical and semiconductor company with substantially more capital is working on platform-level photonic integration. POET's Optical Interposer competes in a space that well-capitalized incumbents are also actively pursuing.
What is the market missing — and what are the open questions?
The optical sector rally in the stock market today is not a mystery once you understand the structural context. The 1.6T supercycle is real. The laser shortage is real. The hyperscaler capex commitments are contracted. The AI infrastructure trade has rotated from the GPU layer toward the interconnect layer, and the stocks moving today — AAOI, COHR, MRVL — are the established names that institutional money has already validated.
The open questions are more interesting than the answers the market has already priced.
At what speed does co-packaged optics displace pluggable transceivers? If the CPO timeline compresses, the implications for every pluggable module maker are significant — and the implications for platform-layer companies building the photonic integration capabilities that CPO requires could be equally significant in the other direction.
Which layer of the optical stack captures the most margin in a scaled, commoditizing market? Component makers face the historical pressure of transceiver price compression as volumes scale. Platform companies that can offer cost reduction at the optical engine level could be positioned differently.
And perhaps most practically: when the market eventually gets around to pricing the 2027 and 2028 production ramps that are already implied by current partnerships and order books, which names on the optical watchlist will look obvious in hindsight — and which ones are already fully priced for what they can deliver?