Alpha SpotlightBack

POET Technologies and the AI Data Center Optical Race

Editorial Team6 min readThursday, June 25, 2026 at 1:39 PM ETBullishBullish Sentiment
POET Technologies and the AI Data Center Optical Race

Share this article

Spread the word on social media

POET Technologies and the AI Data Center Optical Race

The buildout of artificial intelligence infrastructure has turned optical interconnect from a supporting component into core data center plumbing, and the second half of June 2026 made that shift unusually visible. A wave of capacity commitments across the optical-chip supply chain, fresh demand sizing from the major research houses, and a hardening set of grid and community headwinds together reframe the backdrop that POET Technologies is selling into. What follows separates the public, company-agnostic sector picture from POET's own disclosed roadmap, and flags the material risks that any balanced read of the name has to carry.

The capacity race accelerates

The clearest new signal is that the entire upstream is racing to build optical-interconnect capacity. In mid-June, Coherent signed a letter of intent for up to fifty million dollars in U.S. CHIPS Act funding to expand its indium phosphide facility in Sherman, Texas, doubling manufacturing space and quadrupling wafer capacity, with Nvidia's Jensen Huang attending the groundbreaking. Tower Semiconductor is targeting a fivefold increase in monthly silicon photonics wafer output by the end of 2026, backed by roughly 1.3 billion dollars in long-term supply contracts for 2027, while STMicroelectronics plans to quadruple silicon photonics capacity at its Crolles facility by 2027. Coverage of the cluster, compiled by Pandaily, frames it plainly as a global capacity contest for AI data center optical capability now underway. For a company like $POET, whose recent raise is aimed squarely at scaling wafer-level photonic manufacturing, that upstream scramble reads as validation of direction rather than of any single supplier.

Demand sizing and the alignment bottleneck

On the demand side, TrendForce now sizes the AI optical transceiver market at roughly 26 billion dollars in 2026, with North American hyperscale traffic sustaining annual growth above thirty percent. Just as important for the technology narrative, the same analysis flags that supply of critical optoelectronic chips, including electro-absorption modulated lasers and continuous-wave lasers, remains tight, and that high-precision optical alignment is itself a limiter on scalable production. That alignment constraint is the crux of the photonics manufacturing problem, and it is precisely the bottleneck that wafer-scale, alignment-free approaches are designed to attack. The market is, in effect, naming the problem that the newest integration platforms claim to solve.

Capex, the Nvidia engine, and the CPO clock

The macro spend continues to compound. Hyperscalers are expected to commit well over 700 billion dollars to data centers and AI infrastructure in 2026, and Nvidia disclosed at its spring developer conference that its AI compute backlog had swelled from roughly half a billion dollars a year earlier to a solid one trillion. Microsoft added to the run of announcements on June 22 with a new campus in Pecos, Texas, expanding its global capacity by about two gigawatts. The competitive clock matters too: as SemiEngineering reported from this year's conferences, Nvidia now places its scale-up co-packaged optics at 2028, with copper persisting in scale-up in the near term. That timeline defines the window in which deployable pluggable and interposer-based optics can win sockets ahead of co-packaged optics reaching volume.

Headwinds that a balanced read cannot ignore

The buildout is increasingly contested, and the tone of the sector story has shifted with it. Federal energy regulators have ordered major grid operators to fast-track data center interconnections while making the facilities responsible for their own connection costs, against a forecast that data center electricity demand will nearly triple through 2035. Public sentiment has turned sharply negative, with recent polling showing roughly seven in ten Americans opposed to data center construction in their own communities, and analyses noting that a large majority of newly planned U.S. sites sit in drought-affected areas. None of this derails the optical thesis, but it does change the risk frame around the broader buildout that funds it.

POET's disclosed roadmap

Against that backdrop, POET's own data center story rests almost entirely on figures the company has itself released. The anchor is the Lumilens agreement: a strategic supply and joint development arrangement built on an Electrical-Optical Interposer platform, opened with an initial fifty-million-dollar purchase order and a framework that the parties have said could exceed 500 million dollars in cumulative purchases over five years if milestones are met. POET has publicly laid out an operational cadence to match, guiding to more than thirty thousand optical engine modules shipped in 2026, engineering samples of 800G and 1.6T transceivers by year-end, and volume production targeted for 2027. The company's 400-million-dollar registered direct offering is earmarked to roughly decuple manufacturing capacity in support of that ramp. Prior POET coverage and the underlying research notes are tracked on the StockAlpha.ai POET research page, with the optical-interposer thesis developed in more depth in this StockAlpha.ai analysis. Any forward-looking framing should stay tethered to POET's own language; the five-year figure is a milestone-dependent framework, not booked revenue.

Material facts and risk

The same month that sharpened POET's data center narrative also raised the disclosure bar around the name. POET faces an active securities class action covering purchases between April 1 and April 27, 2026, with a lead plaintiff deadline of June 29, 2026, filed after the stock fell roughly 47 percent in a single session when the company disclosed the cancellation of all Celestial AI purchase orders over an alleged confidentiality breach. In our opinion if the multiple law firms have not found a lead plaintiff by now, it’s unlikely they will. We also believe there is a low probability if they did find a plaintiff it would go anywhere considering the stock rallied to $25 from that low & the breach is a grey area as the CFO only reiterated public info. The most recent filings name the chief executive and chief financial officer as individual defendants. The cancellation traces back through Marvell, which had acquired Celestial AI and issued the cancellation notice in late April. On the structural side, a shareholder vote to redomicile to Delaware, expected in late June, would remove the passive foreign investment company classification that has deterred U.S. institutions, and the long-tenured chief financial officer has announced his retirement with a successor search underway. Dilutive overhang from the recent raise remains until orders convert to revenue. A complete picture of POET's data center opportunity includes these facts rather than setting them aside.

POET TechnologiesAI data centeroptical interconnectsilicon photonicsoptical interposerLumilens1.6T transceivershyperscaler capexco-packaged opticsEOI platform

Trade this headline in Alpha Contests.

Free practice contests — earn Alpha Coins
Enter a Contest

Discover more insights

Get curated market analysis and editorial deep dives from our team. The stories that matter most, examined from every angle.

More Spotlight Articles

Compensation Disclosure: Jefferson Equity Derivatives & Intelligence LLC has been compensated for the promotion of POET Technologies Inc. (NASDAQ: POET). POET Technologies Inc. paid three hundred twenty thousand dollars ($320,000) USD Cash for a marketing program (March 1, 2026 through December 31, 2026). As a result, our opinion is neither unbiased nor independent. The publishers hold no securities of the Company. This marketing may increase investor awareness, trading volume, and share price, which may be temporary. Full disclaimers.

Disclaimer: StockAlpha.ai content is for informational and educational purposes only. It is not personalized investment advice. Sentiment ratings and market analysis reflect data-driven observations, not buy, sell, or hold recommendations. Always consult a qualified financial advisor before making investment decisions. Past performance does not guarantee future results.